Market Overview

Wednesday evening Nvidia reported fiscal Q2 results that beat on revenue, earnings, and data centre sales, with Q3 guidance above consensus. Shares initially fell about 2% before reversing during the earnings call, closing the after-hours session more than 4% higher as management laid out a materially stronger long-term outlook.

Thursday delivered the reaction. Nvidia gained 8.7%, its largest single-day advance since April 2025, and the iShares Semiconductor ETF rose 2%. Software did better still. The iShares Expanded-Tech Software Sector ETF surged 7.4%, carried by Salesforce, which jumped 23% on solid quarterly results and an expanded Anthropic partnership. Marvell also reported record revenue and raised both its fiscal 2027 and fiscal 2028 outlooks.

Friday reversed part of it. Warsh delivered his first Jackson Hole address as Fed chair, and bond yields rose as September hike odds solidified. By late afternoon, futures priced a 58% chance of a quarter-point increase at the September meeting, up from 35% the day before. Rate-sensitive technology led the decline. The Nasdaq Composite fell 0.5%, the S&P 500 fell 0.25%, and Nvidia sank 4.6%, erasing much of the prior session’s gain.

On the data front, the updated GDP print showed real output growing at a 1.5% annualised rate from Q1 to Q2. With 97% of the S&P 500 now reported, second-quarter earnings growth stands at 52%, the highest since the pandemic rebound in Q2 2021.

For the week: the Dow gained 0.5%, the S&P 500 gained 0.5%, and the Nasdaq Composite rose 0.9%.

Week Ahead

Monday, August 31

CrowdStrike opens its three-day Fal.Con conference, themed this year around securing the AI revolution, with CEO George Kurtz among the speakers. Broadcom begins its three-day VMware Explore event, focused on using VMware Cloud Foundation to modernise private cloud infrastructure for AI-ready enterprise workloads.

Tuesday, September 1

John Ternus takes over as Apple CEO, with Tim Cook moving to executive chairman after fifteen years in the role. Dell reports, with options pricing a 10% share price swing; the call should be read for AI server demand, order trends, margins, and supply availability. Intel is worth watching alongside it, since Dell accounts for roughly 19% of Intel’s revenue, and the two have traded in the same direction after Dell’s results 88% of the time in recent years.

Palo Alto Networks, MongoDB, Medtronic, and Nio also report, with Credo priced for a double-digit move.

Wednesday, September 2

Broadcom reports, the most consequential print of the week for the AI infrastructure trade and the company whose June guidance sent it down 12% on record AI revenue, as covered in Issue 53.

Snowflake and Hewlett Packard Enterprise also report, giving reads on data platform consumption and AI server demand, respectively.

Thursday, September 3

Tesla holds an invite-only Cybercab launch event in Austin. The Cybercab is the purpose-built robotaxi with no steering wheel or pedals, and reports suggest riders may be able to request one through the existing Robotaxi app.

Zscaler, Ciena, Samsara, Docusign, and Lululemon report.

Friday, September 4

The August jobs report lands at 8:30 a.m. Economists expect 55,000 payroll additions, a 4.2% unemployment rate, and 0.2% monthly wage growth. After July’s 23,000 decline and with September’s hike odds at 58%, this is the number that decides whether Warsh’s committee has cover to move.

Nvidia’s 70 Percent

The quarter was excellent and largely beside the point. Revenue of $96.22 billion grew 106% year-on-year against a $92.29 billion estimate. Non-GAAP EPS of $2.22 beat $2.09. Data centre revenue of $89.0 billion grew 117% year-on-year and 18% sequentially, ahead of the $85.83 billion expected. Gross margin held at 75.0%. Third-quarter guidance of $108 billion, plus or minus 2%, came in above the roughly $104.2 billion consensus and assumes no data centre compute revenue from China at all.

The number that moved the stock came on the call. CFO Colette Kress said Nvidia expects fiscal 2028 revenue to grow roughly 70% year-on-year. The market had been modelling around 45%. That is a 25 point gap on a revenue base approaching $400 billion, and it arrived alongside an acknowledgement that supply remains constrained.

Two disclosures underneath the guidance deserve attention. Vera Rubin began shipping this month and is expected to see the fastest product ramp in the company’s history, with Nvidia estimating that every gigawatt of Rubin compute deployed represents roughly $40 billion of revenue opportunity. Blackwell demand remains strong while Rubin is already entering its ramp, which removes the transition gap that usually accompanies a generational shift. Separately, supply and capacity commitments jumped from $119 billion last quarter to $279 billion, largely to lock in memory and manufacturing capacity, with supply expected to remain the bottleneck at least through fiscal 2028.

That $279 billion figure is the most informative number in the release. A company does not commit $160 billion of incremental capacity in a quarter unless it has visibility on demand it cannot currently serve. It is also, directly, the memory supercycle this newsletter has tracked since Issue 49 showing up as a cost line at the largest buyer in the industry. Q3 gross margin is guided to 74.0%, about a point lower sequentially, and memory costs are the reason. The AI trade and the memory trade are now the same trade viewed from opposite sides of an invoice.

Claudeforce and the End of the Disruption Trade

Salesforce rose 23% on Thursday, its largest single-day gain since August 2020, and dragged the software sector up with it. The iShares Expanded-Tech Software Sector ETF gained 7.4%, outperforming semiconductors on the day Nvidia rose 8.7%.

The results were solid. The reason for a 23% move was the partnership. Salesforce announced an expanded collaboration with Anthropic that gives users access to Salesforce data and functions from within the Claude chatbot, branded Claudeforce. For most of this year the bear case on enterprise software has been that AI agents would make seat-based subscription software obsolete, that if a frontier model can do the work, the vendor sitting between the customer and the work becomes a toll booth without a road.

Claudeforce is an answer to that. Deutsche Bank’s Brad Zelnick wrote that the deepened partnership helps alleviate concerns that frontier models will diminish the need for Salesforce, instead positioning its data, applications, and governance as the trusted foundation through which customers access those models. The distinction matters. Salesforce is not competing with Anthropic for the reasoning layer. It is supplying the customer data, permissions, and audit trail that make the reasoning layer usable inside a regulated enterprise.

Palantir’s US commercial revenue grew 149% by selling AI sovereignty, the ability to run models against proprietary data without handing that data to a third party. Salesforce is selling a version of the same thing to a different customer base. The three-layer framework this newsletter set out in Issue 62 separated infrastructure, platform, and software. What this week established is that the software layer is not being disrupted from above. It is being repositioned as the governance layer, and the market repriced that repositioning by 23% in a session.

Marvell Earnings

In Issue 59, Marvell appeared in this newsletter as one of the worst casualties of the semiconductor bear market, down roughly 33% from its peak. Six weeks later it reported record revenue of $2.739 billion, up 37% year-on-year and 13% sequentially, and raised its outlook twice over.

Fiscal 2027 revenue guidance moved from approximately $11.5 billion to roughly $12 billion. Fiscal 2028 moved from approximately $16.5 billion to approximately $18 billion, a $1.5 billion raise on a horizon more than a year out, delivered a single quarter after the prior number was set. Data centre revenue is now expected to grow more than 60% year-on-year in fiscal 2028. Third-quarter guidance is $3.15 billion plus or minus 5%, with non-GAAP EPS of $1.05 to $1.15.

The mechanism behind the raise is the interesting part. Marvell filed an 8-K disclosing an expanded commercial agreement and an associated warrant with a key hyperscaler, covering custom programs already in execution as well as new design wins and future potential programs. CEO Matthew Murphy described the scope as spanning inference accelerators, storage controllers, network interface cards, memory interface controllers, and near-memory compute, and called the total envelope game-changing. A warrant is equity. What the structure does is align a customer’s economic interest with the supplier’s success over a multi-year design cycle, which is a stronger commitment than a purchase order and a weaker one than an acquisition.

What separates rewarded AI capital from punished AI capital is the presence of a contractual demand anchor. Micron used strategic customer agreements, Microsoft used remaining performance obligations, and CoreWeave used backlog. Marvell has now used warrants. The forms keep changing. The principle does not. Two caveats belong in the same paragraph: gross margin steps down next quarter on custom mix, and management flagged pervasive industry-wide supply constraints, mitigated by roughly $1 billion of capacity prepayments to suppliers this fiscal year.

Bitcoin’s Bounce

Bitcoin has risen approximately 23% over the past thirty days. It also remains roughly 38% below its October 2025 all-time high. Both of those things are true at once, and holding them together is the only honest way to read the move.

The near-term case rests on flows and regulation. The iShares Bitcoin Trust took in roughly $1.5 billion over the past week, which suggests the rally has genuine buying behind it rather than leveraged positioning. The catalyst is the CLARITY Act, which would finally settle which regulator oversees digital assets. That question sounds procedural and is not. Large banks and asset managers have stayed largely on the sidelines because compliance ambiguity makes allocation difficult to justify to a risk committee. Remove the ambiguity and the institutional buyer base widens. The caution alongside that is straightforward: Bitcoin failed to hold above $80,000, and rallies that stall at a round number tend to see profit-taking before they see continuation.

The more interesting case has nothing to do with crypto adoption. It is a bet against the dollar, and it runs directly through the story this newsletter covered in Issue 64. The Treasury is increasing buybacks of longer-dated debt while the government borrows at its heaviest pace in twenty-five years. Total public debt has crossed $40 trillion, net interest costs are approaching $857 billion for the fiscal year to date, and there are growing odds of Japanese investors reducing US bond holdings. That combination points toward currency weakness, and currency weakness is historically good for assets with fixed or constrained supply. Gold has been telling the same story, having posted five consecutive weekly gains through late August.

What makes this argument unusually robust is that it also works in reverse. If the fiscal scenario is avoided and the Fed eventually returns to cutting rates, the dollar weakens through a different mechanism, Treasury yields fall, and capital rotates toward risk assets. Both paths are dollar-negative. For long-term investors the takeaway is not a recommendation on Bitcoin, which sits well outside this newsletter’s fundamentals-first framework and carries volatility that most portfolios should respect. It is that the debasement trade is now being expressed simultaneously in gold, in Bitcoin, and in the term premium on the long bond. Three different markets are pricing the same fiscal concern. That is worth noticing regardless of whether you own any of them.

Closing Thoughts

Nvidia raised its fiscal 2028 growth expectation by roughly 25 percentage points. Marvell raised fiscal 2028 revenue by $1.5 billion. Salesforce added nearly a quarter of its market value in a session on evidence that AI makes it more necessary rather than less. Second-quarter S&P 500 earnings growth came in at 52%, the strongest since 2021. These are not incremental revisions. They are step changes in the cash flows being discounted.

The Bitcoin move sits on the same fraction, just from the other end. Capital rotating into fixed-supply assets is capital expressing doubt about the currency those cash flows are denominated in. Gold, Bitcoin, and the long bond term premium are all pricing a version of the same fiscal question.

For long-term investors the useful question is which side is moving faster and for how long. A quarter-point hike raises the discount rate once. A 25 point upgrade to a multi-year growth trajectory compounds. Neither observation removes the risk in the other, and a market priced for both to keep going has very little room for either to disappoint. Friday’s jobs number is the first test.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

Monday opened lower as the 30-year Treasury yield rose nearly six basis points to 5.31%, its highest since 2007. The S&P 500 fell 0.52% to 7,745.06, the Nasdaq declined 0.32%, and the Dow lost 273 points. The move was global: Canadian 30-year yields hit their highest since 2010 and German rates reached 2011 levels.

Tuesday the long bond pushed to 5.34%. Treasury announced it would at least double its long-end buyback operations, total public debt crossed $40 trillion, and the July FOMC minutes landed at 2 p.m. Yields fell sharply on the buyback news, with the 30-year down nine basis points to 5.196% and the 10-year down 5.7 basis points to 4.647%. That evening Trump announced an economic warfare campaign against Iran.

Thursday reversed all of it. The 30-year climbed back over seven basis points to 5.27%, exactly where it had been before the intervention. Oil rose 3%, Brent topped $93, and the Dow shed 704 points, or 1.32%, its worst session in three weeks. Walmart fell 9%.

Friday recovered some ground. The S&P 500 rose 0.43% to 7,674.37, the Nasdaq gained 0.43% to 26,180.45, and the Dow added 518 points to 53,277.01 after a survey showed US business activity growing at its fastest pace in more than four years. Gold reached $4,569.40, a fifth consecutive weekly gain, and bitcoin hit $77,000 in its best week in two years.

For the week: S&P 500 down roughly 1.4%, Nasdaq down 2.1%, Dow down 0.9%. The first weekly decline for all three since late July.

Week Ahead

Monday August 24

Treasury Secretary Scott Bessent holds a press conference to lay out the mechanics of the Iran economic campaign announced last week. Markets will be listening for whether secondary sanctions explicitly target China, which buys the bulk of Iranian oil.

Tuesday August 25

A quieter session, and the last one before the two events that define the week. Positioning ahead of Nvidia dominates, particularly across the semiconductor complex and the neocloud names whose revenue depends on the same hardware cycle.

Wednesday August 26

July PCE lands in the morning and Nvidia reports after the close. PCE is the Fed’s preferred inflation gauge and the last major price reading before Jackson Hole. Nvidia guided to roughly $91 billion in revenue for the quarter, excluding China compute entirely. Given that the AI capital cycle is currently carrying index-level earnings on its own, as this newsletter argued in Issue 63, Nvidia’s guidance is the most consequential single corporate disclosure of the quarter.

Thursday August 27

The Jackson Hole symposium opens in Wyoming. No policy decisions are made there, but the tone set in the opening sessions will frame how markets read Friday. The second estimate of Q2 GDP also lands.

Friday August 28

Kevin Warsh delivers his first Jackson Hole address as Fed chair. He arrives with three dissenters behind him, a long bond near two-decade highs, a labour market that shed 23,000 jobs in July, and core inflation back at its pre-war pace. He has rejected forward guidance as a policy tool. Whether he uses this platform to provide a framework anyway, or deliberately withholds one, is itself the signal.

Bond Market

On Tuesday the 30-year Treasury yield reached 5.34%, the highest since 2007. On Wednesday total public debt outstanding crossed $40 trillion. Net interest costs reached roughly $857 billion in the first nine months of fiscal 2026, and the debt is currently accelerating at about $2 trillion a year. Also on Wednesday, Treasury auctioned 20-year debt at the second-highest yield since that bond was reintroduced in 2020.

Treasury Secretary Scott Bessent responded by announcing that liquidity support buyback operations in the 10 to 30 year sectors would at least double, from $2 billion to $4 billion per operation, effective September 9. The 30-year fell nine basis points immediately. By Thursday the entire move had reversed and the yield sat back at 5.27%, precisely where it started.

A $2 billion increase per operation, in a Treasury market of roughly $32 trillion, is not a supply intervention. It is a signal. The total additional liquidity support tied to this window comes to around $14 billion, bringing maximum repurchases to approximately $83 billion. That is real money and it is not nothing, but it does not change the deficit, the issuance schedule, or the inflation path. As one strategist put it, the move buys time and signals that Treasury has tools available, without solving the underlying problem.

The significance is what the long end is pricing. A 30-year yield at 5.3%, with inflation stuck above the Fed’s target for five years and issuance rising to fund a widening deficit, is the market demanding compensation for holding duration in a fiscal environment it no longer trusts. That is the risk-free rate. Every discounted cash flow model this newsletter has discussed over the past three months, from Palantir at 141 times earnings to CoreWeave’s $104 billion backlog, runs off it. When the long end reprices, the present value of every distant cash flow reprices with it, regardless of how good the underlying business is.

The Minutes

In the July minutes, many participants judged that higher rates would likely be necessary if inflation did not decline. Some said financial conditions might not be restrictive enough to return inflation to the 2% target. Various participants read the tightening in financial conditions over the intermeeting period as reflecting strong growth and market expectations that the Fed would adopt a more restrictive stance before long. And a few, the three who voted for it, argued that hiking immediately would help forestall the need for further increases later.

In Fed minutes, that ordering is deliberate. Many outranks some, which outranks a few. What the language describes is a committee where the hawkish case had broad conditional support without a majority prepared to act on it in July. One detail matters more than the vocabulary: no member of the Board of Governors joined the dissent. All three dissenters were regional presidents. That is a meaningful split between the Washington-based governors, who are permanent voters, and the rotating regional bank presidents.

The minutes also contain something more structural. Warsh observed that reducing the meeting calendar from eight to six per year, held roughly every two months, might allow more information to accumulate between meetings and give policymakers more time to consider strategic issues. No decisions were made. But a chair who has already declined to submit a dot plot projection and rejected forward guidance is now proposing to meet less often. Taken together, these are consistent moves toward a Fed that communicates less and lets the data do more of the work.

The minutes are backward-looking. Since that meeting the labour market shed 23,000 jobs and core CPI fell to 2.5%, its pre-conflict pace. The conditional hawkishness the minutes describe was conditioned on inflation not declining. It has declined.

Walmart and the Tariff Refund, Again

Walmart reported Thursday morning. Total revenue of $187.9 billion grew 5.9%. Adjusted EPS of $0.81 grew 19.1% and beat consensus. Full-year net sales guidance was raised from 3.5 to 4.5% growth up to 4 to 5%, and full-year adjusted EPS guidance was lifted to $2.80 to $2.87. CFO John David Rainey said the business model is only getting stronger and more durable. The stock fell 9%.

Three things explain the gap. First, US comparable sales grew 2.6%, the slowest pace in more than six years, against expectations closer to 3.5%. Transactions rose 1.5% while average ticket rose only 1.1%, down from 3.1% a year earlier. People are still coming through the doors. They are buying less per visit. Second, the raised EPS guidance of $2.80 to $2.87 still came in below the roughly $2.90 the market expected, so a raise on paper was a downgrade in practice.

Third, and most familiar to readers of Issue 57, is the tariff refund. Walmart was eligible for roughly $2.9 billion in refunds and has received all but just under $100 million. That is a one-time cash benefit flowing through a quarter in which the operating story was decelerating. Rainey said the company plans to deploy the refunds into lower prices, with the effect visible in the third quarter, which is a defensible use of the money and also an acknowledgement that it is not recurring earnings.

Headline beats built partly on one-time items are not operating momentum. The genuinely encouraging parts of this quarter, eCommerce growth of 24% and advertising growth above 40%, are real and structural. The consumer signal underneath is that the trade-down is now visible in the ticket size at the retailer that benefits most from trading down.

Economic D-Day

On Wednesday evening Trump announced what he described as the most crushing economic operation ever taken against any country, promising economic warfare and isolation on an unprecedented scale. He named the channels he wants closed: oil smuggling, swap lines, cash transfers, exchange houses, ship registries, and front companies. Any country allowing its financial institutions, businesses, airports, or government entities to provide a lifeline to Iran would face consequences.

The pivot matters. Months of military strikes and a naval blockade have not reopened the Strait of Hormuz, which has been effectively closed since early March. This is an admission that the military approach has not produced the result, and a substitution of financial pressure for kinetic pressure.

Markets read it as escalation regardless. Brent rose 2.4% to nearly $94 Thursday, its highest since late July, after adding more than 5% over the prior four sessions. US gasoline is averaging $4.11 a gallon, up nearly a dollar from a year ago. Bessent told CNBC the market was misinterpreting the announcement, arguing that maximum economic pressure makes a return to large-scale combat less likely, and said he was not sure why oil had popped. The market’s answer is that the campaign points at China, Iran’s largest oil buyer, weeks before Xi’s planned US visit. Asked directly whether China would be targeted, Bessent said some conversations are best had in private.

The response from Tehran is worth noting for what it targets rather than what it defends. Iran’s foreign minister dismissed the campaign as a diversion from America’s own crisis of unprecedented debt and surging interest costs. That is rhetoric from an interested party. It also happens to describe, almost exactly, the story in the first section of this issue. When the 30-year yield is at a two-decade high and public debt has crossed $40 trillion, an adversary pointing at the bond market is not making an obviously weak argument.

Closing Thoughts

The risk-free rate is the denominator in every valuation. It does not care which economy a company belongs to. A 30-year yield at 5.3% raises the discount rate applied to Palantir’s 2032 cash flows and to Walmart’s, to CoreWeave’s backlog and to Home Depot’s. That is why a week with no bad earnings news produced the first down week since July. The businesses did not deteriorate. The rate against which they are measured moved.

What makes this different from an ordinary rate move is its source. This is not the Fed tightening to cool an overheating economy, which is a policy choice that can be reversed. This is the long end demanding compensation for fiscal deterioration, which no central bank can vote away. Warsh will speak to it on Friday. He has one week to decide how much to say.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

Monday and Tuesday both delivered fresh all-time highs on the S&P 500 as the market extended the rally that followed July’s weak jobs report. The VIX fell to multi-month lows. CoreWeave reported after Tuesday’s close with revenue up 112% and a backlog figure that exceeded almost every published estimate, sending the stock up more than 13% in extended trading.

Wednesday brought July CPI. Headline inflation rose 0.1% for the month and eased to 3.4% annually. Core rose 0.2% and eased to 2.5%, the lowest reading in five months. Both figures matched consensus exactly. Nvidia gained 3.03% and Cisco 2.86% on the session, while Microsoft, Salesforce, and Amazon led decliners.

Thursday was Applied Materials. Record revenue of $9.12 billion, up 25%, record non-GAAP EPS of $3.50, up 41%, and fourth-quarter guidance comfortably above consensus. The stock closed down 2.48% during the session and fell a further 5% after hours.

Friday delivered the week’s most consequential data. July retail sales fell 0.6%, the steepest monthly decline since May 2025 and the first drop in nine months. The University of Michigan’s preliminary August consumer sentiment reading fell to 51 from 55.2, roughly an 8% decline. The Dow shed 108 points, the S&P 500 eased 0.2% from its record, and the Nasdaq fell 0.3%.

For the week: the S&P 500 gained 0.4% to close at 7,785.76, its third consecutive weekly advance. The Nasdaq added 0.1% to 26,729.16. The Dow declined 0.6%. Chevron led Dow gainers at 7.6%, Merck at 5.7%, and Walmart at 4.5%. Home Depot fell 3.9%. Oil held above $87 a barrel.

Week Ahead

Monday August 17

The Empire State Manufacturing Index for August and the NAHB Housing Market Index both land, the first read on regional manufacturing activity and homebuilder confidence following July’s labour market contraction. A light session otherwise, with markets positioning ahead of the heaviest retail earnings week of the quarter.

Tuesday August 18

July housing starts, building permits, industrial production, and capacity utilisation all drop before the open, alongside import and export price indices. Home Depot reports, the first of four major retailers this week and the most direct read available on home improvement demand in an environment of elevated mortgage rates and falling consumer sentiment. Toll Brothers, Keysight, and Baidu also report.

Wednesday August 19

The FOMC minutes from the July 29 meeting are released at 2 p.m. ET. This is the week’s most important document. Three officials dissented in favour of an immediate hike at that meeting, and the minutes will reveal how close the remaining members were to joining them and what specific data thresholds the committee identified. Read alongside the July jobs contraction and this week’s soft CPI, the minutes will show whether the hawkish case was conditional or convicted. Target, Lowe’s, TJX, Analog Devices, and Estée Lauder all report.

Thursday August 20

Walmart reports before the open, the single most important consumer datapoint of the month. As the largest retailer in the world and one that skews toward lower-income households, Walmart’s comparable sales and management commentary on trade-down behaviour will confirm or complicate the retrenchment signal that Friday’s retail sales and sentiment data delivered. Initial jobless claims, the Philadelphia Fed Index for August, and July leading indicators also land. Deere, Ross Stores, and Alibaba report.

Friday August 21

No major scheduled data or earnings. Attention turns to the Jackson Hole symposium the following week, where Kevin Warsh delivers his remarks on August 28. By Friday’s close the market will have the FOMC minutes, four major retail earnings reports, and this week’s inflation and consumption data all in hand, which together form the complete picture Warsh will be responding to.

CoreWeave’s $104 Billion Answer to Meta Compute

In Issue 57, this newsletter covered Meta’s announcement of Meta Compute, a cloud business built to sell excess AI capacity, and CoreWeave falling roughly 12% on the news. The concern was structural: if the hyperscalers with the largest capex budgets started selling compute rather than buying it, the addressable market for specialised neocloud providers would contract sharply.

Six weeks later, CoreWeave signed a $21 billion agreement to supply Meta with AI cloud capacity through 2032, layered on top of a prior $14 billion commitment. The competitor became the customer.

The Q2 results reported Tuesday evening put numbers behind that reversal. Revenue of $2.575 billion grew 112% year-on-year and came in near the top of the company’s own guidance range. Adjusted operating income of $128 million landed above the $30 to $90 million guidance, more than double the midpoint. Adjusted EBITDA reached $1.51 billion at a 59% margin. Revenue backlog stood at $104.2 billion at quarter end, up 246% year-on-year, and management noted this figure excludes more than $25 billion in net new customer commitments secured in the six weeks between June 30 and the August 11 report. A multi-year agreement with Anthropic to provide compute for the Claude model family was also formalised in the period. Full-year revenue guidance was raised to $12.4 to $13.2 billion, and third-quarter guidance of $3.4 to $3.6 billion implies roughly 158% growth at the midpoint, an acceleration against already elevated comparisons.

The other side of the ledger deserves equal attention. Net losses widened to $626 million from $290 million a year ago. Net interest expense reached $640 million, more than double the $267 million recorded in Q2 2025. The company raised over $10 billion in unsecured debt and convertible bonds during the quarter, including its inaugural Eurobond issuance, and 2026 capex guidance now sits at $35 to $39 billion. CoreWeave remains unprofitable on a GAAP basis and is funding an enormous infrastructure build with debt at a scale that leaves little room for a demand slowdown.

This is the sharpest available test of the framework this newsletter set out in Issue 61. Microsoft’s capex was rewarded because $678 billion in contracted RPO justified it. Meta was punished because no equivalent anchor existed. CoreWeave has $104 billion in contracted backlog against $35 to $39 billion in annual capex and $640 million in quarterly interest expense. The backlog is real, and the contracts are binding. Whether the balance sheet can carry the build long enough to convert that backlog into free cash flow is the question that will define the investment over the next two years, and it is a genuinely different question from whether demand exists.

Applied Materials Beat, Raised, and Fell 5%

Applied Materials reported record fiscal Q3 results on Thursday. Revenue of $9.12 billion grew 25% year-on-year, a new quarterly high. Non-GAAP EPS of $3.50 grew 41% and also set a record. Gross margin reached 50.4%, up 1.5 percentage points year-on-year, marking the thirteenth consecutive quarter of gross margin expansion. Fourth-quarter guidance of $3.82 to $4.22 in non-GAAP EPS came in well above the roughly $3.71 consensus. CEO Gary Dickerson raised the calendar 2026 semiconductor systems outlook and said customer visibility points to another strong growth year in 2027, with particularly strong second-half growth expected in DRAM, leading-edge foundry logic, and advanced packaging.

The stock closed down 2.48% during Thursday’s session and fell a further 5% after hours.

This pattern is no longer an anomaly. Broadcom fell 12% on record AI revenue in June. CrowdStrike fell 11% after beating every metric. Micron fell 30% from its peak while maintaining $50 billion in quarterly guidance. TSMC fell 7% on the day it reported a record quarter. Samsung’s 19-fold profit increase sent Korean stocks down 4.91%. Now Applied Materials joins the list with a beat, a raise, and a fall.

What connects these is not weakness in the underlying businesses. It is that all of them entered their reports having already run enormous distances. Applied Materials had rallied roughly 200% before this print and trades at a forward multiple of approximately 44 against a five-year average near 21. When a stock is priced for exceptional results, delivering exceptional results is the baseline rather than the surprise. The move that follows is determined by whether the guidance exceeds what was already embedded in the price, not by whether the quarter was good.

For long-term investors, the analytical discipline this demands is the same one this newsletter has applied to Nike’s tariff refund and Tesla’s investment cycle: separate the business from the price reaction. Applied Materials’ thirteenth consecutive quarter of margin expansion and its raised 2027 visibility are statements about the durability of semiconductor equipment demand.

July CPI: Inflation Is Back to Where the War Started

July CPI rose 0.1% for the month, following June’s 0.4% decline, bringing the annual rate to 3.4% from 3.5%. Core CPI rose 0.2% after holding flat in June, with the annual rate easing to 2.5% from 2.6%, the lowest core reading in five months. Both figures matched consensus exactly.

The detail that matters most is that core inflation has now returned to its February pace, which is to say its pre-Iran conflict rate. The three-month annualised core rate fell to 1.6% from 2.3%, and the six-month rate to 2.4% from 2.6%. Shelter, which accounted for roughly two-thirds of the monthly increase, rose just 0.1% and decelerated to 3.2% annually. Energy fell 1.5% for the month, though it remains 14.7% higher than a year ago with gasoline up 24.6% over that span.

The inflation surge that drove headline CPI to a four-year high in May was, on this evidence, overwhelmingly an energy shock rather than a broadening of underlying price pressure. As the energy component fades from the annual comparison, the core rate has settled back to where it sat before the conflict began. That is close to the best possible outcome for a central bank that spent the first half of the year worrying about whether the shock would become embedded.

Not everyone is persuaded. Cleveland Fed President Beth Hammack, one of the three July dissenters, wrote publicly the day before the print that “now is the time to act,” arguing that the longer the Fed waits to return inflation to 2%, the more difficult and expensive that return becomes. Her position is not about whether inflation is decelerating. It is about the fact that 3.4% headline and 2.5% core both remain above target after an extended period and that credibility erodes with tolerance.

Market pricing after the release left the September hike probability around 40%, essentially unchanged. Morgan Stanley Wealth Management’s Ellen Zentner noted that in-line inflation preserves the no-need-to-hike narrative that took hold after the jobs report. Truist’s Mike Skordeles said the data supports a near-term hold. One more jobs report and one more CPI arrive before the September meeting, with Warsh’s Jackson Hole remarks on August 28 in between. On current data, the hawks have the argument about the level, and the doves have the argument about the direction.

The State Of The Consumer

July retail sales fell 0.6% to $763.6 billion, the steepest monthly decline since May 2025 and the first drop in nine months. Economists had expected a 0.1% increase. Excluding autos, sales fell 0.3% against an expected 0.2% gain. Excluding gasoline, sales still fell 0.6%. Online sales led the decline at 2.2%, followed by a 2% drop at car dealerships. Clothing rose 1.9% on back-to-school buying, and restaurants gained 0.5%. Part of the weakness is explained by Amazon shifting Prime Day from July to June, pulling spending into the prior month, and part by the fading boost from earlier tax refunds. Neither explanation accounts for the full decline.

The University of Michigan’s preliminary August consumer sentiment reading fell to 51 from 55.2 in July, roughly an 8% decline and well below the 55 economists expected. It ended two consecutive months of improvement. Expected business conditions fell 11% for the short run and 17% for the long term. One-year inflation expectations rose to 4.3%, against the 3.4% that prevailed before the Iran war. The deterioration was broad but concentrated among older consumers, lower-income households, and those without college degrees.

Underneath both reports sits the number that explains them. Real average hourly earnings in July were 0.2% lower than a year earlier. Inflation has been outpacing wage growth for four consecutive months. Nominal wage growth of 3.2% against headline inflation of 3.4% means the average American worker is losing purchasing power every month and has been since the spring. Labour force participation, meanwhile, is at its lowest level since 1976 outside the pandemic.

This is the tension that defines the current market, and it is worth stating plainly. The S&P 500 hit all-time highs twice this week. Second-quarter earnings for S&P 500 companies are on pace to rise roughly 50% year-on-year, the strongest growth since 2021, driven overwhelmingly by AI. At the same time, the consumer who generates two-thirds of US economic activity is spending less, feeling worse, and earning less in real terms than a year ago. Both of these are true simultaneously. The AI capital cycle is currently large enough to carry index-level earnings without consumer participation. Whether that remains true if consumer retrenchment deepens into the second half is the most important open question for anyone holding equities into 2027. Next week’s Walmart, Target, Home Depot, and Lowe’s reports will provide the first detailed answer.

Closing Thoughts

There are currently two economies operating in the United States, and the S&P 500 only reflects one of them.

The first economy is the AI capital cycle. CoreWeave’s $104 billion backlog, Applied Materials’ thirteenth consecutive quarter of margin expansion, and second-quarter S&P 500 earnings growth of roughly 50% all describe a boom of genuine historical scale, anchored in binding contracts and physical infrastructure. The second economy is the household. Retail sales at their weakest in over a year, sentiment down 8% in a month, real wages negative for four consecutive months, and participation at levels not seen since 1976.

For long-term investors, the practical question is how long the first can continue to drive index-level returns without the second. The honest answer is that nobody knows and that the historical precedents point in both directions. What can be done is to understand which economy each holding depends on. A memory manufacturer with $100 billion in multi-year customer commitments has a different exposure to consumer retrenchment than a retailer, a restaurant chain, or a consumer hardware company facing rising input costs. That distinction has never mattered more than it does right now.

Wednesday’s FOMC minutes and Thursday’s Walmart report will each address one of the two economies. Together they make next week unusually informative.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

The week began on a busy note with Palantir reporting after Monday’s close. Markets were generally quiet in the meantime as the print was widely expected to be strong, and it delivered – and then some. Revenue of $1.94 billion, up 93%, adjusted EPS of $0.41 versus $0.35, US commercial revenue up 149%, and a full-year guidance raise of nearly $500 million were far above the high-end estimates and drove the stock up over 29% in the following session, eliminating nearly $3 billion in short positions in one day.

SpaceX reported its first-ever quarter on Monday evening alongside Palantir, delivering revenue of $7.8 billion, beating the $6.9 billion consensus by nearly $1 billion, up 92% year-on-year, with Starlink subscribers doubling to 12 million. The stock fell after hours as investors focused on $18.4 billion in quarterly capital expenditure, $15.8 billion of it in the AI segment alone, well above the $13.2 billion analyst estimate. Shares of the rocket company rose over 20% for the week as investors balanced the $14.1 billion in new AI cloud contracts against concerns about the capex.

The market moved lower on Friday on the jobs report, which showed a loss of 23,000 nonfarm payrolls in July, well below the 83,000 to 95,000 consensus and the first negative print since 2025. June employment was simultaneously downwardly revised to just 20,000 from 57,000, removing 60,000 jobs from the picture the markets had been considering. Odds for a September rate hike collapsed as the Fed’s three dissenters voted to raise rates at the July 29 meeting. The Dow and S&P 500 both set new all-time highs.

To briefly summarise the results for the week: The S&P 500 was up 3.6%, the Nasdaq -5.2%, whilst the Dow managed to gain 3.0%. Next week sees the publication of the July CPI, followed by reports from CoreWeave, Cisco and Applied Materials.

Week Ahead

Monday August 10

Simon Property Group, Rocket Lab and Archer Aviation report. Amazon’s Zoox robotaxi begins charging fares in Las Vegas for the first time, representing both the first robotaxi in the city and direct competition with Tesla’s Cybercab.

Tuesday August 11

CoreWeave reports alongside Sea, Cardinal Health, On Holding and CAVA Group. CoreWeave stands out as perhaps the most important report of the week, as the neocloud provider directly competes with Meta’s newly launched Compute business. This represents the first time the market has seen disclosure on revenue or bookings following the launch of the hyperscaler’s direct competitor. The New York Fed publishes its Q2 Quarterly Report on Household Debt and Credit.

Wednesday August 12

July CPI is published at 8:30 a.m. ET. With economists forecasting a moderation to 2.5%, the core inflation print stands to be the most important figure of the week. If anything, it is forecast to be lower than the 3.4% recorded in May, providing a further basis for the Fed to pause and complicating the outlook for rate hikes. Cisco and Coherent report, and OPEC publishes the monthly oil market report.

Thursday August 13

Applied Materials reports after the close and should provide the strongest read on the semiconductor equipment industry this month. A beat and raise will indicate that Micron’s $250 billion commitment and increased capex by TSMC are flowing into equipment purchases. SanDisk hosts its Investor Day, and CEO David Goeckeler is due to preview a multi-year roadmap for the NAND flash business against the backdrop of the AI memory supercycle described in this newsletter since issue 49. July PPI is due at 8:30 a.m. ET.

Friday August 14

July Retail Sales is published at 8:30 a.m. ET, providing the first comprehensive read on the effects of the labour market correction, higher oil prices and the recent hikes in Apple product pricing on consumer spending. The 13F deadline for institutional managers to report long holdings as of June 30 is due, with Nvidia, Meta, Tesla, CoreWeave and Corning among the most watched names. By the end of the week and with CPI, PPI, retail sales and the 13F disclosures, the market will have a near-complete understanding of the forces influencing Warsh’s Jackson Hole speech.

Palantir’s 93% Growth

Palantir CEO Alex Karp has a way with language that tends to obscure as much as it reveals. “This quarter was otherworldly” is not a number, and “demand for AI sovereignty has now been unleashed” is not a financial metric. But beneath the rhetoric, Palantir’s Q2 2026 results are among the most structurally significant AI software prints of this earnings season.

Revenue of $1.94 billion grew 93% year-on-year, beating the $1.81 billion consensus by $123 million and accelerating from 85% growth in Q1. US commercial revenue of $764 million grew 149% year-on-year and 380% since the start of 2024. US government revenue grew 90% to $809 million. GAAP operating income reached $912 million at a 47% margin. Adjusted operating income hit $1.19 billion at a 62% margin. Net income more than tripled from a year earlier. The Rule of 40 score reached 155, one of the highest figures ever recorded by a public software company. Total contract value closed at $3.37 billion, up 49% year-on-year. US commercial remaining deal value reached $6.24 billion, more than doubling from a year ago. Full-year guidance was raised from $7.65 billion to $8.15 billion, a $500 million increase in a single quarter.

The “AI sovereignty” framing, stripped of its rhetoric, describes something commercially specific. Palantir’s platform allows enterprise and government customers to run AI models on their own infrastructure using their own data, without routeing sensitive information through third-party cloud providers. For defence agencies, financial institutions, and healthcare systems, the alternative of sending proprietary data to OpenAI, Anthropic, or Google is often not viable. Palantir occupies the gap between what hyperscalers can offer and what regulated or security-sensitive customers can actually accept.

The US commercial acceleration from 133% in Q1 to 149% in Q2, the 220 deals worth at least $1 million closed in the quarter, and the $6.24 billion in remaining deal value provide a contractual anchor directly comparable to the RPO metrics this newsletter used in Issue 61 to separate Microsoft’s rewarded capex from Meta’s punished capex. The valuation tension is real: at 141 times earnings, Palantir has almost no room for deceleration. Karp’s 18-month visibility comment is the most credible available forward signal, but the quarterly data will be the confirmation.

SpaceX’s First Earnings: Starlink Is Funding xAI

SpaceX’s first earnings report as a public company produced figures that in any other context would be unambiguously celebrated. Revenue of $7.8 billion, up 92% year-on-year, beat the $6.9 billion consensus by nearly $1 billion. Starlink subscribers doubled to 12 million. AI revenue grew 247% to $2.56 billion. Adjusted EBITDA reached $3.5 billion, up 191%. The space segment won over $6 billion in new US government Starshield contracts. $14.1 billion in new AI cloud service agreements were closed. Net losses narrowed to $541 million from $1 billion a year earlier.

And then the capex: $18.4 billion in the quarter, $15.8 billion of it in the AI segment alone, annualising at $73.5 billion, well above the $48.7 billion consensus. The stock fell after hours, following the pattern this newsletter has documented across Alphabet, Meta, and the broader hyperscaler cohort: extraordinary revenue growth paired with capital expenditure the market cannot yet attach to a sufficient contractual backlog.

The three-segment breakdown makes the arithmetic plain. Starlink generated $4.29 billion in revenue at a 38.6% operating margin, the only profitable segment in the company. The space segment generated $962 million in revenue but posted a $542 million operating loss as Starship R&D costs climbed. The AI segment generated $2.56 billion in revenue while consuming $15.8 billion per quarter in capex. Starlink is the only source of operating profit. That profit is being reinvested into xAI at a rate that currently produces $2.56 billion in revenue against $15.8 billion in quarterly spending.

The $14.1 billion in new AI cloud contracts prevents this from being a straightforward Meta-style critique. Unlike Meta’s spending, SpaceX’s AI capex is at least partially attached to signed customer agreements. The approximately $100 billion in cash and marketable securities following the IPO provides a liquidity runway Meta does not have. SpaceX sits between the two extremes this newsletter defined in Issue 61: more demand visibility than Meta, less than Microsoft’s $678 billion commercial RPO. Whether the AI segment can demonstrate revenue growth proportionate to its $15.8 billion quarterly spending in the next two quarters will determine which comparison the market ultimately applies.

July Jobs: The Labour Market Is Now Contracting

Nonfarm payrolls fell 23,000 in July against a consensus range of 83,000 to 95,000, the first negative monthly print since 2025. June was simultaneously revised down to just 20,000 from the previously reported 57,000, removing 60,000 positions from the picture markets that had been working with.

Government employment fell 53,000, led by local government education shedding 50,000 positions as summer seasonal patterns interacted with ongoing federal workforce reductions. Retail trade lost 19,000. Leisure and hospitality shed 40,000. Private payrolls increased 30,000, partially offsetting the government weakness but not reversing the headline direction.

Average hourly earnings rose just 2 cents, bringing the 12-month wage growth rate to 3.2%, the lowest since May 2021 and below the 3.5% forecast. This is the detail that carries the most weight for the Fed. The hawkish case from Hammack, Kashkari, and Logan has centred on services inflation, which is most directly tied to wage dynamics. A 3.2% reading, decelerating from 3.5%, removes one of the primary arguments for immediate rate action. Labour force participation fell to 61.4%, the lowest in over five years. The employment level has fallen by 833,000 since January.

The implication for the Fed is direct. The three dissenters who voted to hike at the July 29 meeting based their case on a resilient labour market. That case has been materially weakened. The base case for September is now a hold. What Wednesday’s CPI print determines is whether any residual justification for the hawkish position survives, or whether a core reading near economists’ 2.5% consensus makes the dissenters’ September vote harder still to defend. Jackson Hole on August 27 is where Warsh synthesises all of it publicly, knowing the labour market is moving in a direction that substantially complicates the hawks’ argument.

The AI Software Monetisation Layer Is Now Visible

Since Issue 48, this newsletter has analysed the AI trade through the lens of infrastructure: hyperscaler capex commitments, the memory supercycle, semiconductor manufacturing investment, and the cloud revenue acceleration that began confirming demand in recent issues. Those layers of the trade are now well documented. What this week adds is the layer that has been hardest to see until now: the software monetisation layer, where AI capabilities get tied to specific enterprise workflows and sold as recurring revenue with contractual commitments.

The past four weeks of data make a three-layer structure visible for the first time with enough specificity to be analytically useful. The infrastructure layer, anchored by Micron’s $250 billion US manufacturing commitment and $100 billion in Strategic Customer Agreements, confirmed in Issues 56 to 58 that the physical compute being built is backed by contractual demand. The platform layer, seen in Issues 59 to 61 through AWS at 37%, Azure at 43%, and Google Cloud at 82%, showed that infrastructure is generating accelerating cloud revenue as enterprises shift AI workloads from experimentation to production. This week’s Palantir results add the third layer: software built on top of those platforms, where AI capabilities are converted into enterprise-specific workflows and sold under multi-year contracts with remaining deal value disclosures.

Palantir’s $6.24 billion in US commercial remaining deal value, more than doubling year-on-year, is the software layer’s equivalent of Micron’s $100 billion SCA backlog or Microsoft’s $678 billion commercial RPO. SpaceX’s $14.1 billion in new AI cloud contracts adds another data point to the same picture.

Each layer carries a different risk profile and a different valuation framework. Infrastructure names like Micron trade at approximately 8 times forward earnings, reflecting cyclicality risk that the SCA structure is reducing. Platform names like Microsoft trade at approximately 35 times forward earnings, reflecting the compounding nature of contracted cloud revenue at scale. Software names like Palantir trade at approximately 141 times forward earnings, reflecting hypergrowth and the uncertainty about its duration. All three can be simultaneously rational, because they reflect genuinely different risk-reward profiles within the same underlying demand cycle. The investor’s task is to understand which layer each holding sits in, what the contractual demand anchors are, and what deceleration in those anchors would mean for the thesis. The past four weeks of earnings have provided enough data to make that assessment with real specificity for the first time.

Closing Thoughts

Three weeks ago, Warsh walked out of the July FOMC meeting with three dissenting votes, a 30-year yield at its highest since 2007, and markets pricing a 41.9% probability of a September hike. This week the labour market shed 23,000 jobs, wage growth slowed to its lowest since May 2021, and the employment level fell for the seventh consecutive month. The speech Warsh was going to give at Jackson Hole three weeks ago is not the speech he will give now.

What has not changed is the AI story. Palantir at 93%, SpaceX at 92% revenue growth, AWS at 37%, Azure at 43%, Google Cloud at 82%: the demand cycle across infrastructure, platform, and software is intact and accelerating. The gap between an economy that shed 23,000 jobs in July and a technology sector that closed $14.1 billion in new AI cloud contracts in the same period is not a paradox. It is the normal condition of a structural technology transition: extraordinary value being created in a concentrated set of businesses while the broader economy absorbs the disruption that the same transition causes. IBM’s 25% fall on enterprise software demand shifting to AI was the same story told from the other side.

For long-term investors, the framework this newsletter has been building since Issue 49 now has enough data behind it to function as a genuine analytical foundation. The AI trade is real, layered, contractually anchored, and macro-sensitive in the specific way that all growth investing is macro-sensitive: through the discount rate. A weakening labour market and a likely September hold change that rate in a direction that is, all else equal, positive for growth valuations. Jackson Hole will tell us whether Warsh agrees.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

Monday opened cautiously. The Dow gained 0.51%, the S&P 500 was essentially flat, and the Nasdaq lost 0.18% as semiconductor stocks extended their recent losses. Oil prices fell as the US and Iran paused their attacks while mediators resumed negotiations, providing modest relief from the prior week’s $100 intraday spike.

Tuesday built momentum. The Dow climbed 537 points led by Sherwin-Williams up 8% and Coca-Cola up 5% on strong earnings. Apple briefly crossed a $5 trillion market capitalisation for the first time, making it the most valuable publicly traded company in the world ahead of its Thursday earnings report. The VanEck Semiconductor ETF fell more than 3% for a fourth consecutive session.

Wednesday was the most consequential day of the week. The FOMC held rates steady at 3.5% to 3.75% for the fifth consecutive meeting, but three regional Fed presidents dissented in favour of an immediate hike: Beth Hammack of Cleveland, Neel Kashkari of Minneapolis, and Lorie Logan of Dallas. The Dow dropped more than 840 points. The 30-year Treasury yield hit its highest level since 2007. After the close, Microsoft reported a record quarter and Meta reported an EPS miss.

Thursday reversed sharply. The Dow gained 613 points, the S&P 500 rose 1.66%, and the Nasdaq jumped 2.78% as Microsoft surged on its results. Apple and Amazon both reported beats. Meta fell 9.64%.

Friday extended the gains. Amazon jumped 15.3%, Alphabet gained 6.9%, and Microsoft added 3%. Apple fell 7.13% on chip shortage concerns affecting production volumes. For the week: Dow and S&P 500 each gained approximately 1%, Nasdaq advanced roughly 1.6%. For July overall: the S&P 500 slipped 0.1% and the Nasdaq declined 3.2%, with the Dow the only major index to post a monthly gain at 0.3%.

Week Ahead

Monday August 3

ISM Manufacturing PMI for July lands at 10 a.m. ET, the first major economic read of August and the first manufacturing gauge to fully capture the period in which oil crossed $100 and the Fed held with three dissents. Palantir and Caterpillar report, a pairing that covers AI software monetisation and industrial demand in the same morning.

Tuesday August 4

JOLTS job openings for June land alongside factory-order data. Uber and Airbnb both report, giving a read on consumer mobility and travel spending heading into peak summer. Airbnb’s commentary on forward bookings will be particularly watched given rising oil prices and the potential consumer confidence drag from the Iran conflict’s persistence.

Wednesday August 5

ADP private payrolls for July drop at 8:15 a.m. ET, the first read on labour market conditions in the month after the Fed held with three dissents. ISM Services PMI also lands. Walt Disney reports after the close, providing a read on theme park attendance, streaming subscriber trends, and whether consumer entertainment spending is holding up under the current inflation and oil price environment.

Thursday August 6

Initial jobless claims land alongside the preliminary Q2 productivity and unit labour cost data, the most direct measure of whether wage growth is translating into labour cost inflation that would justify the three dissenters’ position. Eli Lilly reports, as does Shopify, giving a read on GLP-1 drug demand and small business e-commerce health.

Friday August 7

The July Employment Situation report lands at 8:30 a.m. ET. This is the most important single data point before Jackson Hole on August 27 to 29. After June’s 57,000 miss, a rebound toward consensus would reopen the September hike debate that the three dissenters have already voted to accelerate. A second consecutive soft print would give Warsh the space to hold again and use Jackson Hole to set a more deliberate communication framework. The jobs report effectively sets the agenda for the most consequential Fed speech of the year.

Microsoft vs. Meta: The AI Capex Verdict

Microsoft reported fiscal Q4 2026 results on Wednesday evening that were, by almost any measure, the strongest quarter in the company’s history. Revenue of $90.01 billion, up 18% year-on-year, beat the $87.62 billion consensus. Adjusted EPS of $4.74 beat $4.24. Azure grew 43% in the quarter, crossing $100 billion in annual revenue for the first time. The Intelligent Cloud segment generated $39.31 billion in revenue, up 32%. Commercial remaining performance obligations surged 84% to $678 billion, meaning Microsoft now has $678 billion in contracted future revenue on its books. Microsoft 365 Copilot crossed 30 million paid seats, up from 20 million in April. The stock rose more than 8% after hours and added approximately $260 billion in market value.

There are two details worth flagging before moving to the investment conclusion. First, the reported EPS included a $3.2 billion gain from Microsoft’s investment in Anthropic and lower-than-expected costs from a voluntary retirement programme. Strip those items and the underlying beat is still real, but the headline profit growth figure of 31% year-on-year overstates the operational performance. Second, full-year capital expenditure reached $115.95 billion, a record. Free cash flow declined. These are not trivial numbers to absorb. What makes Microsoft’s spending different from Meta’s is not the scale. It is the contractual visibility tied directly to that spending.

Meta reported Q2 2026 results the same evening. Revenue of $60.8 billion beat the $60.2 billion consensus by 0.9% and grew 28% year-on-year. On every revenue metric, the result was strong. On every profitability metric, it was not. EPS of $6.18 missed the $7.23 consensus by 14.5%. Total expenses climbed 55% to $42 billion, including $2.4 billion in legal charges and $1.2 billion in severance. Operating margin collapsed from 43% in Q2 2025 to 31% in Q2 2026. Free cash flow fell to just $784 million for the quarter, from $13.7 billion a year earlier. Reality Labs posted another $4.62 billion operating loss on $431 million in revenue. The stock fell 9.64% in after-hours trading.

The comparison is not primarily about the headline numbers. It is about what each company’s spending is attached to. Microsoft’s $678 billion commercial RPO is a binding contractual commitment from enterprise customers to purchase Azure, Copilot, and cloud services over future periods. Every dollar Microsoft spends on AI infrastructure is being pulled by customers who have already signed contracts specifying what they will buy and at what price. That is not a speculative investment in future demand. It is a capital deployment in response to demand that has already been contractually confirmed.

Meta’s $42 billion in quarterly expenses includes $2.4 billion in legal charges and $1.2 billion in severance that are entirely unrelated to AI infrastructure. Strip those items and the picture improves, but Meta’s AI spending still lacks the contractual demand anchor that Microsoft has in its RPO. Meta Compute, the cloud business announced in early July as a potential monetisation vehicle for excess compute capacity, has not yet generated meaningful disclosed revenue. Reality Labs has now accumulated more than $80 billion in total operating losses since late 2020 with no disclosed path to profitability. The capex guidance of $130 to $145 billion for 2026 is being deployed into infrastructure that, unlike Azure or AWS, does not yet have a confirmed commercial backlog to justify it at the scale being committed.

This is the verdict the market delivered this week, and it is worth stating plainly for long-term investors: AI capital expenditure is not being punished uniformly. It is being rewarded where it is attached to contractual revenue commitments and punished where it is not. That distinction, not the absolute dollar amount of spending, is the analytical lens that separates the AI infrastructure trade into its component parts. Microsoft has $678 billion in contracted future revenue. Meta has $784 million in quarterly free cash flow and a capex bill of $130 billion for the year. The market is reading both accurately.

Amazon’s AWS at 37%: The Cloud Acceleration

The single most important number in this entire week’s earnings slate is not Microsoft’s $678 billion RPO or Meta’s $6.18 EPS miss. It is AWS’s 37% revenue growth rate.

Amazon Web Services generated $42.2 billion in revenue in Q2 2026, growing 37% year-on-year and accelerating from 28% growth in Q1. That acceleration, nine percentage points in a single quarter, is the largest sequential re-acceleration in AWS’s history outside of the period immediately following the pandemic cloud adoption surge. It beat the $40.5 billion consensus estimate by nearly $2 billion, and AWS described it as its fastest growth in 18 quarters.

Put this alongside Azure at 43% and Google Cloud at 82%, both reported in the prior week, and what emerges is the clearest available picture of AI-driven cloud demand: all three major hyperscale cloud platforms are accelerating simultaneously. This has not happened since 2021. The conventional concern that AI infrastructure spending would eventually create a supply glut and compress cloud pricing has not materialised in the revenue data available through the end of June 2026. The demand is absorbing the supply as it comes online.

Amazon’s overall Q2 results carried one significant footnote that this newsletter applies the same discipline to as it applied to Nike’s tariff refund and Microsoft’s Anthropic investment gain. Amazon’s reported EPS of $1.85 beat the consensus exactly. But the result included a $2.1 billion gain from Amazon’s investment in Rivian. Strip that item and the underlying EPS is approximately $1.58, below consensus. The beat is real on the operational top line, including AWS. It is partially an accounting artefact on the bottom line. Long-term investors who understand this distinction will make better sense of both why the stock jumped 15% and why that reaction is not purely an endorsement of the underlying earnings quality.

AWS operating income reached $14.5 billion, up 65% year-on-year, at an operating margin of 34.3%. CEO Andy Jassy said on the call that AWS is “seeing accelerating demand, particularly from generative AI workloads,” and confirmed that the company is still supply-constrained in certain regions despite significant capacity additions. That supply constraint is the mirror image of Micron’s SCA disclosures and TSMC’s CoWoS sold-out commentary: the AI infrastructure build-out, measured in actual commercial cloud revenue rather than capex announcements, is still in a demand-constrained phase rather than a supply-glut phase.

For long-term investors, the AWS re-acceleration is the most important single data point in this earnings season for understanding where the AI demand cycle actually stands. The semiconductor selloff that followed Kimi K3’s release three weeks ago was driven by concerns that efficient AI models would reduce compute demand. AWS’s 37% growth rate, delivered in the quarter immediately following those efficiency model releases, is the empirical answer to that concern. Cheaper inference costs did not reduce AWS’s revenue. They expanded the addressable market for cloud AI services and accelerated the deployments that flow through AWS’s infrastructure. Jevons’ paradox in the cloud computing context is no longer theoretical. It is in the quarterly results.

Apple at $5 Trillion

On Tuesday July 28, Apple briefly crossed a $5 trillion market capitalisation for the first time, making it the most valuable publicly traded company in the world, passing Nvidia. On Friday August 1, the stock fell 7.13%, its worst single session since April 2025, wiping approximately $360 billion in market value in a day. Both things happened within four days of each other, and both are worth understanding precisely.

The earnings themselves, reported Thursday after the close, were genuinely strong on most metrics. Revenue of $109.4 billion, up 16% year-on-year, beat the $108.2 billion consensus. EPS of $2.02 beat the $1.87 consensus by 8%. iPhone revenue of $44.8 billion beat estimates. Mac revenue of $10.7 billion beat estimates. iPad revenue of $8.1 billion beat estimates. Wearables beat estimates. Gross margin of 46.5% was above guidance.

The miss was in services, and it was meaningful. Services revenue of $30.74 billion grew 12% year-on-year but came in below the $31.39 billion consensus, the first services miss in several quarters. Services is Apple’s highest-margin, most predictable revenue segment, and it is the primary justification for the premium multiple on Apple’s stock. A miss there carries disproportionate weight relative to its dollar magnitude.

The production constraint disclosure is the detail that drove Friday’s 7% fall. Tim Cook confirmed on the call that chip shortages, specifically the DRAM and NAND scarcity this newsletter has tracked since Issue 49, constrained iPhone production in the June quarter. Apple could not build as many units as it could sell. The company declined to give formal revenue guidance for Q4, which Cook described as a reflection of “unprecedented supply uncertainty” rather than weak demand. The implication for the September iPhone 18 launch, which Cook has said will carry higher prices due to memory costs, is that Apple enters its most important product cycle of the decade with a constrained supply chain and a premium pricing strategy simultaneously.

Tim Cook confirmed on the call that this was his final earnings call as CEO. He steps back to executive chairman in September. The incoming CEO, whose identity has not yet been publicly confirmed by Apple, inherits a company at $5 trillion in market cap, in the middle of a memory-driven cost crisis, deploying its first AI product suite under the Apple Intelligence brand, and navigating a geopolitical environment in which its largest manufacturing base remains concentrated in a country that is the subject of ongoing US-China technology tensions. It is not a simple inheritance.

For long-term investors, the Apple story this week is a study in the gap between headline valuation and underlying operational complexity. A $5 trillion market cap implies a level of predictability and durability that the June quarter, with its services miss and production constraint, does not straightforwardly confirm. That does not make Apple a bad business. It is one of the greatest businesses in the history of capitalism. It does suggest that the current multiple requires either a reacceleration in services growth or a successful iPhone 18 cycle that overrides the memory cost headwind, and probably both. Cook has delivered both many times before. Whether his successor can is a question the market has not yet fully priced into the stock.

The Fed’s Three Dissenters and What Warsh Said Next

The Federal Reserve held rates steady at 3.5% to 3.75% on Wednesday, as expected. What was not expected, or at least not fully priced, was the scale of the internal dissent: three regional Fed presidents voted to hike immediately, the most dissents in favour of tightening since September 2016. Cleveland Fed president Beth Hammack, Minneapolis Fed president Neel Kashkari, and Dallas Fed president Lorie Logan each voted for an immediate 25 basis point increase.

The market’s reaction was immediate. The Dow dropped more than 840 points in the hour following the decision and Warsh’s press conference. The 30-year Treasury yield hit its highest level since 2007, a moment worth sitting with. The 30-year yield at its highest since 2007 is the bond market pricing a structurally higher long-term inflation path than it has at any point in the current tightening cycle. It is not a short-term rate call. It is a long-duration structural statement about where the market believes inflation will settle over the next decade.

Warsh’s press conference language was characteristically sparse and carefully chosen. He confirmed that press conferences would continue through the remaining 2026 meetings as committed to by his predecessor, but signalled this structure may evolve. He described the committee’s internal debate as healthy rather than alarming: “I asked for a good family fight and I got one.” On the specific question of whether three dissents represent a committee on the verge of hiking, Warsh said he was “laser focused on the direction of travel in the data, not relying on any single print.” That framing is significant: it suggests Warsh is managing a committee whose hawkish members are impatient while he maintains a data-dependent holding pattern rather than committing to either path.

The September probability shift is the most actionable piece of information to emerge from Wednesday’s decision. Before the meeting, markets priced roughly a 24% probability of a September hike. By Thursday morning, that figure had moved to 41.9%. A single meeting did not tip the probability over 50%, but the three-dissent revelation moved it materially in one direction. The data between now and September 15 to 16, specifically the July jobs report on August 7 and July CPI in mid-August, will determine whether September becomes the live meeting that the dissenters clearly believe it should be.

Jackson Hole on August 27 to 29 is now the most consequential speech of Warsh’s tenure. It is not formally a policy announcement, but in practice every Fed chair has used Jackson Hole to signal the committee’s thinking ahead of the September meeting. Given that Warsh has explicitly rejected forward guidance as a policy tool, the question is whether he uses Jackson Hole to provide the kind of framework communication that markets have come to expect, or whether he deliberately withholds it as a statement about his new operating philosophy. Either choice sends a signal. The bond market, with the 30-year at its highest since 2007, is already positioning for the possibility that Warsh chooses the latter and that the answer, when it comes, is hawkish.

For long-term investors, the three-dissent vote changes the baseline assumption about the second half of 2026. Through the first half, the question was whether Warsh would hike at all this year. Through the second half, the question is whether he can hold the committee together against three members who have already publicly stated their preference for immediate action. The July jobs report on August 7 is the first decisive data point. A rebound from June’s 57,000 would hand the dissenters a strong argument. A second consecutive soft print would give Warsh the runway to hold through September and use Jackson Hole to reframe the committee’s communication approach on his own terms.

Closing Thoughts

Is the AI infrastructure build-out generating revenue proportionate to the capital being deployed? The answer, as delivered by Microsoft, Amazon, and Google Cloud’s combined results, is yes. AWS at 37% growth and Azure at 43% growth, both accelerating in the same quarter, are not the numbers of a demand cycle running out of room. They are the numbers of a demand cycle still in expansion.

The more nuanced answer, delivered by Meta’s 9.6% post-earnings fall, is that the yes applies specifically to infrastructure spending that is attached to contractual demand. The $678 billion in Microsoft’s commercial RPO and the $514 billion in Alphabet’s Cloud backlog are the anchors that justify the capital. Meta’s $130 billion capex plan for the year is not attached to an equivalent contractual foundation, and the market priced that distinction with precision this week.

The next question is already forming. Three Fed officials just voted to hike into an economy where the cloud is accelerating and AI adoption is genuinely commercial. The July jobs report will tell us whether the labour market is giving those three officials the cover they need to build a majority by September. And Jackson Hole will tell us whether Warsh intends to lead the committee or manage it. Both answers are coming within the next four weeks. This newsletter will be ready.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

Monday opened positively as chip stocks bounced from the prior week’s Kimi K3 sell-off. The Nasdaq jumped 1.3%, the S&P 500 gained 0.9%, and the Dow added 0.7%. Nvidia disclosed a stake in Nebius. Mediators proposed a 10-day ceasefire between the US and Iran, and Brent crude hovered around $90 a barrel, easing marginally from the prior week’s spike.

Tuesday reversed. Oil rose further to settle at $94.07, the highest level in over a month, after the eleventh consecutive round of US strikes against Iran. Secretary of State Marco Rubio said Iran was “not serious about talks”. The S&P 500 dipped 0.14% and the Nasdaq slipped 0.57%. After the close, Alphabet reported its Q2 results.

Wednesday extended the oil-driven pressure. Brent crude pushed higher again, and the broader market struggled. Alphabet’s stock fell despite its earnings beat as investors focused on capex concerns. Tesla reported after the close, missing EPS estimates by 39% and sending its stock down 14.4% the following session. Intel also reported after the close with its strongest quarter in fifteen years.

Thursday was the worst session of the week. Oil crossed $100 a barrel for the first time since early spring as the conflict entered its thirteenth consecutive night of US strikes. The Dow fell 550 points, the S&P 500 dropped 1.4%, and the Nasdaq sank 2.5%. The 10-year Treasury yield rose to its highest level since January 2025. Intel surged 12.4% on its results. Tesla fell 14.4%.

Friday saw a partial recovery. The Dow added 236 points, the S&P 500 inched higher, and the Nasdaq 100 lost 1.1% as oil settled below $98, easing slightly from its intraday peak but remaining sharply elevated on the week. For the week overall: the S&P 500 down approximately 1.2%, Nasdaq down approximately 2.2%, Dow roughly flat. The week ahead brings the July 29 FOMC decision, the most consequential since Warsh’s June debut, alongside Meta and Amazon earnings.

Week Ahead

Monday July 27

Markets open with the FOMC meeting beginning its two-day session. No decision until Wednesday, but positioning ahead of the most uncertain Fed meeting of the year begins immediately. Apple reports after the close alongside Visa, in what will be the first major consumer hardware and payments read since Apple’s price hike announcement in June and the subsequent surge in oil prices.

Tuesday July 28

Meta reports after the close. With Q2 results already in for Alphabet and pending for Amazon, Meta’s numbers will complete the picture on digital advertising health in Q2 and deliver the first formal update on Meta Compute’s commercial progress since the announcement in early July. Consensus expects approximately $60 billion in revenue, up 33% year-on-year. The operating margin trajectory, which compressed to 41% from the 48% peak in Q4 2024, is the detail most closely watched. S&P Case-Shiller home price data and Conference Board consumer confidence also land.

Wednesday July 29

The Federal Reserve announces its rate decision at 2 p.m. ET, followed by Kevin Warsh’s press conference. With Brent crude having touched $100 this week, June PCE still pending on Friday, and a committee that the June minutes described as a family fight, the July meeting is the most consequential since Warsh took the chair. Markets are currently pricing a hold, but the oil shock materially changes the inflation picture the committee is responding to. Amazon reports after the close.

Thursday July 30

The Employment Cost Index for Q2 drops, the most comprehensive measure of labour cost inflation available and a key input into the Fed’s services inflation assessment. Initial jobless claims also land. Merck, Mastercard, and Caterpillar report, giving insights on pharmaceuticals, consumer spending velocity, and industrial demand heading into August.

Friday July 31

June PCE lands at 8:30 a.m. ET, the final major inflation data point of the month and the one that will confirm or complicate the Fed’s Wednesday decision in retrospect. With oil back near $95, headline PCE is now almost certain to show re-acceleration from the June reading. Core PCE is the number that matters: whether the oil shock is bleeding into core services inflation will determine how much room Warsh has to hold rates steady at the August 26 to 27 Jackson Hole symposium without losing credibility.

Alphabet’s $514 Billion Cloud Backlog

For eighteen months, the central question hanging over the AI trade has been whether the extraordinary capital being deployed into AI infrastructure is generating proportionate revenue. Alphabet’s Q2 2026 results, reported Tuesday evening, are the most comprehensive available answer to that question from any company in the world.

Total revenue: $119.8 billion, up 24% year-on-year, the twelfth consecutive quarter of double-digit growth. Operating income: $40.8 billion at a 34% operating margin, two percentage points above the prior year. Google Search revenue: $63.3 billion, up 17%, despite every prediction over the past two years that AI-native search competitors would erode Google’s dominance. YouTube advertising: $11.1 billion, up 13%. These are not the numbers of a company losing ground. They are the numbers of a company whose core businesses are growing faster in absolute dollar terms than they were before the AI transition began.

Google Cloud is where the AI investment is most directly visible as revenue. Cloud revenue was $24.8 billion, up 82% year-on-year, more than eleven percentage points above the 64% growth analysts had forecast and an acceleration from the 63% recorded in the prior quarter. Cloud operating income reached $8.8 billion, up from $2.8 billion a year earlier, a 213% increase that reflects the operating leverage beginning to emerge as the business scales past breakeven. The cloud backlog at quarter-end was $514 billion, up more than $50 billion sequentially. That backlog is contractual: it represents signed customer commitments to purchase Google Cloud services over future periods. It is not a wish list. It is a binding revenue pipeline.

The number that caused the stock to fall despite the beat is full-year capital expenditure guidance, raised to $195 to $205 billion from the prior range of $180 to $190 billion, with Q2 capex of $44.9 billion alone representing a 100% year-on-year increase. Free cash flow turned negative for the quarter at minus $5.9 billion. Alphabet raised $49.6 billion by issuing stock in June and $20.3 billion in senior unsecured notes during Q2 to fund this spending, a financing structure that tells you Alphabet’s management believes the returns on this capital will exceed the cost of the debt.

The bull-bear divide on Alphabet is cleanly visible in the earnings reaction. The bears see $44.9 billion in quarterly capex producing negative free cash flow and ask whether the returns will ever justify the investment. The bulls see $24.8 billion in cloud revenue growing at 82% with a $514 billion backlog and a 213% increase in cloud operating income and answer that they already are. Both sides are looking at the same numbers. The difference is in how they weight the forward returns against the current cash cost.

For long-term investors, the Alphabet quarter settles the AI monetisation question at least for Google Cloud. Eighty-two percent growth against a base that was already $13.6 billion a year ago is not a number produced by pilot programmes and cautious enterprise trials. It is a number produced by large-scale commercial deployments at hyperscale customers who have concluded that AI infrastructure spending is generating measurable returns on their own businesses. The $514 billion backlog is the customers’ vote on the durability of that demand. CEO Sundar Pichai’s comment that nearly 90% of the Fortune 100 are using Gemini Enterprise is the commercial adoption data point that underpins that vote.

Tesla’s 39% EPS Miss

Tesla reported Q2 2026 results Wednesday evening. Revenue of $28.24 billion beat the $26.43 billion consensus by 6.8%. Record deliveries of 480,126 vehicles. Energy storage deployments are rising strongly year-on-year. On every volume metric, Q2 was a success.

The profitability picture was a different story entirely. Non-GAAP EPS of $0.33 missed the $0.54 consensus by 39%, the largest earnings miss in several quarters. Operating income fell 57% year-on-year to $398 million. Operating margin compressed to 1.4% from 4.1% a year ago. Automotive gross margin fell to 16.8% from 19.4%. Free cash flow turned negative at minus $1.1 billion, the first negative quarter in more than two years. Capital expenditure surged 142% to $5.8 billion. The stock fell 14.4% the following session, wiping approximately $140 billion in market value.

Understanding what happened requires looking at where the money is going. Operating expenses climbed 47% to $4.35 billion, driven almost entirely by research and development spending on Cybercab, Optimus, and full self-driving infrastructure. The Cybercab, Tesla’s two-seat autonomous robotaxi, has begun production at Gigafactory Texas. Elon Musk confirmed on the earnings call that the robotaxi service is now operating in seven major metropolitan areas. Optimus, the humanoid robot, is in early production ramp, with Musk acknowledging on the call that “everything on the robot is new” and “there is no existing supply chain” for it. Tesla’s full 2026 capital expenditure guidance is above $25 billion, roughly triple its historical spending level.

The parallel to Nike in Issue 57 is instructive but has an important distinction. In Nike’s case, the headline beat rested on a one-time tariff refund item that inflated reported EPS with no operational substance behind it. Tesla’s situation is the inverse: the headline miss reflects genuine, intentional, real operational spending on a future business that does not yet generate revenue proportionate to its cost. One is an accounting artefact. The other is a deliberate strategic decision whose payoff timeline is measured in years rather than quarters.

The honest analytical question is not whether Tesla’s investment is real. It clearly is. The question is whether CyberCab and Optimus will generate returns sufficient to justify the margin compression required to build them and on what timeline. The largest order backlog since 2023 and the seven-city robotaxi deployment are the early evidence that demand exists. The 142% capex surge and the negative free cash flow are the current cost of building the supply side to meet it. Long-term investors need to be clear-eyed about both and about the fact that Tesla’s current valuation of approximately 340 times earnings prices in an outcome that has not yet been delivered.

This newsletter has applied the same discipline here that it applied to Nike and to Micron: look through the headline number to what it actually reflects. In Tesla’s case, the miss is real, but its cause is strategic rather than operational deterioration. Whether that distinction justifies the current valuation is a question each investor must answer for themselves, with a full understanding of what the numbers actually mean.

Intel’s 25% Revenue Growth: The Comeback Continues

Intel reported Q2 2026 results Wednesday evening, and the numbers deserve far more attention than they received in a week dominated by Tesla’s miss and Alphabet’s capex. Revenue of $16.1 billion, up 25% year-on-year, was the strongest quarterly revenue growth Intel has reported in more than fifteen years. It beat the $14.42 billion consensus by approximately 12%. The stock rose 12.4% the following session, in the same week the Philadelphia Semiconductor Index was falling.

CEO Lip-Bu Tan described the result directly: “Our Q2 results represent our strongest revenue growth in more than fifteen years, enabled by greater speed, accountability, and customer focus.” CFO Dave Zinsner confirmed the company exceeded its own financial guidance on what he called “robust demand and improved execution, including volume upside driven by higher factory yields and improved cycle times”.has been Q3 guidance of $15.8 to $16.8 billion was issued alongside the results.

The most significant disclosure in the release was the update on Intel’s 18A manufacturing process, the node that CEO Lip-Bu Tan has staked his turnaround on since taking the helm. Yields on 18A have now risen to approximately 85%, up from 65% last quarter. In semiconductor manufacturing, yield improvement of that magnitude in a single quarter is a genuine operational milestone, representing both the reliability of the process and the economics of production moving into ranges where external customers can commit to volume orders with confidence.

The customer confirmation matters even more. Intel announced that a major cloud service provider has confirmed as the first external customer for 18A in volume production. The identity was not disclosed, but separate reporting has confirmed it is a hyperscaler operating at the scale of Google, Amazon, or Microsoft. In addition, Apple and Microsoft have been confirmed as 18A design partners, meaning both companies are actively designing chips intended to be manufactured on Intel’s most advanced node. ASML separately confirmed that Intel is the first company globally to deliver high-volume logic chips on High-NA EUV lithography, the most advanced photolithography technology currently available in commercial production.

The relevance to the broader semiconductor narrative this newsletter has tracked is direct. Issue 59’s semiconductor bear market section noted that Intel’s decline was partly valuation-driven and partly competitive-concern-driven, reflecting uncertainty about whether 18A could actually compete for advanced foundry customers against TSMC. Wednesday’s disclosure begins to answer that question. An 85% yield rate and a confirmed hyperscaler customer in volume production are not signs of a process still struggling. They are signs of a process approaching commercial maturity.

Intel’s turnaround is not complete. GAAP EPS remains negative. The foundry business is not yet at the scale where its economics fully cover its cost base. And competing with TSMC for the most advanced AI chip production from a standing start is a multiyear challenge that Wednesday’s results do not eliminate. But for long-term investors who tracked Intel’s difficulties through 2023 and 2024, and who watched the stock fall more than 30% from its recent highs alongside the broader semiconductor selloff in the past four weeks, Wednesday’s result changes the analytical baseline materially. The 18A story is no longer speculative. It is beginning to be operational.

Iran: A War Story

The June MOU has collapsed. As of Sunday, July 26, the US has now conducted thirteen consecutive nights of strikes against Iranian military targets. Brent crude, which had fallen to $71.57 on July 1, the lowest since before the conflict began, closed Friday below $98 a barrel after trading above $100 intraday on Thursday, a surge of more than 35% in three weeks. The Houthi rebels in Yemen have declared a maritime blockade on Saudi shipping and claimed attacks on Saudi oil tankers in the Red Sea. The Caspian Pipeline Consortium has suspended crude loadings at its Black Sea terminal following tanker attacks, disrupting approximately 80% of Kazakhstan’s oil exports. Asian buyers have begun discussing rerouting Saudi crude shipments through the Suez Canal and around Africa. The disruption is no longer confined to the Strait of Hormuz.

The human cost has also become visible in the data in a way that earlier escalations did not. At least four US service members have been killed in the most recent wave of Iranian strikes, according to the Pentagon. Iran’s health ministry reports 59 people killed and 666 injured inside Iran since June 27. Trump responded on Truth Social: “Every time Iran kills an American soldier, they will pay for that killing many times over.” He subsequently threatened “major military punishment” against Iran and the Houthis over Red Sea attacks and said he was considering a “massive attack” on Iran. Neither side has verbally signalled a return to diplomacy as of Sunday.

The diplomatic picture is fractured. Mediators from Qatar, Pakistan, and Iraq proposed a 10-day ceasefire to the parties on July 20, which a senior Iranian official confirmed Tehran had received. The New York Times reported Iran rejected the proposal. Iran’s foreign ministry denied that characterisation, calling the report “diversionary and misleading”. Iraq’s prime minister’s office also disputed it. A US official told CNN the rejection account was “not accurate”. What is accurate is that neither the US nor Iranian officials have publicly signalled a return to the table, and Secretary of State Rubio said this week that Iran is “not serious about talks”. The gap between what each side is saying publicly and what is happening in back-channel mediation is wide enough that the situation could shift quickly in either direction.

The economic transmission is already visible in the data. The 10-year Treasury yield rose to its highest level since January 2025 during Thursday’s session. Brent at $95 to $100 a barrel, if sustained through July, changes the arithmetic on the Fed’s inflation picture materially. The energy component of CPI fell 5.7% in June and was the single largest contributor to the headline deceleration this newsletter covered in Issue 59. That tailwind will reverse sharply in July’s data. Whatever relief the June CPI print provided to the Fed’s rate deliberations, the oil surge has partially undone it before the ink on the June numbers is dry. Chinese officials are reportedly increasingly concerned that attacks on Gulf states and Strait disruptions are damaging their economic interests, adding a new diplomatic dimension that was not present in earlier escalation cycles.

For long-term investors, the situation as of Sunday, July 26 is more serious than any prior point in this conflict’s inflation transmission story. This is no longer a ceasefire-monitoring exercise. It is an active war with no visible near-term diplomatic off-ramp, expanding geographic scope beyond the Strait into the Red Sea and the Black Sea, and a US president who has explicitly threatened a “massive attack”. The FOMC meeting opening Monday July 27 begins with an energy shock that the committee did not model in its June projections. Warsh’s family fight has a new variable that nobody on the committee priced in three weeks ago.

Closing Thoughts

Kevin Warsh walks into the July 29 FOMC decision with Brent crude having touched $100 intraday this week, a 10-year yield at its highest since January 2025, an active US military campaign now in its thirteenth consecutive night, an expanding geographic disruption that has moved from the Strait of Hormuz into the Red Sea and the Black Sea, and an earnings season that has confirmed AI infrastructure demand is accelerating while the capex required to sustain it is growing faster than anyone modelled.

Holding rates leaves the Fed open to criticism that it is tolerating an oil-driven re-acceleration in headline inflation without response, at a moment when the June CPI relief that appeared to validate patience has already been partially reversed by the energy market. Hiking into an active war, with consumer confidence fragile and Tesla’s investment cycle and Intel’s turnaround both at their most capital-sensitive moments, risks triggering a growth shock on top of an already deteriorating geopolitical backdrop. Neither option is comfortable. Neither can be made comfortable by the data available. That is the honest state of the FOMC meeting this newsletter is watching from.

What this newsletter has argued consistently is that the long-term investor’s response to this kind of uncertainty is to own businesses whose value compounds through rate environments in either direction and geopolitical shocks of any magnitude: businesses with pricing power sufficient to pass through cost increases, with balance sheets strong enough to absorb a tightening cycle, and with competitive positions durable enough that a year of margin compression does not change the ten-year outcome. Alphabet’s $514 billion cloud backlog and Micron’s $100 billion in SCA commitments are two examples of that kind of structural durability. They do not make the FOMC decision any easier, or the war any less serious. They make the question of what happens in the next three months somewhat less central to the long-term investment case than the headlines this week would suggest.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

All five major US banks reported Q2 results before the open simultaneously with June CPI. CPI came in at 3.5% year-on-year, well below the 3.8% consensus, with core CPI flat on the month, the softest monthly reading since January 2021. JPMorgan posted the highest quarterly profit in US banking history. Goldman Sachs nearly doubled its EPS. All five banks beat estimates. Semiconductor stocks rallied 2.5%. The S&P 500 gained 0.38%, and the Nasdaq rose 0.9%. IBM fell 25% after warning that enterprise clients are actively shifting spending from software and mainframes toward AI infrastructure.

Wednesday continued the momentum. The S&P 500 gained a further 0.38% to 7,572, the Nasdaq rose 0.62%, and BlackRock posted record results. After the close, Netflix reported Q2 revenue of $12.56 billion, up 13.4% year-on-year and in line with estimates, but Q3 guidance of $12.86 billion missed the $13 billion consensus, sending the stock lower.

Thursday delivered TSM’s record quarter. TSM’s ADR fell 1.55% after hours anyway, as Moonshot AI’s Kimi K3 announcement began circulating.

Friday was the worst session of the week. China’s Moonshot AI unveiled Kimi K3, a 2.8 trillion-parameter open-weight model, at the World AI Conference in Shanghai. The Philadelphia Semiconductor Index fell into bear market territory, down more than 20% from its June 22 record high. The index lost 10% for the week, its worst weekly performance since April 2025. TSM fell 7% on the day it reported a record quarter. Micron, which crossed $1 trillion in market cap in late May, fell below that threshold again. SoftBank fell 9%. For the week: S&P 500 down 1.6%, Nasdaq down 2.9%, Dow down 0.9%.

Week Ahead

Monday July 21

A lighter open before the week’s heavy earnings flow begins. Markets will still be processing Friday’s semiconductor selloff and the Kimi K3 announcement, with pre-market sentiment likely shaped by any further commentary from hyperscalers or chip companies on AI spending commitments. NXP Semiconductors reports after the close, an early read on automotive and industrial chip demand heading into the core earnings week.

Tuesday July 22

Alphabet and Tesla both report after the close. Alphabet’s results will be the most important AI monetization read of the quarter: how much of Google Cloud’s AI infrastructure investment is generating revenue, whether Search is holding its market position against AI-native competitors, and what YouTube’s advertising growth says about consumer attention in the current environment. Tesla’s delivery numbers are known, but management commentary on autonomous vehicle progress and energy storage demand will drive the stock reaction. Intuitive Surgical and Lockheed Martin also report.

Wednesday July 23

Meta reports after the close. After announcing Meta Compute and absorbing significant stock volatility on the news, this will be the first formal earnings disclosure that puts numbers behind its $145 billion capex commitment. Analyst attention will focus on AI-driven advertising revenue growth, Reality Labs losses, and any update on Meta Compute’s commercial progress. Boeing reports the same morning, giving a read on the aerospace supply chain. ServiceNow also reports, a bellwether for enterprise AI software adoption.

Thursday July 24

Amazon reports after the close alongside Merck and Hasbro. Amazon’s results will add the final major hyperscaler data point to the AI infrastructure picture: AWS growth rate, capital expenditure guidance, and management commentary on AI demand will all be read against the backdrop of Kimi K3 and the broader question of whether the compute build-out is generating proportionate returns. The preliminary S&P Global PMI readings for July also drop, giving the first read on economic activity in the month after the semiconductor selloff.

Friday July 25

The primary focus will be on digesting Alphabet, Meta, and Amazon’s combined AI commentary from the prior three evenings, which will collectively form the most comprehensive single-week read on the state of AI monetization yet available. June PCE also drops on Friday, the final major inflation data point before the July 28-29 FOMC meeting.

The Banks Had Their Best Quarter in History

On Tuesday morning, the five largest US banks reported simultaneously and produced a combined result that has no precedent in modern financial history. JPMorgan Chase posted net income of $21.2 billion, the highest quarterly profit ever recorded by a US bank. Goldman Sachs reported diluted EPS of $20.98, nearly double the prior year, on revenue of $20.34 billion, a 39% increase. Bank of America’s profit rose 27% to $9.1 billion. Wells Fargo posted EPS of $2.00 against a $1.72 consensus. Citigroup delivered results ahead of estimates across the board.

The drivers were consistent across all five: record or near-record trading revenues fuelled by geopolitical volatility, a surge in investment banking activity led by SpaceX IPO fee income distributed across all five banks as co-underwriters, and consumer credit quality that remained materially better than feared given the inflation environment. JPMorgan CFO Jeremy Barnum said that “consumers and small businesses remain resilient, with consumer spend growth continuing above last year’s pace.” The bank’s provision for credit losses fell to $2.5 billion, roughly half a billion less than analysts had expected.

The trading revenue story deserves specific attention. Market volatility generated by the Iran conflict, the Fed’s hawkish pivot, and the repeated AI model release shocks has been, for Wall Street trading desks, a source of extraordinary revenue rather than a source of risk. Every geopolitical shock, every inflation surprise, every semiconductor selloff has generated the kind of client activity and spread income that trading desks exist to capture. The banks are, in a meaningful sense, beneficiaries of the same uncertainty that has made long-term investing more difficult for individual investors this year.

IBM’s 25% fall on the same morning put the contrast in sharpest possible relief. The company warned that Q2 profits will be lower than expected due to soft demand in its software and infrastructure businesses, explicitly citing a shift in enterprise spending toward AI infrastructure. Clients are not cutting IT budgets. They are redirecting them. The money that used to flow into IBM’s mainframe software and hybrid cloud services is being redeployed into AI compute, AI infrastructure, and the new generation of AI-native software platforms that are beginning to replace legacy enterprise tools. IBM is not being disrupted by a better competitor in its own category. It is being disrupted by the emergence of an entirely new category that its clients now consider more important.

For long-term investors, the bank earnings and IBM’s warning together tell a story about the structural nature of the current AI transition. It is not simply a technology upgrade cycle where incumbents evolve and the market grows. It is a spending reallocation cycle where dollars are actively leaving some categories, legacy software, traditional infrastructure, and mainframe computing, and concentrating in others. The companies sitting in the path of that reallocation, whether as recipients or as those being displaced, are experiencing the most consequential repricing of the current market environment. Understanding which side of that reallocation your portfolio sits on is one of the most important analytical questions a long-term investor can ask right now.

TSM’s Record Quarter

If there is a single data source in the public markets that provides the most reliable, granular, real-time read on global AI hardware demand, it is TSM’s quarterly earnings release. The company manufactures the most advanced chips for Nvidia, AMD, Apple, Broadcom, and virtually every other name in the AI supply chain. When TSM’s revenue grows, AI hardware production is growing. When TSM raises its full-year outlook, customers are committing to more volume. The numbers cannot be gamed by narrative, because they are not narrative. They are physical wafers produced at prices set by supply and demand.

With that context, Thursday’s release deserves to be read for what it actually says rather than how the market reacted to it on a day when Kimi K3 was simultaneously announced.

Revenue: $40.2 billion, up 36% year-on-year, a single-quarter record. Net profit is up 77.4% year-on-year, also a single-quarter record, extending TSM’s consecutive streak of double-digit profit growth to nine quarters. Gross margin: 67.7%, above the company’s own guidance ceiling and above analyst expectations, reflecting the pricing power that comes from holding approximately 73% of the global advanced foundry market with no credible near-term challenger for the customers that matter most. EPS of $4.31 per ADR. Q3 guidance: $44.6 to $45.8 billion, implying approximately 37% year-on-year growth.

The revenue mix shift is where the structural story is clearest. High-Performance Computing, the segment anchored by AI accelerators for cloud data centers, rose 20% sequentially in a single quarter and now accounts for 66% of total wafer revenue, up from under 50% as recently as 2024. Smartphones, which generated the largest share of TSM’s revenue as recently as 2022, have fallen to 22%. Advanced process nodes at 7 nanometers and below represented 77% of all wafer revenue. The 3-nanometer node alone accounted for 30% of wafer revenue, and the 5-nanometer node for 33%.

TSM also raised its full-year capital expenditure guidance to approximately $18.75 to $20 billion, a significant increase from prior guidance, and committed an additional $100 billion to its Arizona fabrication facilities. CEO C.C. Wei said: “Our conviction in the multi-year AI megatrend remains very high.” CoWoS advanced packaging, a critical process for assembling AI chips, is reportedly sold out through year-end.

The stock fell 7% on its earnings day. The reason had nothing to do with the earnings. It had everything to do with Kimi K3 and the question the market could not resolve in a single afternoon: if efficient open-weight models can be built without incremental US chip purchases, does the demand trajectory TSM is reporting require revision? The answer this newsletter gives in the next section is no. But the important point here is that TSM’s numbers themselves, the revenue, the margin, the mix shift, and the guidance are the strongest available public evidence that AI hardware demand was still accelerating through the end of June. Whatever happens from here, that is the baseline the market is now discounting.

The Semiconductor Bear Market in Context

The Philadelphia Semiconductor Index entered bear market territory on Friday, defined as a 20% or more decline from a recent peak. From its record close on June 22, the index has now fallen more than 20% in less than four weeks. For the week alone it fell 10%, its worst weekly performance since the April 2025 tariff sell-off.

Before drawing any conclusions, the full-year context matters: the index is still up approximately 65% year-to-date. That 65% gain and the subsequent 20% correction are not contradictory. They are the mathematical consequence of a sector that ran 105% from its March low to its June peak in a period measured in months rather than years. When a sector moves that far, that fast, corrections of 20% or more are not unusual. The question is whether this particular correction is a crowded-trade unwind, a genuine fundamental reassessment, or something in between.

The individual stock moves reveal which part of the semiconductor complex has been hit hardest. Marvell Technology, ARM Holdings, and Intel have each fallen more than 30% from their recent peaks. These are three very different businesses with very different exposure to the AI cycle. Marvell and ARM are both direct AI beneficiaries, with Marvell’s custom silicon and networking chips and ARM’s processor architecture sitting at the center of the AI hardware build-out. Their 30%-plus declines are straightforwardly valuation-driven: both stocks had risen to multiples that priced in growth trajectories that were going to be very difficult to sustain even in the most bullish scenarios.

Intel’s situation is structurally different. The company has been undergoing a manufacturing turnaround while simultaneously trying to rebuild its competitiveness in AI chips, where it has fallen significantly behind Nvidia, AMD, and even custom silicon competitors. Intel’s decline reflects a compound of valuation compression, competitive concerns, and the broader sector rotation, but the underlying business case is less straightforward than for ARM or Marvell.

Micron has fallen approximately 30% from its late-June peak of approximately $1,213, briefly dipping below the $1 trillion market cap threshold it crossed just weeks ago. This is the most instructive individual stock story in the current selloff because the Micron fundamental case has not changed. The company’s FQ4 guidance of approximately $50 billion in revenue still stands. Its Strategic Customer Agreements covering $100 billion in remaining performance obligations are still binding contracts. HBM supply is still sold out through the end of 2026. The selloff in Micron is not a response to any change in the underlying business. It is a response to multiple compressions in the sector, driven by questions about AI demand sustainability that Micron’s own financials do not support.

The key distinction for long-term investors is between names where the selloff is correcting genuine overvaluation and names where the selloff is creating genuine undervaluation relative to a fundamentally unchanged business case. Both exist in the current correction. Separating them requires the same work this newsletter has always argued for: understanding the specific demand drivers, the contractual underpinnings, and the competitive position of each business rather than treating the semiconductor sector as a monolithic trade in either direction.

One useful frame: the VanEck Semiconductor ETF has now posted three weekly declines in four weeks, losing nearly 25% from its peak. During the same period, TSM reported a record quarter, Micron maintained $50 billion in Q4 guidance, Samsung reported a 19-fold profit increase, SK Hynix raised $26.5 billion in the largest foreign listing in US history, and Micron committed $250 billion in US manufacturing investment through 2035. The sentiment and the fundamentals are not moving in the same direction. That gap is where long-term investing decisions get made.

Kimi K3: The New DeepSeek Moment

On Friday, July 17, China’s Moonshot AI unveiled Kimi K3 at the World Artificial Intelligence Conference in Shanghai. The model carries 2.8 trillion parameters, making it the largest open-weight AI model in the world at the time of its release. Moonshot claimed it performs comparably to Anthropic’s Claude Fable 5 and substantially outperforms OpenAI’s GPT-5.6, with particular strength in front-end coding. Because it is open-weight, developers can download, study, and run it themselves without paying inference fees to any US company.

The market’s reaction was immediate and severe. The Philadelphia Semiconductor Index fell more than 5% on the day. Traders explicitly called it a “new DeepSeek moment,” drawing the comparison to January 2025, when DeepSeek’s R1 model triggered a similar selloff by demonstrating that frontier AI performance could be achieved at dramatically lower compute cost than US models had suggested. TSM fell 7% on the same day it reported a record quarter. SoftBank fell 9%. Global semiconductor stocks shed approximately $3.3 trillion in market value.

The original DeepSeek moment was in early 2025, and the framework applied then is worth revisiting now because the mechanism is identical. The market’s fear is that efficient open-weight models reduce the demand for compute by enabling AI capabilities to be deployed at lower cost, which in turn reduces the revenue trajectory for chip manufacturers and the returns on hyperscaler infrastructure spending. The logic, at its core, is: if you need less compute per AI task, you will buy fewer chips.

The counterargument, which history has consistently supported, is that cheaper AI capability expands the total addressable market for AI rather than contracting it. When AI inference costs fall, the number of applications that are economically viable increases. When a business can deploy an AI coding assistant for 95% less cost using an open-weight model, it does not stop there. It deploys ten more AI tools it could not previously justify. The aggregate demand for compute therefore rises even as the cost per unit of capability falls. This dynamic has played out across every major efficiency gain in computing history. More efficient transistors did not reduce demand for chips. Cheaper cloud computing did not reduce demand for servers. More efficient AI models have not, so far, reduced demand for AI compute.

There are also specific reasons to be cautious about the Kimi K3 claims specifically. The model’s benchmark performance has not been independently verified. Moonshot is a venture-backed startup with strong incentives to make its model sound as capable as possible at launch. The original DeepSeek moment triggered a severe market reaction, and in the months that followed, the actual impact on hyperscaler AI spending was essentially zero: Alphabet, Meta, Microsoft, and Amazon all raised their AI capex guidance rather than reducing it. The demand for US AI infrastructure did not fall because a Chinese model was efficient. It kept accelerating.

What does seem genuinely relevant about the Chinese AI competitive picture is the longer-term trend rather than any single model release. Bloomberg Intelligence data cited this week shows that Chinese firms plan to allocate 46% of AI accelerator budgets to domestic suppliers over the next 12 months, up from 30% currently, as Beijing pushes a nationwide data center build-out worth approximately $295 billion over five years. That is not a Kimi K3 story. It is a structural story about the bifurcation of the global AI supply chain, and it is a more substantive competitive concern than any individual model’s benchmark claims. It is also a trend that plays out over years, not sessions.

For long-term investors, the appropriate response to a Kimi moment is the same as the appropriate response to a DeepSeek moment: examine the specific claims carefully, apply historical precedent to the demand question, and resist the impulse to draw permanent conclusions from single-session selloffs. The semiconductor bear market that has developed over the past four weeks is real. The valuation reset it represents in some names is rational. But the underlying demand for advanced AI compute, as confirmed by TSM’s record quarter on the same day as Friday’s selloff, has not changed because a Chinese startup released a large open-weight model at a conference in Shanghai.

Closing Thoughts

This week produced data that this newsletter believes will eventually be remembered as among the most significant of 2026. TSM’s record quarter confirmed that AI hardware demand was accelerating through the end of June. The bank earnings confirmed that the financial sector is generating extraordinary returns from the volatility and deal flow that the AI transition is creating. IBM’s warning confirmed that the transition is actively disrupting incumbent enterprise technology. And Kimi K3 reminded markets that China is not a passive observer of the AI race.

None of these facts contradict each other. They are all simultaneously true. The semiconductor selloff has created a market environment where the fundamentals and the prices are, for a period, moving in opposite directions. That is not a new dynamic. It is the oldest dynamic in investing. What changes from one cycle to the next is the specific names, the specific catalysts, and the specific valuations at which the divergence resolves.

Next week brings Alphabet, Meta, and Amazon within a three-evening window. Together they will provide the most concentrated single-week read on AI monetisation yet available in any public earnings season. Whether those results confirm or challenge the demand picture that TSM’s revenue numbers imply will be the question that shapes the second half of July. This newsletter will be watching closely.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

A Week Of Reversals

Monday was a strong open. The Dow reached a fresh record above 53,000 for the first time, the S&P 500 gained 0.72%, and the Nasdaq added 1.12%. Big Tech led the recovery from the prior week’s selloff. Foxconn’s stronger-than-expected quarterly sales, reported over the weekend, provided an early positive signal on AI hardware demand.

Tuesday turned. The semiconductor complex sold off broadly, with the VanEck Semiconductor ETF falling more than 3%. Samsung’s record quarterly profit figure, released early in the Seoul session, initially prompted fears of a cycle peak rather than celebration. Oil prices began climbing again as Iran attacked a tanker near the Strait of Hormuz. The Nasdaq fell 1.16%.

Wednesday was the worst session of the week. Trump told the NATO summit in Turkey that the ceasefire with Iran was “over”, oil surged more than 5%, European markets fell sharply, and the Dow dropped 576 points. The FOMC minutes from the June meeting were also released in the afternoon, revealing a committee split more evenly between those favouring hikes and those preferring to hold than the dot plot alone had suggested.

Thursday recovered. Micron announced a major acceleration of its US manufacturing commitment, the Nasdaq gained 1.30%, and the semiconductor complex bounced. Oil pulled back from its Wednesday peak as de-escalation signals re-emerged. For the week overall: the S&P 500 ended roughly flat, the Nasdaq slipped marginally, and the Dow managed a small gain. The Roundhill Memory ETF fell more than 10% for the week despite Micron’s announcement, as Samsung’s profit report and the renewed Iran tensions competed for the market’s attention.

The week ahead opens Tuesday with June CPI alongside Kevin Warsh’s testimony before the Senate Banking Committee, the most important combination of inflation data and Fed communication since the June FOMC meeting. Q2 bank earnings from JPMorgan, Bank of America, Citigroup, and Wells Fargo all report Tuesday before the open.

Week Ahead

Monday July 14

A light open on data. Markets will be positioning ahead of Tuesday’s simultaneous combination of June CPI and Warsh’s Senate testimony, which together represent the most concentrated single-morning policy signal since the June FOMC meeting itself.

Tuesday July 15

June CPI lands at 8:30 a.m. ET. This is the first inflation print to fully capture the energy price easing that followed the initial Iran ceasefire framework: Brent crude averaged meaningfully lower in June than in April or May. Headline CPI should fall from May’s 4.2%. Core CPI is the number that matters for the Fed’s July 28-29 decision. At 10 a.m., Kevin Warsh testifies before the Senate Banking Committee, his first appearance before Congress as Fed chair. The combination of a fresh inflation print and Warsh’s live commentary in the same morning session will make Tuesday one of the most watched data days of the summer. JPMorgan, Bank of America, Citigroup, and Wells Fargo all report Q2 results before the open, making Tuesday the single heaviest data and earnings morning of the month.

Wednesday July 16

June retail sales land alongside industrial production and capacity utilisation data for June. Retail sales will give the clearest read yet on whether consumers began pulling back in June, the first full month after Apple’s price hike announcement and the second consecutive month of negative real wages. Goldman Sachs and Morgan Stanley report Q2 results before the open. ASML reports from Europe, providing the most granular read on semiconductor equipment demand available in any earnings release.

Thursday July 17

Initial jobless claims drop alongside the Philadelphia Fed Manufacturing Index for July. Netflix reports after the close, the first major streaming earnings of Q2 and a useful read on consumer subscription spending in an environment of compressed discretionary budgets. Taiwan Semiconductor Manufacturing also reports, providing the clearest available data on advanced chip production volumes heading into the second half of the year.

Friday July 18

The preliminary University of Michigan Consumer Sentiment Index for July drops. After two months of negative real wages, renewed Iran tensions, and Apple price hikes, this number will tell us whether the consumer is beginning to internalise the inflation and rate environment in their confidence and spending intentions.

Iran Ceasefire, Round Two: Trump Calls It Over at NATO

At the NATO summit in Turkey on Wednesday, President Trump said the ceasefire with Iran was “over.” His words: “I think it’s over. I don’t want to deal with them anymore. They’re scum.” He later threatened to “hit them hard tonight.” Oil surged more than 5%, European stock markets fell sharply, and the Dow dropped nearly 600 points.

By Thursday, the familiar pattern reasserted itself. De-escalation signals re-emerged, oil pulled back from its peak, and markets partially recovered. Diplomats noted that US-Iran talks remained ongoing in Doha, and that Wednesday’s statements from Trump were more likely to reflect frustration with Israeli strikes on Hezbollah in Lebanon than a formal termination of the ceasefire framework itself.

This is now the fourth time since February that markets have cycled through escalation fear and de-escalation relief in the Iran conflict. And something has shifted in how markets are pricing each iteration. The oil price response to Wednesday’s escalation, while significant at 5%, was smaller than the equivalent moves in March or May. The equity selloff, while sharp intraday, was partially recovered the same day. The VIX rose but did not spike. The pattern itself is becoming visible, and markets are beginning to price it as such.

The question of whether that desensitisation is rational or dangerous is genuinely unresolved. The case for rational adjustment is straightforward: each previous escalation has been followed by de-escalation, the Strait has remained at least partially open throughout, and energy prices have trended lower from their peak despite the repeated flare-ups. Markets updating their probability of a full-scale war downward and reducing their sensitivity to individual incidents accordingly is not obviously wrong.

The case for dangerous complacency is equally straightforward. The ceasefire framework that was signed in France is contested and fragile. The Strait of Hormuz mine clearance is still incomplete. The harder negotiations on sanctions, enrichment, and proxy networks have not advanced. And Israel, which is not party to the agreement and has continued striking Hezbollah in Lebanon, remains a wildcard that neither Washington nor Tehran can fully control. A single incident involving a US or allied vessel could escalate in ways that the market’s current pricing does not reflect.

For long-term investors, the lesson from this pattern is not primarily about Iran. It is about how to think about geopolitical risk more broadly. Markets consistently underweight tail risks that have not yet materialised and overweight recent precedent. Every previous de-escalation makes the next escalation feel less likely. That dynamic can persist for a long time. Until it does not. The appropriate response is not to position for a worst case that may never arrive but to own businesses whose long-term value is not contingent on the Iran situation resolving cleanly and to maintain enough liquidity that a genuine deterioration creates opportunity rather than distress.

The FOMC Minutes

Kevin Warsh described the June FOMC meeting internally as a “family fight”. The minutes released Wednesday confirmed that characterisation in detail.

The committee is genuinely split. “Many participants” saw the appropriate year-end rate as within or slightly below the current target range, suggesting a preference for holding. “Many other participants” saw it finishing the year above the current range, implying at least one hike. “A few participants” were ready to hike in June itself but agreed to hold given the uncertainty. The minutes confirmed that the unanimous vote to hold masked a committee nowhere near consensus on the path ahead.

Three specific inflation drivers dominated the internal discussion: AI-related investment demand, the ongoing Middle East conflict, and tariffs. The AI driver is the one that has received least attention in mainstream coverage and deserves the most. Multiple participants noted that the sheer scale of AI infrastructure spending, the $690 billion in hyperscaler capex committed for 2026 and the associated demand for semiconductors, data centre power, and construction labour, is itself generating inflationary pressure through investment demand. This is a fundamentally different inflation mechanism from the energy shock or tariffs. It is not a supply disruption raising prices temporarily. It is demand-driven, structural, and directly correlated with the continued acceleration of AI spending that every company and investor in the space is treating as a positive.

The implication is uncomfortable: the same AI infrastructure build-out that is driving extraordinary earnings at Micron, Dell, Broadcom, and Nvidia is also, in the committee’s own analysis, contributing to the inflation that makes rate hikes more likely. The Fed’s own minutes now explicitly name AI investment demand as an upside inflation risk. That is a signal worth taking seriously regardless of how Tuesday’s CPI prints.

Warsh’s deliberate absence from the dot plot is the other detail the minutes illuminate. He confirmed in his press conference that he chose not to submit a projection, and the minutes reveal why: Warsh has a fundamental philosophical objection to central bank forward guidance, believing it reduces flexibility and creates unnecessary market dependence on Fed communication rather than underlying data. The minutes confirm this was not a procedural decision but a deliberate signal about how he intends to run the committee. Markets accustomed to parsing every word of post-meeting statements for clues about the next move will need to recalibrate. Under Warsh, the data itself, not the Fed’s interpretation of the data, is intended to do the signalling.

Tuesday’s testimony before the Senate Banking Committee will be the most extensive public discussion of these dynamics since the June meeting. With June CPI landing simultaneously, the morning session will effectively be Warsh’s first public opportunity to respond to a fresh inflation print in real time. That combination of data and live commentary makes Tuesday the single most concentrated policy signal of the summer.

Micron’s $250 Billion Commitment and SK Hynix’s Historic US Debut

On Thursday, Micron announced it is raising its planned US investment to $250 billion through 2035, up from $200 billion committed just twelve months ago. The company poured the first concrete at its Clay, New York facility, which will be the largest semiconductor manufacturing plant in US history. A separate $3 billion commitment to strengthen the US supply chain included $500 million in financing to GlobalWafers for its Texas operations alongside a ten-year raw silicon wafer supply agreement, domesticating a critical raw material that the US semiconductor industry has historically sourced almost entirely from overseas. Micron also signed a long-term strategic memory supply agreement with Ford.

The $250 billion figure, taken over its nine-year horizon, represents one of the largest manufacturing investment commitments in the history of American industry. The goal: to produce approximately 40% of Micron’s DRAM output domestically, a target that would fundamentally change the geographic concentration risk in the global memory supply chain. CEO Sanjay Mehrotra said the investment would create more than 90,000 jobs and put Micron at the centre of the US government’s semiconductor sovereignty agenda under the CHIPS Act framework.

This arrived simultaneously with SK Hynix’s US debut on the Nasdaq, which raised approximately $26.5 billion in the largest foreign listing in American history. SK Hynix simultaneously committed $8.6 billion to acquire advanced EUV lithography equipment from ASML, accelerating its own manufacturing capacity expansion. Samsung, for its part, guided to a record approximately $73 billion in 2026 semiconductor investment and is lifting its HBM production capacity toward 250,000 wafers per month by year-end.

Taken together, these three announcements in a single week represent the clearest evidence yet of what this newsletter has called the memory supercycle since Issue 49. Micron, SK Hynix, and Samsung are each making nine-figure manufacturing commitments on the basis of demand visibility that they believe extends well beyond the current fiscal year. These are not reactive capacity additions based on spot market signals. They are structural bets on a sustained, multi-year supercycle driven by AI infrastructure demand.

The Ford supply agreement deserves specific attention because it extends the memory supercycle narrative beyond the data centre and into the automotive sector. Micron’s automotive memory revenue grew 311% year-on-year in fiscal Q3, driven by advanced driver assistance systems and in-vehicle AI processing that require orders of magnitude more memory than previous generations of automotive electronics. The domestication of that supply chain, with Ford committed to long-term Micron supply and Micron committed to a $500 million investment in domestic silicon wafer supply, is a structural shift in how the US automotive industry sources its most critical AI-enabling components.

For long-term investors, the $250 billion commitment changes the investment calculus for Micron in a specific way. Capital expenditure at this scale, sustained over nine years, creates a manufacturing moat that is very difficult for competitors to replicate quickly. It also creates a US government stakeholder relationship that, in a world of increasing semiconductor geopolitical competition, provides Micron with a form of strategic protection unavailable to overseas competitors. The risk is that demand does not sustain at the levels required to justify the capital deployment. The Strategic Customer Agreements covering approximately $100 billion in remaining performance obligations are the primary evidence that demand visibility is genuine rather than assumed.

Samsung’s 19-Fold Profit Surge Triggered a Selloff.

Samsung Electronics reported a 19-fold year-on-year increase in quarterly operating profit on Tuesday. The number, approximately $58 billion in annualised terms, is one of the most extraordinary single-quarter profit figures in the history of the global semiconductor industry.

South Korea’s Kospi fell 4.91% on the news. The VanEck Semiconductor ETF fell more than 3% in the US session that followed.

The market’s interpretation was almost unanimous: a result this extraordinary must represent the peak. If Samsung earned 19 times more than a year ago, the comparison base is now so high that future growth rates must inevitably compress. What has happened has been priced in. What comes next, by definition, cannot be as good. The selloff was a classic peak-earnings reaction, the same dynamic that sent Broadcom down 12% on record AI chip revenue in early June.

This newsletter observed this dynamic in Issue 56 in the context of the Kospi as a leading indicator. It is worth revisiting with more precision. Samsung’s profit surge was driven almost entirely by HBM memory pricing, which has risen dramatically as AI data centre demand for high bandwidth memory has vastly exceeded the available supply from all three major manufacturers simultaneously. The question the sell-off was really asking is not “Is this as good as it gets?” in an abstract sense. It is asking a specific structural question: is HBM pricing at or near its peak, and if so, when does the margin compression begin?

The evidence available this week cuts both ways. On the side of peak pricing: Samsung itself acknowledged that consumer DRAM prices are beginning to stabilise after months of extraordinary gains. Bernstein noted a modest rise in standard DRAM and NAND contract prices, suggesting the rate of increase is moderating. On the side of sustained pricing, HBM supply from all three major manufacturers remains fully committed through at least the end of 2026; Micron’s $250 billion investment announcement confirmed that the structural demand case extends through 2035; and Samsung’s own guidance to $73 billion in 2026 semiconductor investment signals its own confidence in sustained elevated demand.

The more useful framework for long-term investors is not whether this quarter’s profit represents the peak. It is whether the structural demand drivers, AI model complexity growing each generation, and each generation requiring substantially more HBM and hyperscaler capex committed at $690 billion for 2026 alone are durable enough to support a cycle measured in years rather than quarters. Micron’s SCA structure, covering $100 billion in remaining performance obligations with binding customer commitments, is the strongest currently available evidence that at least one major manufacturer and its largest customers believe the answer is yes.

The Seoul leading indicator point from Issue 56 also proved relevant this week in both directions. Tuesday’s Kospi selloff on Samsung’s results front-ran the US semiconductor decline by several hours, giving investors who monitor Korean market openings a useful early read on the direction of US chip stocks at the open. Thursday’s Micron-driven recovery in Seoul similarly preceded the US semiconductor bounce. The pattern continues to hold: for investors with material exposure to the AI memory trade, the Korean market’s opening session is worth monitoring as a directional guide before NYSE opens.

Closing Thoughts

This week asked two questions that are in fundamental tension with each other, and it asked them simultaneously.

The first question: is the AI infrastructure investment cycle durable enough to justify the scale of capital now being committed? Micron’s $250 billion through 2035, SK Hynix’s $26.5 billion US debut, and Samsung’s $73 billion 2026 semiconductor investment plan, these are nine-figure bets that the structural demand case holds. The people making them have more information about their order books and customer commitments than any external observer.

The second question: is the AI infrastructure investment cycle itself generating inflationary pressure that will eventually force the Fed to act in ways that compress the valuations of the very companies driving it? The FOMC minutes named AI investment demand as an explicit upside inflation risk alongside Iran and tariffs. If the Fed hikes because AI spending is overheating the economy, the discount rates applied to AI companies rise, and the valuations that justified the investment commitments get tested.

Both questions are live. Both have genuine uncertainty. And both will begin to get partial answers on Tuesday morning, when June CPI, Kevin Warsh’s Senate testimony, and the first major bank earnings of Q2 all land within the same two-hour window. This newsletter will cover all of it next week.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

A Holiday Week That Packed in a Full Month of News

Monday opened with a strong relief rally. The Nasdaq rose 2%, the S&P 500 gained 1.2%, and the Dow put on 0.6% to close above 52,000 for the first time in its history. Alphabet joined the Dow Jones Industrial Average that morning, replacing Verizon, and rose more than 4% on its first day as a component. A Supreme Court ruling held that Federal Reserve governor Lisa Cook would remain in her position, rejecting the Trump administration’s attempts to remove her, with the court carving out an explicit exception for central bank independence. On the Iran front, easing tensions helped lift markets after the US and Iran agreed to stop tit-for-tat attacks, with Trump announcing the two nations would meet in Doha on Tuesday for renewed talks. Tesla surged 8%, SpaceX gained 2.3%, and semiconductor names bounced sharply from the prior week’s selloff.

Tuesday brought the biggest structural AI story of the week. Bloomberg reported that Meta is planning to launch a cloud business, dubbed Meta Compute, to sell excess AI compute capacity to third parties. Meta shares surged on the news. CoreWeave and Nebius each fell approximately 12%, as the announcement signalled that one of their largest potential customers would now become a direct competitor. The same day, Rocket Lab announced it would acquire Iridium in an $8 billion deal, sending Iridium up 25%, Rocket Lab up 16%, and space sector names broadly higher.

Wednesday introduced fresh AI jitters. Reports emerged that OpenAI had begun discussions about selling a 5% equity stake to the US government via a proposed Public Wealth Fund. The news added to a broader reconsideration of AI valuations and capital allocation dynamics. Meta fell 5% and Tesla fell 7.5% despite reporting strong delivery numbers. Micron sank 7%, Applied Materials fell 7.4%, and AMD dropped 4.3%. The S&P 500 dropped 0.4% and the Nasdaq 100 slipped more than 2% as chipmakers sold off for a second consecutive session.

Thursday delivered the week’s defining number. June nonfarm payrolls came in at 57,000, well below the 115,000 consensus and revised down from the prior two months by a combined 74,000. The soft print immediately removed September from the rate hike timeline. The Dow climbed 595 points to a new record at 52,844 as Apple gained 4.8%, Walmart rose 3%, and McDonald’s gained 4%. The S&P 500 was roughly flat and the Nasdaq 100 fell 0.8% as chip names continued to sell off. For the week: S&P 500 up 1.8%, Nasdaq up 2.1%, Dow up 2%. Markets were closed Friday for Independence Day.

Week Ahead

July 7 to 11: CPI, FOMC Minutes, and Bank Earnings

Monday July 7

Markets reopen after the Independence Day holiday. SpaceX joins the Nasdaq 100 on this date, following Nasdaq’s confirmation of its fast-tracked addition. The addition will trigger forced buying from index-tracking funds, which could provide a near-term technical tailwind for SPCX shares that has nothing to do with the underlying business fundamentals. A light economic calendar makes positioning for the week’s heavier events the primary activity of the session.

Tuesday July 8

JOLTS job openings for May land, the last major labour market read before the week’s main events. Consumer credit data for May also releases. PepsiCo reports before the open, giving one of the first major consumer staples reads of the second half earnings season and a useful read on whether pricing power is holding against the inflation backdrop.

Wednesday July 9

June CPI lands at 8:30 a.m. ET. This is the first inflation print that should begin reflecting the easing in energy prices following the Iran ceasefire framework. May’s headline CPI was 4.2%, driven overwhelmingly by the energy shock. With Brent crude now trading in the low $70s after falling from above $100, the energy component should pull headline CPI meaningfully lower. Core CPI is the number that matters most for the Fed’s rate decision calculus. The FOMC minutes from the June meeting, Kevin Warsh’s first as chair, also drop Wednesday. The minutes will give a behind-the-scenes read on the committee’s internal deliberations that the shortened post-meeting statement deliberately obscured.

Thursday July 10

Major bank earnings begin. JPMorgan Chase, Wells Fargo, and Citigroup all report before the open. Bank earnings are one of the most reliable leading indicators for credit conditions, loan demand, and the health of the consumer and corporate balance sheet. In the current environment, with rate hike odds having shifted dramatically following Thursday’s jobs report, bank commentary on net interest margin outlook and loan growth will be particularly closely watched. Delta Air Lines also reports, offering an early read on travel demand heading into peak summer season.

Friday July 11

June PPI drops alongside the University of Michigan’s preliminary July Consumer Sentiment reading. Producer prices will give a forward-looking read on whether goods inflation has genuinely peaked or whether the supply chain pressures flagged by Apple’s price hike announcement are still working their way through. Goldman Sachs and BlackRock report, completing the first week of the second half earnings season.

Jobs And Interest Rates

Three consecutive months of stronger-than-expected jobs data had built a near-ironclad case for a Fed rate hike before year end. April came in at 148,000 revised. May came in at 129,000 revised. Both were well above consensus at the time of their initial release. Heading into Thursday, the market was pricing a meaningful probability of a September hike, nine of eighteen FOMC members had already pencilled one in on the dot plot, and Kevin Warsh’s own inaugural press conference had been read as hawkish.

Then June arrived at 57,000.

The Bureau of Labor Statistics reported that the US economy added just 57,000 jobs in June, below the revised 129,000 in May and less than half the 115,000 consensus. It was the weakest single month of job creation since early 2025. Prior months were revised down simultaneously: April cut by 31,000 from 179,000 to 148,000, and May cut by 43,000 from 172,000 to 129,000, removing 74,000 positions previously reported. The household survey was even weaker, with 507,000 fewer people reporting themselves at work, a figure that sent the labour force participation rate down 0.3 percentage points to 61.5%, its lowest level since March 2021.

The breakdown offers some structural context. Professional and business services added 36,000, social assistance added 25,000, and healthcare rose by 22,000. These are the stable, secular-growth categories that have driven job creation consistently through the past two years. What collapsed was leisure and hospitality, down 61,000, a reversal the BLS attributed to weaker than usual seasonal hiring, a pattern that analysts have since connected to the FIFA World Cup effect. Seasonal adjustment models expect certain hiring patterns from the hospitality sector in June each year. In 2026, those patterns were distorted by World Cup-related hiring in May and early June, with the adjustment pulling forward jobs that did not materialise in the usual way when measured against the seasonal baseline.

Average hourly earnings rose 0.3% for the month and 3.5% year-on-year, both in line with forecasts. Wage growth is holding steady but not accelerating, which is precisely the combination the Fed needs to avoid the services inflation spiral that would make a rate hike unavoidable regardless of the headline jobs number.

The market’s read was immediate and decisive. Following the jobs print, traders took a potential September hike off the table. Futures still point to a possible October increase, but that probability has fallen meaningfully. Jefferies senior economist Thomas Simons put it plainly: “For the Fed, this number is fine. The pace of job growth is plenty strong enough to maintain a steady unemployment rate and average hourly earnings are solid, but not accelerating. There is no imperative on their part to do anything with rates immediately, and the softening in the pace of job growth suggests that rate hikes are very unlikely to be necessary this year.”

Warsh himself has repeatedly said he will not provide forward guidance and is “not committed to any type of policy path.” That posture, combined with a June jobs report that removes urgency without signalling deterioration, gives the Fed exactly the space it needs to stay on hold through the summer and let the inflation picture clarify. Wednesday’s CPI and the FOMC minutes will tell us whether the committee is inclined to use that space or push through with the hike the dot plot suggested.

Meta Compute

Meta’s announcement this week that it is launching a cloud business to sell excess AI compute capacity was, on its surface, a story about Meta’s capex strategy. Underneath the surface, it is one of the most significant structural shifts in the AI infrastructure market this year, and it has direct implications for every investor with exposure to neocloud names like CoreWeave, Nebius, or any company whose business model depends on hyperscalers being consumers of compute rather than sellers of it.

The background: Meta has committed $145 billion in capital expenditure for 2026, the largest single-year infrastructure investment in the company’s history. That spending has built a network of data centres and GPU clusters that is, by some estimates, larger than what the company currently needs for its own AI model training and inference workloads. CEO Mark Zuckerberg first signalled the possibility of a cloud move at the Q3 2025 earnings call. Bloomberg’s report this week confirmed that the initiative, named Meta Compute and led by infrastructure head Santosh Janardhan alongside Meta Superintelligence Labs leader Daniel Gross and president Dina Powell McCormick, is now operational rather than theoretical.

The competitive implications are direct and serious. CoreWeave and Nebius each fell approximately 12% on the news. The logic is straightforward: Meta entering the market as a seller of raw compute capacity creates supply that competes directly with the inventory CoreWeave and other neocloud providers have built at enormous cost. CoreWeave’s entire business model rests on the premise that there is a gap between what hyperscalers provide and what AI-focused customers need, a gap filled by purpose-built GPU cloud infrastructure from specialised providers. If Meta is now filling part of that gap itself, using capacity it has already paid for, the addressable market for CoreWeave and its peers contracts.

There is a more constructive read on this development, and it is worth taking seriously. Meta’s decision to sell excess compute rather than simply absorb the cost of unused capacity is evidence that the hyperscaler build-out has genuinely overshot near-term demand in at least one corner of the market. That is not a new concern. CoreWeave’s own guidance miss in May and OpenAI’s revenue shortfalls earlier in the year both pointed in the same direction. What Meta’s announcement adds is the clearest confirmation yet that even the most aggressive capex spenders have built more than they currently need, and are now trying to recoup those costs through commercial channels rather than waiting for internal demand to catch up.

For long-term investors, the Meta Compute announcement asks a question this newsletter has framed repeatedly since Issue 48: is the $690 billion committed to AI infrastructure in 2026 creating genuine, durable, proportionate demand, or is it creating a supply glut in certain categories that will eventually require a painful realignment? The answer, based on this week’s evidence, is more nuanced than either the bulls or bears are acknowledging. Demand is real. Supply has overshot demand in the near term. The mechanism by which supply excess is resolved, through commercial monetisation like Meta Compute, through write-downs, or through demand eventually catching up, will determine which names in the AI infrastructure ecosystem compound through this period and which ones do not.

Nike, and the Tariff Refund.

Nike reported fiscal Q4 2026 earnings on Tuesday, and the headline numbers looked extraordinary. Diluted EPS of 72 cents crushed the 13-cent consensus. Gross margin expanded 890 basis points to 49.2%. Net income surged 407% year-on-year to $1.1 billion.

Strip out the one-time item and the picture changes entirely.

The 72 cents per share included a 52-cent benefit from the expected recovery of IEEPA tariff refunds, a $986 million one-time accounting item related to the anticipated repayment of import duties collected under International Emergency Economic Powers Act provisions. Remove that item and Nike’s underlying EPS was 20 cents. The 890 basis point gross margin expansion was almost entirely attributable to the same tariff refund, which contributed roughly 900 basis points on its own. Without it, gross margin was essentially flat.

The operational numbers underneath the tariff benefit told a consistent story of a company still in the middle of a difficult turnaround. Q4 revenue came in at $10.97 billion, down 1% reported and down 4% on a currency-neutral basis. Greater China sales fell 12% to $1.30 billion. Nike Direct revenues fell 7%, with digital sales down 12% and Nike-owned stores down 7%. Converse revenue dropped 32% to $244 million, down across all territories. Full-year revenue of $46.4 billion was flat on a reported basis and down 2% on a currency-neutral basis.

CEO Elliott Hill was direct about where things stand. “We continue to face top-line headwinds. Overall, the results aren’t there yet.” Outgoing CFO Matthew Friend offered guidance expecting revenue to decline low to mid-single digits in the near term, with Q2 of fiscal 2027 sequentially weaker due to prior year anomalies. Earnings were expected to be “flattish” through the next three quarters, excluding any further tariff recovery benefit.

The one bright spot in the quarter was performance categories. Running, Football, and Basketball all grew in the mid-single digits, suggesting that the brand’s core athletic credibility remains intact even as the lifestyle and streetwear segments, particularly Sportswear and Jordan, continue to face sell-through challenges in both the US and China.

This newsletter returns to a principle that runs through every earnings analysis it publishes: headline beats that rest on one-time items are not the same as operational momentum. The distinction matters enormously for long-term investors trying to assess whether a turnaround is real or whether it is being masked by accounting items that will not recur. In Nike’s case, the tariff refund boosted a single quarter’s EPS by 2.6 times the underlying result. Investors who read the headline number and concluded that Nike’s recovery is accelerating drew a very different conclusion than those who read through to the currency-neutral revenue decline and the 12% fall in China sales.

The turnaround case for Nike remains credible over a longer horizon. The performance category momentum is real. The brand retains extraordinary recognition globally. The new management team is making structural changes to the marketplace and supply chain that have a multi-year payoff timeline. But the turnaround is not yet showing up in the operational numbers in a way that would justify confidence that the most difficult phase of the reset is behind them. That is the honest read on a quarter the headlines described as a beat.

OpenAI’s Public Wealth Fund Proposal

On July 2, the Financial Times reported that OpenAI is in talks to voluntarily hand the US government up to 5% of its equity, worth approximately $42.6 billion at the company’s current valuation of roughly $852 billion, via a new structure called the Public Wealth Fund. The proposal, championed by CEO Sam Altman, would place that equity block into a sovereign-wealth-style fund designed to distribute AI-generated returns directly to US citizens, rather than routing AI profits through the tax system in the conventional way.

This arrived in the same week that Altman reportedly rejected a lower-valuation IPO path in favour of waiting until 2027 for a listing at or above $1 trillion, and in the same week that SoftBank closed a second $10 billion tranche of its planned $30 billion follow-on investment in OpenAI. The timing is not coincidental. It reflects a company navigating an increasingly complex set of pressures simultaneously: regulatory scrutiny from the DOJ and FTC, antitrust questions about its market dominance in consumer AI, political scrutiny about its relationship with the Trump administration, and the capital markets reality that its most recent private valuation of $730 to $850 billion is well below the $1 trillion target Altman insists on for any public listing.

The mechanics of the proposed fund raise questions the reporting leaves largely unresolved. Whether the 5% stake vests immediately or is contingent on milestones such as an IPO or revenue thresholds. Who controls the voting rights, and whether a passive Treasury holding would function differently from an active government board seat. How the fund would distribute returns, whether through per-capita citizen dividends, retirement account contributions, or infrastructure spending. And whether a government equity stake in the most commercially dominant AI lab in the world creates novel antitrust complications that have no precedent in modern regulatory history.

For investors watching the broader AI capital markets story, the Public Wealth Fund proposal has two readings in tension. The optimistic read is that Altman is making a sophisticated regulatory arbitrage move, pre-emptively creating a political constituency among US citizens who would benefit from OpenAI’s commercial success, thereby reducing the probability of aggressive antitrust action and smoothing the path to a $1 trillion IPO in 2027. The more skeptical read is that OpenAI is facing enough regulatory and competitive pressure that it has determined proactive political positioning is less costly than the alternative.

What the proposal unambiguously signals is that Altman does not believe a straightforward commercial IPO at the valuation he is targeting is achievable in the current environment without some form of structural political accommodation. That is a meaningful statement about where AI valuations and public market appetite actually stand. Anthropic, which has filed confidentially and is widely expected to pursue a listing in late 2026, will be watching closely to understand what precedent the OpenAI-government relationship sets for its own regulatory positioning. And any long-term investor considering AI exposure through the IPO wave that was supposed to define 2026 now has a more complicated picture to navigate than the straightforward growth story the pre-IPO hype suggested.

Closing Thoughts

The first half of 2026 closed with an S&P 500 that had navigated an Iran conflict, a new Fed chair, accelerating inflation, a record IPO, and one of the most extraordinary earnings seasons in semiconductor history. It delivered a positive total return despite all of it, which is itself a remarkable statement about the resilience of corporate earnings as an anchor in a volatile macro environment.

The second half opens with the same questions the first half could not fully resolve. Is the Iran ceasefire durable or is the Strait of Hormuz the next point of escalation? Will Warsh hold rates or hike, and does a 57,000 job print change the calculus that nine of his committee members had already committed to? Is the AI infrastructure build-out generating returns proportionate to the spending, or is Meta Compute the first visible crack in the assumption that every dollar of hyperscaler capex translates into lasting competitive advantage?

Wednesday’s CPI will be the first data point that begins to answer the inflation question for the second half. The FOMC minutes will tell us whether the committee is leaning toward using the space the soft jobs print has given them. And Nike’s quarter, read carefully rather than at the headline, is a reminder that the discipline of looking through one-time items to the operational reality underneath them is the work this newsletter exists to do.

The edge is built one week at a time. The second half has just begun.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

Supplier Vs Consumer

Monday opened with Big Tech under pressure as markets reopened after the Juneteenth holiday. Alphabet fell 5%, SpaceX declined for the third consecutive session after its post-IPO peak, and the broader large-cap technology complex sold off. The Dow gained 0.2% while the Nasdaq fell 1.3% and the S&P 500 dropped 0.4%. Iran-US peace talks in Switzerland reported “encouraging progress” with both sides agreeing to a roadmap for a final deal within sixty days, and Brent crude fell further towards $78 a barrel.

Tuesday extended the losses. Bank of America issued a note raising its rate hike probability, and Asian markets collapsed overnight, with South Korea’s Kospi triggering an intraday circuit breaker before closing down 5.8%, its worst single session in over two years. The semiconductor complex bore the brunt in both Seoul and New York. The VanEck Semiconductor ETF fell more than 5% for the week overall.

Wednesday was Micron’s day. After the close, Micron reported fiscal Q3 2026 results that defied almost every model on Wall Street. The stock rose 14.55% in after-hours to $1,199.52. During the session, Brent crude fell 4.33% to $73.74, its lowest level since before the Iran conflict began in February, and the 10-year Treasury yield fell below 4.5%.

Thursday was Apple’s day and a brutal one. Apple announced price increases of up to $500 on MacBooks and iPads, citing “unprecedented” memory cost pressures driven by AI data centre demand. Apple fell 6.12%, its worst single session since April 2025, wiping approximately $265 billion in market value. The move triggered broader technology selling, with the AI price hike narrative spreading to other consumer electronics names. May PCE also landed: headline 4.1% year-on-year, the highest since April 2023; core 3.4% year-on-year, slightly above the 3.3% forecast. Q1 GDP final estimate came in at 2.1%, revised up from 1.6%. Personal income rose a stronger-than-expected 0.7%.

Friday closed the week with OpenAI’s IPO delay report. The New York Times reported that OpenAI is leaning towards delaying its IPO to 2027, with CEO Sam Altman reportedly rejecting any valuation below $1 trillion and advisers citing SpaceX’s post-IPO volatility, from a peak of $225.64 to an intraday low of $147.11 in just three sessions, as evidence that market appetite for mega-cap AI listings is more limited than previously assumed. Chip stocks fell again.

For the week: S&P 500 down nearly 2%, Nasdaq down 4.6%, Dow up 0.6%.

Week Ahead

Monday June 29

ISM Manufacturing PMI for June drops at 10 a.m. ET. After weeks of inflation data and macro noise, this gives the first read on factory activity for the month in which the Iran ceasefire framework was signed and oil prices fell sharply. Nike reports after the close.

Tuesday June 30

JOLTS job openings for May land, the first labour market read of the week ahead of Thursday’s payrolls data. Constellation Brands and General Mills report, adding consumer staples context to the week. Eurozone CPI also drops, which will be watched for signs of whether the energy shock from the Iran conflict has spread beyond US borders and at what pace it is now easing.

Wednesday July 1

ADP private payrolls drop alongside the ISM Services PMI for June, the last major preview before Thursday’s NFP. Both will be read closely against Warsh’s hawkish dot plot from the June FOMC meeting. Any sign that the labour market is softening, or that services activity is decelerating, would complicate the case for the rate hike nine FOMC officials projected for later this year.

Thursday July 2

The June Employment Situation report drops at 8:30 a.m. ET. This is the most consequential data point of the week. The prior two months, April at 172,000 and May at 179,000 revised, both nearly doubled their respective consensus estimates. A third consecutive strong print would make the case for the Fed’s first rate hike since the current tightening cycle began almost inarguable, regardless of how far energy prices have fallen since the Iran ceasefire. Markets close early at 1 p.m. ET ahead of the Independence Day holiday.

Friday July 4

US markets are closed for Independence Day. Investors head into the long weekend with the first half of 2026 behind them and a jobs report, a contested Iran ceasefire, a hawkish Fed, and an extraordinary earnings season all in context simultaneously.

Micron’s Quarter That Rewrote the Memory Playbook

Revenue: $41.46 billion. Up from $23.86 billion in the prior quarter. Up from $9.30 billion a year ago. That is a 346% year-on-year increase in a single quarter for a company that was already the world’s third-largest memory manufacturer. Non-GAAP gross margin reached 84.9%, the highest in the company’s history. Operating income reached $33.68 billion on a non-GAAP basis, an operating margin of 81.2 per cent. Non-GAAP EPS of $25.11 beat the $20.49 consensus by 22.6%. Adjusted free cash flow reached $18.3 billion for the quarter alone. Cash on the balance sheet at quarter end stood at $24.99 billion, up from $9.64 billion just nine months earlier.

Q4 guidance of $50 billion in revenue with a gross margin of approximately 86% was more than double what most Wall Street models had projected for this time last year and meaningfully above the $43.58 billion consensus. If achieved, Micron’s fiscal full-year revenue will exceed approximately $170 billion, a figure that would make it one of the fastest top-line expansions in the history of the semiconductor industry.

The segment breakdown is where the structural story lives. Cloud Memory Business Unit: revenue of $13.77 billion, gross margin of 83%, and operating margin of 78%. Core Data Center Business Unit: revenue of $11.52 billion, gross margin of 87%, and operating margin of 83%. Mobile and Client Business Unit: revenue of $11.52 billion, gross margin of 87%, operating margin of 86%. Automotive and Embedded: revenue of $4.63 billion, gross margin of 79%, operating margin of 75%. Every segment delivered operating margins above 75%. That is not a memory company in a cyclical upturn. That is a memory company that has structurally repriced its products across every customer category simultaneously.

The Strategic Customer Agreement disclosure is the detail that changes the long-term investment thesis most profoundly. Micron announced it has now signed sixteen SCAs with data centre operators, automakers, and other enterprise customers. For those SCAs with defined pricing, the company disclosed remaining performance obligations of approximately $100 billion at quarter end. Those obligations include binding commitments to purchase volumes of Micron’s chips over multi-year periods. The financial commitments from customers amount to $22 billion in deposits and advance payments. CEO Sanjay Mehrotra said on the call that when completed, he expects approximately half or more of the company’s revenue to be under these agreements. CFO Mark Murphy called it “transformational”.

The SCA structure is worth understanding in detail for long-term investors because it changes the fundamental risk profile of Micron as a business. The memory industry has historically been one of the most brutal in technology: hyper-cyclical, with periods of extraordinary profitability followed by periods of severe losses as supply caught up with and then exceeded demand. What the SCA structure does is remove a meaningful portion of that revenue from the spot market entirely. When half of Micron’s revenue is under multi-year, binding, volume-committed, fixed-or-floored-price agreements with hyperscalers and automakers, the quarterly swings that have historically defined this industry are removed. That is not a minor change. It is a business model transformation with direct implications for valuation, earnings predictability, and investor confidence in long-term forecasting.

HBM remains fully booked. Micron said tight supply conditions are expected to persist beyond calendar 2027. HBM4, built on 1-beta DRAM technology, is in high-volume shipments for the lead customer’s platform, with qualification samples shipped to multiple additional customers. HBM4E development, built on 1-gamma DRAM technology, is well underway, with volume production expected in calendar 2027. The progression from HBM4 to HBM4E, and the capacity commitments already secured for each generation, is the clearest available evidence that the structural demand cycle this newsletter identified has not peaked. It is still in its early expansion phase.

Apple’s Price Hikes and What They Confirm

On Thursday June 25, Apple raised prices on MacBooks, iPads, iMacs, HomePods, and Apple TV. The MacBook Air rose from $1,099 to $1,299. The MacBook Pro from $1,699 to $1,999. The iPad Air from $599 to $749. The Mac Studio with M3 Ultra jumped from $3,999 to $5,299, a 33% increase overnight. The average increase across affected products landed at approximately $246. The iPhone lineup was not touched. Apple’s stock fell 6.12%, its worst single session since April 2025, wiping approximately $265 billion of market value in a day. Microsoft raised Xbox console prices the same day, citing memory cost increases of more than 2.5 times and projecting another doubling by fall 2027.

Apple’s statement was explicit about the cause: “The rapid expansion of AI data centres has created an extraordinary surge in demand for memory and storage. We have never seen a component price increase this much, this quickly.” That framing, from the company that negotiates the largest consumer electronics memory contracts in the world, is not a complaint. It is a confirmation of the structural argument at the centre of this newsletter’s memory supercycle.

The iPhone exclusion is almost certainly temporary and strategic. IDC estimates that all new iPhone models moving to 12GB of RAM, required to support the full Apple Intelligence feature set, combined with the underlying DRAM cost increase, could add roughly $200 per unit to Apple’s bill of materials. Analysts at Counterpoint Research expect iPhone price increases of $150 to $200 across the lineup for the September iPhone 18 launch, weighted more heavily towards higher-memory configurations. IDC senior director Nabila Popal was unambiguous: “The iPhone isn’t spared; a price hike is coming. The storm isn’t over yet; this is just the beginning.” CEO Tim Cook will step back from the CEO role in September to become executive chairman, meaning the iPhone 18 pricing decision will likely be his last major product pricing call in the role.

Counterpoint Research director Tarun Pathak put the broader picture plainly: “Apple held it off for at least two quarters, protecting its user base from price inflation, but it has reached the point where it could no longer absorb the cost increase. We believe the situation is unlikely to improve for at least the next two years.” TrendForce data shows DRAM prices increasing up to 98% in the first quarter of 2026, with a further 58% to 63% increase expected in the current quarter.

For long-term investors, the Apple price hike announcement is significant for a reason that goes beyond Apple’s own stock price reaction. It is the clearest possible confirmation that the memory supercycle has crossed the boundary between AI infrastructure and consumer products. When the most powerful consumer hardware company in the world, with decades of supply chain leverage and the strongest supplier relationships in the industry, cannot absorb the cost increase and must pass it to customers, the companies on the supply side of that trade are in a structurally advantaged position. Micron reported the same week. The connection is not coincidental. It is causal.

For Apple itself, the long-term question is different. Price increases of this scale risk volume compression, particularly in markets outside the United States where Apple’s pricing is already at a premium to local alternatives. The MacBook Neo’s price increase specifically removes the one Apple product that was genuinely competitive on price with Windows and Chromebook alternatives. That is a strategic concession at the entry level that could have durable consequences for market share among younger, price-sensitive first-time buyers.

OpenAI Blinks. And Seoul Told You First.

By the time US markets opened on Tuesday, South Korea’s Kospi had already told the story. The index fell sharply enough intraday to trigger its circuit breaker mechanism, a decline that automatically halts trading to prevent panic selling, before closing down 5.8%. The dominant drivers: Samsung and SK Hynix, the two largest memory semiconductor companies in the world after Micron, sold off hard as the market reconsidered AI infrastructure valuations in the context of Apple’s anticipated price hike announcement and broader technology multiple compression.

This is worth understanding as a structural feature of how the memory and AI semiconductor trade works, not just as a one-week anecdote. South Korea’s stock market opens at 9 a.m. Seoul time, which is 8 p.m. the prior evening in New York. That means Korean market participants are digesting overnight US news, after-hours earnings releases, and early morning data before NYSE opens the following morning. For investors in Micron, Nvidia, AMD, or any company with significant exposure to the same semiconductor supply chain that runs through Samsung and SK Hynix, the Kospi and the performance of its major technology constituents can function as a leading indicator of how those names will trade when US markets open hours later.

In practical terms: when SK Hynix is down 7% in Seoul at midnight New York time, the probability that Micron opens lower in New York is meaningfully higher than it would otherwise be. When Samsung reports unexpected margin pressure, it tells you something about pricing dynamics that will eventually show up in Micron’s next quarter. The Korean market is not a perfect predictor, and there are company-specific factors that create divergence, but as a directional guide to the overnight sentiment shift in the AI memory trade, Seoul has earned attention from any long-term investor with exposure to the space. Monitoring Korean equity markets before the NYSE open is one of the most underappreciated informational edges available to retail investors who know what they are looking for.

This week illustrated the dynamic in both directions. Tuesday’s Kospi collapse front-ran the US semiconductor selloff by a full trading day. By Wednesday, when Micron reported results that obliterated every estimate, the after-hours move in MU immediately signalled a reversal in the narrative. By Thursday, when Apple announced its price hikes and the market reacted with a broad technology selloff, Seoul again gave an early read on the direction of travel.

The OpenAI IPO delay, reported Thursday evening, added a separate but related layer to the capital markets story. According to the New York Times, OpenAI’s advisers presented Sam Altman with two options: delay to 2027 at a target valuation of approximately $1 trillion, or list sooner at a lower valuation. Altman reportedly rejected any reduction from the trillion-dollar target, making the 2027 path the more likely outcome. The trigger for the reconsideration: SpaceX’s post-IPO trajectory. SPCX debuted at $135, rallied to a peak of $225.64, and then fell to an intraday low of $147.11 in the three sessions that followed, a 34% peak-to-trough move in days. For OpenAI’s bankers, that volatility was a data point about retail and institutional appetite for mega-cap AI listings that carried direct implications for OpenAI’s own offering.

The delay, if it holds, has cascading implications. SoftBank, which has approximately $65 billion committed to OpenAI and whose stock fell 13% on the news, had been counting on OpenAI’s IPO as a monetisation event that would validate its AI investment thesis. Anthropic, which has also filed confidentially for a US listing, must now recalibrate its own timing in the context of a market that has shown it will price even the most anticipated AI IPO with more scepticism than the pre-listing hype suggested. And for the broader AI trade, an OpenAI delay removes one of the largest potential sources of new AI-focused equity supply from the market this year, which is not uniformly negative for existing public AI names but does raise questions about what the delay signals about private market confidence in near-term AI monetisation timelines.

For long-term investors, the OpenAI story and the Kospi circuit breaker are two versions of the same question: how much future growth is already priced in to the AI trade, and what happens when the capital formation pipeline shows signs of friction? The answer this week was not reassuring for short-term holders. But for investors with a seven-plus-year horizon, weeks like this one, when the Nasdaq falls 4.6% and Seoul triggers a circuit breaker and OpenAI blinks on its IPO timeline, are when the businesses underneath the headlines can be evaluated more clearly than during the euphoria that preceded them.

May PCE and the Inflation Peak Question

May PCE came in at 4.1% year-on-year on the headline, the highest reading since April 2023, with core PCE at 3.4%, the highest since October 2023 and slightly above the 3.3% forecast. The monthly headline reading of 0.4% was below the 0.5% expected, while the core monthly reading of 0.3% matched forecasts. Services inflation accelerated to 0.5% monthly from 0.3% in each of the two prior months, the most concerning detail in an otherwise mixed report. Personal income rose a stronger-than-expected 0.7%, personal spending rose 0.7%, and the personal saving rate edged up to 3%.

The headline number matters because of what it likely confirms: May was almost certainly the peak of this inflationary cycle, and not because the Fed has done anything to bring it down. The energy shock from the Iran conflict drove the overwhelming majority of the headline acceleration. Brent crude has now fallen to $73.74, below where it was before the conflict began in February. That drop in oil prices does not yet appear in May’s PCE data, which measures prices paid through the end of May. It will begin showing up in June’s and July’s readings. The broader case for May as a peak is credible but not guaranteed. If services inflation continues at this pace through June and July, even a sharp drop in headline energy-driven PCE may not be enough to bring core comfortably below 3% by year’s end.

Deutsche Bank now expects two rate hikes this year, bringing the federal funds rate to 4.1%, with cuts not beginning until 2028. That call, which would have seemed extreme three months ago, is now within the range of mainstream forecasts. Nine of eighteen FOMC members already project at least one hike in Warsh’s own dot plot. The June jobs report on Thursday this week will be the next decisive data point: a third consecutive strong print would make the case for the Fed’s first hike almost inarguable, regardless of what happens to headline energy prices from here.

For long-term investors, the inflation picture heading into the second half of 2026 is more nuanced than either the bears or the bulls are acknowledging. The energy shock is easing, and headline PCE will likely fall meaningfully if the Iran ceasefire holds. Core inflation is stickier, driven by services and shelter, and is not responding to lower oil prices in any meaningful way. The Fed’s credibility question, how long it tolerates core PCE above 3% before acting, is what determines whether the second half of 2026 looks like 2023’s soft landing or something more disruptive. Warsh’s inaugural dot plot made clear which direction his committee is leaning. Thursday’s jobs number will tell us how much ammunition they have to act on it.

Closing Thoughts

The Same Trade, Seen From Both Sides

This week illustrated that the AI memory supercycle is not a single-company story, or even a single-sector story. It is a structural force that is now showing up simultaneously in Micron’s income statement, Apple’s retail pricing, South Korea’s stock exchange, OpenAI’s IPO timeline, and the Federal Reserve’s inflation projections.

Micron’s numbers, drawn directly from its SEC filing, are among the most extraordinary in semiconductor history. But they are also the mirror image of Apple’s price hike announcement. The same supply constraint that produced $41.46 billion in Micron revenue and an 84.9% gross margin is the same supply constraint that forced Apple to raise MacBook prices by $200 overnight. Both stories are true. Both have long-term investment implications. And they point in opposite directions depending on which side of the memory trade you are sitting on.

The OpenAI IPO delay and the KOSPI circuit breaker are the capital markets’ expression of the same underlying question the technology sector has been asking since Broadcom’s guidance miss three weeks ago: how much of the AI growth story is already priced in, and what happens to valuations when the capital formation pipeline shows friction? This newsletter does not have a confident short-term answer to that question. What it does have is a clear long-term framework: the businesses with structural, contractually embedded positions in the AI supply chain, Micron’s SCAs being the clearest current example, are materially better positioned to compound through that uncertainty than those dependent on spot market dynamics or IPO timing.

The second half of 2026 opens with a jobs report on Thursday, a contested ceasefire, a hawkish Fed chair with a dot plot that leans towards hiking, and a technology sector that has just had one of its most volatile fortnights since April 2025. That is not a comfortable starting position. It is, however, exactly the kind of environment in which the difference between businesses that compound through uncertainty and those that merely ride momentum becomes most clearly visible. That distinction is what this newsletter is built to identify, one week at a time.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

Monday delivered the surprise this newsletter has been tracking for weeks. The US and Iran announced a framework deal to end the war and reopen the Strait of Hormuz. Trump declared on social media that “the Deal with the Islamic Republic of Iran is now complete,” authorising the immediate removal of the US naval blockade. Markets rallied hard: the S&P 500 gained 1.9% and oil prices dropped almost 5%. By the close, the Dow had hit a record high while Brent crude sank 4.76% to $83.17, its lowest settle price since early March.

Tuesday brought a pause for digestion. Stocks edged higher ahead of Wednesday’s Fed decision, with the Dow notching another record while the Nasdaq and Russell 2000 backslid modestly as investors positioned for the main event.

Wednesday was Kevin Warsh’s first FOMC meeting as chair, and it delivered a genuine surprise of its own. The Fed held rates steady at 3.50% to 3.75% in a unanimous 12-0 vote, a marked change from the 8-4 split at the prior meeting under the previous chair. But the Summary of Economic Projections told a different story. The median dot for year end 2026 jumped to 3.8%, up from 3.4% in March, with nine of eighteen officials now projecting at least one rate hike before year end and six projecting two. Year end PCE inflation projections jumped to 3.6%, up sharply from 2.7% in March. Markets fell hard on the news, with the S&P 500 and Nasdaq posting their worst session of the week.

Thursday, the final trading day of the week, brought a recovery as markets reassessed the new information. The Nasdaq surged 1.91% to 26,517.93, the S&P 500 rose 1.08% to 7,500.58, and the Russell 2000 led with a 2.12% gain to close the week at 2,979.77, helped by a modest decline in Treasury yields even despite Wednesday’s hawkish tone. The formal signing of the US-Iran memorandum of understanding in France added to the relief, with energy stocks declining further as the agreement was made official.

Markets were closed Friday June 19 for the Juneteenth federal holiday, giving investors an extended weekend to process both the geopolitical resolution and the Fed’s new posture before trading resumes.

Come Saturday, and the Strait of Hormuz was closed by Iran again, a never-ending cycle.

Week Ahead

Monday June 22

Markets reopen after the long weekend with no major earnings or scheduled data, giving traders a full session to reposition around both the Iran deal and the Fed’s hawkish dot plot before the week’s heavier data and earnings flow begins.

Tuesday June 23

FedEx reports after the close its first earnings as a leaner company following the recent spinoff of its Freight division. As one of the closest things markets have to a real time read on global trade volume, FedEx’s commentary on shipping demand will be watched closely for signs of whether the Iran resolution is translating into renewed activity through previously disrupted trade routes. Carnival also reports.

Wednesday June 24

Micron reports after the close, in what is now one of the most closely watched earnings events of the month given the memory supercycle story running through this entire issue. May new home sales data and the Fed’s annual bank stress test results also land. Paychex and Jefferies also report.

Thursday June 25

The week’s defining release. May PCE, the Federal Reserve’s preferred inflation gauge, drops at 8:30 a.m. ET alongside the third and final estimate of Q1 2026 GDP and May durable goods orders. This is the first major inflation print to land inside Warsh’s new communication framework, and it will be read directly against the hawkish dot plot delivered Wednesday. April’s reading came in at 3.8% headline and 3.3% core. Darden Restaurants also reports, giving a consumer dining read alongside the inflation data.

Friday June 26

Nike closes out the week’s earnings, alongside Walgreens Boots Alliance. Nike’s results will offer a global consumer discretionary read at a moment when the macro backdrop, energy prices, and rate expectations have all shifted meaningfully within a single week.

The Iran Deal Was Signed. The Strait Just Closed Again.

For three months, this newsletter has tracked the Iran conflict as the single most important transmission mechanism between geopolitics and the inflation data driving Fed policy. This week, that thread reached a genuine turning point, and then, within 48 hours, a genuine complication.

The US and Iran announced a framework agreement on Monday to end the fighting and reopen the Strait of Hormuz, one of the world’s most critical oil transit routes. Trump’s language was characteristically emphatic: “I hereby fully authorize the toll free opening of the Strait of Hormuz, and, simultaneously herewith, authorize the immediate removal of the United States Naval blockade. Ships of the World, start your engines. Let the oil flow!” Iran’s Supreme National Security Council confirmed the deal had been reached “following a difficult and intensive period of negotiations lasting several months.” The agreement was formally signed Thursday in France, digitally, by Trump and Vice President Vance on the US side and Parliament Speaker Mohammad Bagher Ghalibaf on the Iranian side.

Markets responded immediately and decisively. Brent crude fell 4.76% on Monday to $83.17 a barrel, its lowest settle since early March. The S&P 500 gained 1.9% on the day. By Thursday, with the formal signing complete, energy stocks declined further as the agreement moved from framework to reality.

The substance of the deal, however, was always narrower than the celebration suggested. The agreement covers an end to sporadic attacks, the reopening of the Strait, and a sixty day window to negotiate the issues that actually determine whether this becomes a durable peace: sanctions relief, Iran’s uranium enrichment program, and the future of Iran’s regional proxy network including Hezbollah. Israel, notably, was not a party to the agreement, and Prime Minister Netanyahu’s first public comment was that he and Trump “do not always see eye to eye.”

Council on Foreign Relations analysis published this week frames the agreement as best understood as primarily a ceasefire rather than a comprehensive peace, one that could in principle be extended indefinitely while the harder negotiations continue. That framing has already proven prescient. On Saturday, just two days after the formal signing in France, Iran’s top joint military command announced the Strait of Hormuz was closed again, citing continued Israeli strikes on Hezbollah in southern Lebanon as a violation of the memorandum’s first clause, which called for an end to fighting on all fronts including Lebanon. Israeli strikes killed at least sixteen people in southern Lebanon overnight Saturday, hours after diplomats had briefed reporters on a separate Israel-Hezbollah ceasefire. Neither Israel nor Hezbollah are signatories to the US-Iran deal, and Israel has continued operating in southern Lebanon despite American pressure to stand down, with Vice President Vance publicly criticising Israeli officials on Thursday for what he characterised as a lack of appreciation for US support.

The market and policy response to Saturday’s announcement has been genuinely split. CENTCOM disputed that any closure is actually in effect, with spokesperson Captain Tim Hawkins stating that fifty five merchant vessels transited the strait on Saturday carrying more than seventeen million barrels of oil, and that “Iran does not control the Strait of Hormuz. Traffic continues to flow.” Vice President Vance said on Saturday he had seen no evidence the strait remains closed, though he acknowledged that mines in the waterway could still be disrupting some commercial routes regardless of Iran’s formal position. Iran’s Revolutionary Guard Navy, for its part, has warned ships not to approach the waterway and said vessel safety could not be guaranteed. Special envoys Steve Witkoff and Jared Kushner were already on the ground in Switzerland for technical talks on implementing the memorandum when the announcement came, with broader US-Iran talks now scheduled to resume Sunday in Burgenstock alongside mediators from Pakistan and Qatar.

This is precisely the dynamic this newsletter flagged when the deal was first announced. The agreement covers a halt to direct US-Iran hostilities and an initial reopening of the strait, but Israel was never a party to it, and the harder, unresolved issues, sanctions, enrichment, and the fate of Iran’s regional proxy network, were always going to determine whether the ceasefire holds. Less than a week after the framework was signed, the gap between what was agreed and what is actually happening on the ground in Lebanon has reopened the single channel, energy markets, through which this conflict has consistently transmitted into inflation data and Fed policy.

For long-term investors, the lesson from this week’s price action is a familiar one from this newsletter, and Saturday’s developments only reinforce it. The transmission mechanism from this conflict into headline inflation runs through energy, and Monday’s nearly 5% drop in oil prices showed exactly how directly that mechanism operates in one direction. Whether Saturday’s announced closure, contested as it is between Iran’s stated position and CENTCOM’s tracked shipping data, proves to be a temporary skirmish over Lebanon or the first sign that the broader ceasefire is unraveling will likely be clearer by the time markets reopen Monday. Either way, this newsletter’s core point from Issue 54 stands: a meaningful portion of the inflation and rate uncertainty embedded in asset prices is tied to a binary geopolitical outcome that can resolve or unresolve with almost no warning.

Warsh’s First Surprise: The Dot Plot Flips Hawkish

Markets expected a hold. They got one, and it was unanimous, a marked shift from the 8-4 split that defined the prior meeting under the previous chair. What markets did not fully price in was just how hawkish the rest of Wednesday’s communication would be.

The headline number is the dot plot itself. The median FOMC projection for the federal funds rate at year end 2026 jumped to 3.8%, up from 3.4% in the March projections, a quarter point above the current target range and an explicit signal that the committee now sees a rate hike, not a cut, as the more likely path for the rest of the year. Of the eighteen officials who submitted projections, nine saw the rate finishing above the current range, with three projecting a single quarter point hike and six projecting two. Just one official saw a cut. The Fed’s own inflation projection moved in lockstep: year end PCE inflation is now expected at 3.6%, sharply higher than the 2.7% projected back in March.

Equally significant was what the Fed removed from its communication. The post-meeting statement dropped language that had previously signalled a bias toward future rate cuts, and did so within a dramatically shortened statement overall. Warsh described the change directly in his press conference: “It’s a bit shorter, a bit simpler and it dispenses with some older language. That statement just gives you the facts, as best we can judge it.”

Warsh’s own remarks left little room for a dovish read. He called persistently high inflation “a burden for the American people” and was unambiguous about the committee’s priorities: “I am pleased to report that members of the FOMC are unambiguous and unanimous; this committee will deliver price stability.” Asked directly whether the Fed might reconsider its long-held 2% inflation target given how long it has run hot, Warsh’s answer was characteristically blunt: “The ‘two’ is the left of the decimal point. For now, ‘zero’ is to the right.” He added that he saw no reason to revisit the target “until we have reestablished our commitment and ability to deliver” on it.

One detail added an extra layer of intrigue. The published dot plot appeared to be missing one submission, and Fed watchers widely speculated the missing dot belonged to Warsh himself, who confirmed during the press conference that he had not provided a projection, though he said he encouraged his colleagues to do so. A new chair declining to put his own number on the board, in his very first meeting, while simultaneously delivering a more hawkish committee message than markets expected, is itself a signal worth sitting with. It suggests a chair still calibrating how directly he wants to commit himself publicly, even as he is comfortable letting the committee’s collective output run more hawkish than the market had priced.

History offers a relevant precedent here. Analysis this week noted that the average sell off in the 2 year Treasury during a new Fed chair’s first meeting runs around 6 basis points, roughly six times the average move across all FOMC meetings generally. New chairs tend to generate outsized market reactions in their early appearances simply because markets do not yet have a reliable model for how their language translates into policy. Wednesday was the first data point in that calibration process for Warsh, and based on the initial market reaction, that calibration is still very much underway.

CrowdStrike Revisited: When the Selloff Becomes the Setup

Two weeks ago, this newsletter covered CrowdStrike’s results alongside Micron as an example of good news becoming bad news in the current market regime. It is worth checking back in.

CrowdStrike closed Wednesday at $682.96, down from the all-time high of $782.17 set earlier in June but still up roughly 80% since its April lows. The stock has stabilised rather than continued falling, trading in a range over the past two weeks that suggests the immediate post-earnings reaction has been digested rather than extended.

The analyst picture remains genuinely divided, which is itself informative. The consensus rating sits at Buy, with one tracking service putting the average 12 month price target at $673.65, essentially flat with the current price, while another puts the median target as high as $725 and the top individual target at $825. At the other end, Berenberg downgraded the stock to Hold this month, but the basis for that downgrade was explicitly valuation, not any concern about the underlying business. That distinction matters. When a downgrade cites price rather than fundamentals, it is a statement about what has already been priced in, not a revision of the investment case itself.

Commentary around the stock since the selloff has reflected the same split. Some market participants have characterised the post-earnings drop as algorithmic, high-frequency systems selling first on a headline detail and discerning the broader quality of the print only afterward. Others have continued to flag the stock’s valuation, now trading at a multiple that prices in a great deal of future growth, as the more durable constraint on near term upside regardless of execution quality.

What has not changed in two weeks is the underlying business. CrowdStrike continues to expand its platform, including a newly expanded partnership with Amazon Web Services on AI and cloud security tooling, and continues to report enterprise customers standardising onto its Falcon platform. The structural demand drivers for cybersecurity spending, which this newsletter has noted tend to hold up better than general IT spending during periods of economic uncertainty, remain intact.

The honest answer to the question this section opened with is that it is too early to say definitively. A stock that fell sharply on its earnings reaction and then stabilises within a tight range for two weeks is neither confirming that the selloff fully resolved the valuation tension nor confirming that it failed to. What it does suggest is that the market has settled, for now, into a holding pattern around a price that reflects genuine disagreement about how much future growth is already baked in. That disagreement, not a clean verdict in either direction, is the most accurate way to describe where CrowdStrike sits two weeks on.

Micron and the Apple Warning Nobody Saw Coming

On the same day Kevin Warsh delivered his hawkish dot plot, a comment from an entirely different corner of the market may have said just as much about where the economy actually stands.

Apple CEO Tim Cook told the Wall Street Journal, in comments published Wednesday June 17, that the company can no longer absorb the rising cost of memory chips and that price increases across its product lineup are now “unavoidable.” Cook’s framing was striking for its scale: “This is a 100-year flood. I’ve never seen anything like it in any area in over 40 years,” referencing his career across IBM, Compaq, and Apple. The root cause, as this newsletter has covered repeatedly since Issue 49’s memory supercycle piece, is the wholesale reallocation of DRAM and NAND manufacturing capacity toward AI server demand, leaving consumer electronics fighting over a shrinking remaining supply.

The numbers behind Cook’s comment are stark. Memory chip costs have roughly quadrupled since last year. DRAM and NAND pricing combined are estimated to have risen more than 300% since 2023. Micron has discontinued production of consumer memory products entirely to focus on commercial AI customers, a decision that, multiplied across the industry, is the direct mechanical cause of the shortage Cook is describing. S&P Global Ratings warned in a report published just days before Cook’s comments that memory prices are “poised to stay elevated on tight supply stemming from surging AI demand,” projecting the tight conditions will persist well beyond 2026.

Micron sits on the other side of exactly this trade. The stock closed at $1,127.94 this week, up over 8% on the day Deutsche Bank raised its price target from $1,000 to $1,500 and Citi lifted its own target from $840 to $1,200, pushing Micron’s market capitalisation to roughly $1.27 trillion, within striking distance of Meta. Micron has signed its first five year Strategic Customer Agreement, designed specifically to dampen the boom-and-bust cycles that have historically plagued the memory industry, launched a $5.4 billion debt buyback, and raised its dividend by 30%, all while trading at roughly 8.4 times forward earnings, well below the broader semiconductor industry average of 24 times.

What Cook’s comment makes visible is something this newsletter has argued from the start of the memory supercycle coverage: the AI infrastructure build out does not stay contained to data centres. It is now showing up directly in the cost structure of the most valuable consumer hardware company in the world, and by extension, in the price tags Apple’s customers will pay for the iPhone 18 lineup expected this September. Apple is one of the largest buyers of memory and storage anywhere in the world, and even a company of its scale and negotiating leverage has concluded it cannot fully absorb this shift.

For long-term investors, the read here is two-sided and both sides matter. For Micron, Samsung, and SK Hynix, sustained pricing power of this magnitude is exactly the structural tailwind this newsletter identified months ago, now confirmed by one of their largest customers in the most public way possible. For Apple, and for any company that consumes memory as an input rather than producing it, this is a genuine margin headwind arriving at a moment when the broader inflation picture, as this issue’s other sections make clear, is already complicated. The same structural force, AI driven memory scarcity, is simultaneously the best argument for owning Micron and a real source of uncertainty for owning Apple. Long-term investors holding both need to understand they are now, in a very direct sense, on opposite sides of the same underlying trade.

Closing Thoughts

A Resolution and a Reminder, in the Same Week

This week delivered the geopolitical resolution this newsletter has been tracking since the Iran conflict first began disrupting energy markets months ago, and then, within 48 hours of the formal signing, delivered a reminder of exactly how fragile that resolution is. It also delivered the clearest signal yet that the Federal Reserve, under new leadership, intends to take inflation seriously even at the cost of disappointing a market that had spent much of the year pricing in eventual relief.

Neither story is resolved, and the weekend made that more explicit than this newsletter could have anticipated even a few days ago. The Iran deal is a framework with a sixty day clock attached, and Saturday’s disputed closure of the Strait of Hormuz, contested between Iran’s stated position and CENTCOM’s tracked shipping data, shows just how quickly that framework can be tested by events entirely outside US-Iran control. The Fed’s hawkish dot plot is a projection, not a commitment, and Thursday’s PCE print next week will be the first real test of whether the data supports the path the committee has now signalled.

What ties the week together, and what Tim Cook’s comment on memory prices captures perhaps better than any single market data point could, is that the forces driving this market, AI infrastructure demand, energy geopolitics, and monetary policy, are no longer separate stories. They are increasingly the same story, showing up simultaneously in a central bank’s dot plot, a consumer electronics company’s pricing strategy, a memory chip maker’s stock price, and a contested shipping lane half a world away, all within the same week.

For long-term investors, the discipline required by a week like this has not changed. Own the businesses whose value compounds regardless of which way the Iran negotiation, the Lebanon ceasefire, or the Fed’s next move ultimately resolves. This week gave genuine, if incomplete, answers to two of the biggest questions hanging over markets and a weekend reminder that those answers remain provisional. It is worth taking the time to understand them properly before the next set of questions arrives, which, based on this week, may not take long.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

From a 900 Point Drop to SpaceX’s Record Debut

The week began with a modest recovery from the prior Friday’s brutal session. The S&P 500 rose 1.19% on Monday, the Nasdaq gained 1.57%, and the VIX fell sharply to 19.82 as chip stocks rebounded. The calm did not last.

Tuesday and Wednesday were dominated by Iran. On Tuesday, Trump indicated that strikes on Iran might resume after the country allegedly targeted a US helicopter, sending stocks lower intraday before a modest recovery. The Nasdaq fell 0.97% to 25,678.82, and the S&P 500 dropped to 7,386.65, with technology and energy the only sectors in the red.

Wednesday brought the worst session of the week. The Dow dropped 900 points as the US signalled more strikes on Iran were imminent. The same morning, May CPI was released: headline inflation at 4.2% year-on-year, the highest since April 2023, in line with the Dow Jones consensus. But the detail underneath told a more complicated story. Core CPI rose just 0.2% month-on-month, below the 0.3% forecast and down from April’s 0.4% increase, with the annual core rate at 2.9%. Energy costs, up 23.5% year-on-year with gasoline up 40.5%, drove the overwhelming majority of the headline acceleration. Real wages fell for a second consecutive month, down 0.7% year-on-year.

Thursday brought a turn. Stocks climbed and oil prices rose as investors weighed early signs that the US and Iran might be moving toward de-escalation rather than further conflict.

Friday was the big one. SpaceX made its public market debut on the Nasdaq under the ticker SPCX. Shares priced at $135, opened at $150, and closed at $161.11, up 19% on the day. The intraday high briefly pushed the company’s market capitalisation to approximately $2.21 trillion, within striking distance of Amazon’s roughly $2.54 trillion. Trading volume topped 207 million shares with dollar volume just under $33 billion, more than the combined dollar trading volume of QQQ and SPY, the two most heavily traded ETFs in the world, on the same day. The broader market also rallied on hope that a US-Iran peace deal could lift sanctions and reopen the Strait of Hormuz. Brent crude fell 2% toward $85. The S&P 500 gained 0.5% to 7,431, the Nasdaq rose 0.3%, and the Dow added 0.7% to 51,202, led by financials, with Goldman Sachs up 2.57% and JPMorgan up 2.25%.

For the month, the S&P 500 remains down slightly. Year-on-year, it remains up 24.34%.

Week Ahead

June 15 to 19: Warsh’s First FOMC Meeting

Monday June 15

A quiet open ahead of the main event. Markets will continue to digest SpaceX’s post-IPO trading, where early price action in the days following a debut of this scale often tells its own story about institutional versus retail positioning. Any further developments on the US-Iran situation, in either direction, will dominate sentiment heading into Tuesday.

Tuesday June 16

The FOMC meeting begins. Markets are pricing a 97 to 98% probability of no rate change, with the federal funds rate expected to remain at 3.50% to 3.75%. The two day format means Tuesday is largely a positioning day, with the real action arriving Wednesday afternoon.

Wednesday June 17

Kevin Warsh’s first FOMC meeting as chair concludes with the rate decision at 2 p.m. ET, followed by the Summary of Economic Projections, the dot plot, and Warsh’s first press conference. The decision itself is close to a formality. What matters is the language: whether the committee shifts from an easing bias toward a neutral stance, how the dot plot reflects internal disagreement following the four dissents at the prior meeting, and how Warsh handles his first extended public exposure in the role. May retail sales data also lands the same morning, giving a read on whether the consumer absorbed April and May’s energy driven price increases without pulling back materially.

Thursday June 18

May PPI drops, the final inflation data point before markets digest the full picture from Wednesday’s Fed decision alongside this week’s producer price trends. Weekly jobless claims also land. Markets will spend the session interpreting Wednesday’s FOMC communication in light of the freshest available data.

Friday June 19

Markets close the week digesting a full seven days of Fed communication, inflation data, and whatever has emerged on the Iran situation. With SpaceX now a week into public trading and the FOMC meeting behind us, the focus shifts toward how durable this week’s relief rally proves to be heading into the second half of June.

The Core CPI Surprise Nobody Is Talking About

Headline CPI hit 4.2% year-on-year in May, the highest reading since April 2023. The headline number was exactly what economists expected, and exactly what the energy shock from the Iran conflict had been signalling for weeks. Gasoline prices rose 40.5% year-on-year. Energy overall rose 23.5%. Fuel oil rose 58.9%. None of that is surprising. All of it has been visible in the data for months.

What was surprising, and what received far less attention than the 4% headline figure, is what happened underneath it. Core CPI, which strips out food and energy, rose just 0.2% month-on-month. Economists had forecast 0.3%. April’s monthly core reading was 0.4%. This is the second consecutive month in which the monthly core print came in below expectations and below the prior month’s pace.

The annual core rate ticked up slightly to 2.9% from 2.8% in April, a new high since September 2025. But the monthly trajectory is what matters for understanding momentum, and the monthly trajectory is decelerating. If May’s pattern continues into June and July, the annual core rate could plateau or even begin to ease, even as the headline rate remains elevated due to energy.

This distinction matters enormously for how the Federal Reserve thinks about its mandate. Energy driven headline inflation, while painful for consumers at the pump, is generally treated by central banks as a supply shock rather than a demand driven, monetary phenomenon. Raising interest rates does very little to bring down the price of oil when the cause is a geopolitical conflict disrupting the Strait of Hormuz. What raising rates can address is core inflation, the broad based price pressures across goods and services that reflect underlying demand conditions in the economy.

A core CPI print that decelerates on a monthly basis, even as headline inflation rises on energy, is about as constructive a setup as the Fed could hope for heading into a meeting. It allows the central bank to acknowledge that headline inflation is uncomfortable without concluding that aggressive tightening is the appropriate response. Kevin Warsh walks into his first FOMC meeting with a CPI report that, read in full, gives him genuine room to project a measured, watchful tone rather than an urgently hawkish one.

The real wage story remains the most uncomfortable part of the report. Average hourly earnings grew 3.4% year-on-year, trailing the 4.2% headline inflation rate by 0.8 percentage points. This is the second consecutive month that real wages have fallen, following April’s negative reading. For households, the headline number is what shows up at the gas pump and the grocery store. The core number is an abstraction. Both are true simultaneously, and both matter, just to different audiences.

SpaceX’s $2.2 Trillion Friday

SpaceX’s public market debut on Friday was, by any measure, a historic event. Shares priced at $135, implying a valuation of $1.77 trillion at the IPO price. The stock opened at $150, immediately up 11% from the offer price, and proceeded to climb further. By the close, shares stood at $161.11, up 19% on the day. At the session high of $168.75, up 25% from the IPO price, SpaceX’s market capitalisation reached approximately $2.21 trillion, putting it within striking distance of Amazon’s roughly $2.54 trillion.

The scale of the offering itself was unprecedented. SpaceX sold 555.6 million shares for a $75 billion raise, with underwriters holding an option for a further $11.2 billion. That raise alone is larger than the entire market capitalisation of most S&P 500 companies. Trading volume on the day exceeded 207 million shares, with dollar volume just under $33 billion, more than the combined dollar trading volume of QQQ and SPY, the two most heavily traded ETFs in the world, on the same day.

Elon Musk retains over 82% voting control following the offering, an unusual structure for a company of this size but one that reflects the unique dynamics of a founder-controlled business going public at a scale where few traditional governance norms apply. The company’s February merger with xAI, which valued the combined entity at $1.25 trillion at the time, means SpaceX’s public listing also represents the public market debut of the Grok AI model and xAI’s broader AI ambitions, layered on top of the Starlink and launch businesses that generate the bulk of current revenue and profitability.

The ripple effects across the sector were immediate and significant. Space and satellite stocks broadly declined as SpaceX began trading: Firefly Aerospace fell more than 18%, Rocket Lab, Redwire, and Intuitive Machines each dropped at least 10%, Karman Holdings slumped 4%, and Virgin Galactic, trading around a $400 million market cap, plunged 34%. Satellite names including Planet Labs and AST SpaceMobile also fell. This is a common pattern around large, high profile IPOs in a sector: capital and attention rotate toward the new listing, often at the direct expense of smaller existing players in the same space.

Behind the headline valuation, the S-1 filing gave the clearest picture yet of what SpaceX actually is as a business. Total 2025 revenue came in at approximately $18.7 billion, up roughly 33% from $14.1 billion in 2024. Adjusted EBITDA reached $6.6 billion for the year. But the company also reported a net loss of approximately $4.9 billion, a reminder that scale and profitability are not the same thing.

The segment breakdown is where the real story sits. Starlink, the satellite internet business, generated approximately $11.4 billion in revenue, around 61% of the total, and was the only segment that was genuinely profitable, contributing roughly $4.4 billion in operating profit. The Space segment, SpaceX’s original launch business including Falcon 9 and the national security launch contracts that built the company’s reputation, generated around $4.1 billion in revenue but posted an operating loss. The AI segment, formed from the February 2026 merger with xAI and now referred to as SpaceXAI, contributed roughly $3.2 billion in revenue but lost approximately $6.4 billion at the operating level for the year.

In plain terms, Starlink is funding xAI. The satellite internet business that most people associate with home broadband in rural areas is generating the cash flow that subsidises one of the most expensive AI build-outs in the world. That is not unusual for a company at this stage of an AI investment cycle, several of the largest technology companies in the world are doing some version of the same thing, but it does mean that an investor buying SPCX at $161 is buying exposure to two very different businesses with very different risk profiles, bundled into a single security.

Not everyone is convinced the valuation reflects the underlying business. Keith Snyder, senior analyst at CFRA Research, told CNBC he remains skeptical of whether SpaceX is really worth $1.77 trillion. That skepticism is worth taking seriously, not because SpaceX is not an extraordinary business, but because the question of what a business is worth and what the market is willing to pay for it on day one of trading are frequently very different things, particularly for an offering this large and this anticipated.

For long-term investors, SpaceX’s debut is significant for reasons beyond the company itself. It is the first of what is expected to be a wave of major AI and space adjacent listings, with OpenAI and Anthropic both reportedly laying groundwork for their own offerings later in 2026. The market’s appetite for SpaceX on day one, and the valuation it was willing to support for a business that combines a profitable satellite internet operation with a loss-making AI division, will be read by every banker and every prospective issuer in this category as a signal about what comes next.

Warsh’s First Meeting: What Wednesday Actually Means

On Wednesday June 17, Kevin Warsh chairs his first FOMC meeting as Federal Reserve Chair. The vote itself is close to a formality. CME FedWatch data puts the probability of no rate change at 97 to 98%, with the federal funds rate expected to remain at 3.50% to 3.75%, unchanged across what would be the fourth consecutive meeting.

If the decision is essentially pre-determined, why does this meeting matter so much? Three reasons.

The first is the dot plot. Alongside the rate decision, the Fed releases its Summary of Economic Projections, including the so-called dot plot showing where each FOMC member expects rates to be at the end of 2026, 2027, and beyond. The prior meeting produced four dissents, an unusually high number that reflects genuine disagreement within the committee about the appropriate path forward. Whether that disagreement narrows or widens under Warsh’s first meeting, and where the median dot sits relative to where the market is currently pricing the path of rates, will shape expectations for the rest of the year regardless of what the headline decision says.

The second is the language. Strategists at major banks have been explicit that they expect an explicit move away from a bias toward easing to a more neutral stance on rates. That shift in framing, even without any change to the actual rate, is itself a meaningful signal. A Fed that has spent much of 2026 signalling an eventual path toward cuts, even as it has held rates steady, communicating a pivot to neutral tells markets that the bar for the next move, in either direction, has shifted. Given this week’s CPI report, with its encouraging core trend but alarming headline figure, how Warsh frames that tension in his prepared remarks and in the press conference will be closely parsed.

The third is Warsh himself. This is his first extended public appearance in the role, his first press conference, and the first opportunity for markets to observe his communication style under pressure. History offers a relevant precedent: when Jerome Powell took over from Janet Yellen in 2018, markets reacted sharply to what they perceived as a more hawkish tone in his early communications, even though the underlying policy stance had not shifted dramatically. It can take several meetings for markets to calibrate how a new chair’s language translates into policy expectations. Wednesday is the first data point in that calibration process for Warsh, and given the political context of his confirmation, a closely divided 54 to 45 Senate vote amid an already fraught environment, the scrutiny on his tone will be intense.

For long-term investors, the outcome that matters least is the rate decision itself, because it is already known with near certainty. What matters is everything around it: the dot plot’s signal about the path of rates over the next 18 months, the framing of the easing to neutral shift, and the market’s read on Warsh’s communication style. All three of those inputs will influence the discount rate environment for risk assets well beyond this single week, and all three carry genuine uncertainty heading into Wednesday afternoon.

The Week the Iran Story Almost Changed Everything

Consider the span of just three trading days. On Tuesday, reports that Iran had targeted a US helicopter prompted Trump to indicate that strikes might resume, sending markets lower intraday before a modest recovery. On Wednesday, the Dow dropped 900 points as the US signalled more strikes were imminent, the worst session of the week by a wide margin. By Thursday, sentiment had turned, with stocks climbing and oil rising on early signs the US and Iran might be moving toward de-escalation. By Friday, the market was rallying on hope that a peace deal could lift sanctions and reopen the Strait of Hormuz entirely, with Brent crude falling 2% toward $85.

That is a 900 point Dow swing on escalation fears followed by a broad market rally on de-escalation hope, within 48 hours, in the same week. This whiplash is not noise to be filtered out. It is the single most important variable for understanding the macro environment heading into next week’s FOMC meeting, and arguably for the rest of 2026.

The transmission mechanism from the Iran conflict to the broader economy runs almost entirely through energy. Gasoline prices, up 40.5% year-on-year, are the dominant driver of headline CPI’s acceleration to 4.2%. Every escalation in the conflict raises the probability of further disruption to the Strait of Hormuz, one of the world’s most critical oil transit routes, which pushes Brent crude higher, which pushes gasoline prices higher, which pushes headline inflation higher. Every de-escalation signal works in reverse. Friday’s 2% drop in Brent toward $85, in response to peace deal hopes, is a direct illustration of how tightly coupled these markets have become.

This matters enormously for the Fed’s calculus heading into Wednesday’s meeting. If the de-escalation signals from Thursday and Friday prove durable, and the Strait of Hormuz situation genuinely stabilises, headline CPI could begin falling meaningfully in the coming months purely on energy base effects, even without any change in core inflation trends. That would be the cleanest possible outcome for Warsh: a Fed that held steady through the worst of an energy shock, watched headline inflation ease as the shock resolved, and never had to choose between fighting inflation and supporting growth.

If, instead, this week’s de-escalation proves temporary and the conflict reignites, as it has done multiple times already in 2026, the opposite dynamic plays out. Headline inflation stays elevated or worsens, the core CPI deceleration becomes harder to read as a durable trend versus noise, and the Fed’s path becomes genuinely more difficult regardless of what Wednesday’s dot plot signals.

For long-term investors, the lesson from this week’s whiplash is not to try to predict the next 48 hour swing in the Iran conflict. Nobody can do that reliably, and the market has already demonstrated this week that it cannot either. The lesson is to recognise that a meaningful portion of the inflation and rate uncertainty currently embedded in asset prices is tied to a binary geopolitical outcome that could resolve in either direction with little warning. Positioning portfolios with that uncertainty in mind, rather than betting heavily on either outcome, remains the more durable approach.

Closing Thoughts

A Week That Asked More Questions Than It Answered

The highest CPI reading in three years, with a core trend that quietly improved underneath it. The largest IPO in history, briefly putting a rocket company within reach of Amazon’s market cap. And a Federal Reserve meeting next week that is simultaneously a foregone conclusion and one of the most closely watched events of the year.

None of these stories resolved this week. The Iran situation remains genuinely binary. The core CPI improvement is one data point, not yet a trend. SpaceX’s valuation will be tested by months of trading, not a single Friday. And Warsh’s tone on Wednesday will be interpreted, re-interpreted, and recalibrated for weeks afterward.

What this week demonstrated clearly is how interconnected these threads have become. The same conflict that drives headline inflation also drives the energy prices that flow through to the Fed’s calculus, which flows through to the discount rates applied to companies like SpaceX, AMD, and every other name in this newsletter. There is no clean separation between geopolitics, macro, and company fundamentals anymore, if there ever was one.

For long-term investors, the response to a week like this is not to have a confident view on how Wednesday’s FOMC meeting or the Iran situation resolves. It is to own businesses whose long-term value does not depend on getting either of those calls right. That has been the position of this newsletter for 54 issues. It remains the position heading into what could be one of the more consequential weeks of the year.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

Nine Weeks Ended.

The week opened with momentum. The S&P 500 rose 0.26% on Monday to 7,599.96 and the Nasdaq gained 0.42%. Tuesday pushed further, with the S&P 500 closing above 7,600 for the first time ever at 7,609.78, and the Dow hitting a fresh all-time high at 51,307.79. Alphabet weighed slightly after announcing an $85 billion AI infrastructure fundraise through new stock sales.

Wednesday was the pivot. Broadcom reported after the close: AI chip revenue of $10.8 billion, up 143% year-on-year. Total revenue of $22.19 billion, a record, up 48%. Adjusted EPS of $2.44, beating the $2.40 consensus. And yet the stock fell 12% in after-hours trading. The reasons: Broadcom left its full-year AI chip revenue forecast unchanged at $56 billion, and Q3 AI guidance of $16 billion came in below the $17.2 billion analysts had expected. CrowdStrike also reported Wednesday, beating on every metric including a 4-for-1 stock split announcement, and fell 11% in after-hours anyway.

Thursday spread the damage across the sector. Broadcom fell 12.6% on the session, dragging AMD and Intel down around 11%. The S&P 500 closed marginally higher only because non-tech sectors held. The Nasdaq ended fractionally lower. The VIX began climbing.

Friday delivered the week’s biggest shock. The May jobs report showed 172,000 positions added, nearly double the 85,000 consensus, with unemployment steady at 4.3%. The 10-year Treasury yield surged above 4.5%, and the 30-year crossed 5% for the first time in over a year. The Nasdaq fell 4.18%, its worst single session since April 2025. The S&P 500 dropped 2.64% to 7,383.74. The Dow lost 695 points. Semiconductor stocks bore the brunt: Marvell fell 16%, Micron 13%, AMD and Intel 11%, Nvidia nearly 6%. The VIX surged 34%, closing above 20 for the first time since April.

For the week: the S&P 500 fell more than 2%, the Nasdaq lost 4.7%. Nine consecutive weeks of gains. Over in one session.

Week Ahead

Monday, June 8

A quiet open. Campbell’s reports. Global manufacturing PMI readings give a read on industrial activity heading into what is a data-heavy week. Treasury markets will still be absorbing Friday’s yield surge, and any further movement in the 10-year or 30-year will set the tone for the session.

Tuesday, June 9

May existing home sales land, giving a read on whether elevated mortgage rates and inflation are cooling the housing market further. J.M. Smucker and Casey’s General Stores report, adding to the consumer picture. Bond markets will be the story, not equities, as traders position ahead of Wednesday’s CPI.

Wednesday, June 10

The most important data release of the week, and arguably of the summer. May CPI drops at 8:30 a.m. ET. Headline inflation is expected to come in at 4.2% year-on-year, up from 3.8% in April. That would be the highest headline CPI reading since early 2023. Core CPI is the critical number heading into the June 16 to 17 FOMC meeting. The trend is clear: March core was 2.6% and April core was 2.8%. If May continues that acceleration, the Fed’s first meeting under Kevin Warsh becomes the most consequential in the current cycle. Oracle reports after the close, providing a further read on enterprise cloud and AI infrastructure demand.

Thursday, June 11

May PPI drops alongside the ECB rate decision. The ECB is expected to hold rates, but its commentary on European inflation and energy costs will add a global dimension to the monetary policy picture. Adobe reports, giving a window into enterprise AI software adoption. Weekly jobless claims also land, the last labour market read before the FOMC meeting.

Friday, June 12

The preliminary University of Michigan consumer sentiment reading for June drops. After a week of inflation data, a surging VIX, and bond yields at multi-year highs, this will tell us whether the consumer is beginning to feel it. It is also the last major data release before the June 16 to 17 FOMC meeting. Whatever it shows, Kevin Warsh will be reading it alongside the CPI, PPI, and jobs prints when the committee convenes.

Broadcom and the New Rules of the AI Trade

Broadcom delivered one of the most impressive quarters in semiconductor history. AI chip revenue of $10.8 billion, up 143% year-on-year. Total revenue of $22.19 billion, a record, up 48%. Adjusted EPS of $2.44, beating estimates. Record free cash flow of $10.26 billion, representing 46% of revenue. The company ended the quarter with $19.6 billion in cash. By any conventional measure, it was an extraordinary result.

The stock fell 12% anyway.

Understanding why reveals something important about where the AI trade is now. Broadcom left its full-year 2026 AI chip revenue forecast unchanged at $56 billion. Its Q3 AI chip guidance of $16 billion came in below analyst expectations of $17.2 billion. CEO Hock Tan disclosed that Google would likely draw on multiple chip suppliers rather than Broadcom exclusively, raising questions about customer concentration risk. And Tan acknowledged that surging AI semiconductor sales were diluting the company’s overall gross margins as the product mix shifted.

None of these disclosures change the fundamental reality that Broadcom is one of the most important AI infrastructure companies in the world. It designs custom AI accelerators for some of the largest organisations on the planet, including Anthropic, Google, Meta, and OpenAI. Its networking chips are essential to every large-scale AI cluster. Its FY2027 AI chip revenue target of more than $100 billion, reiterated on the call, represents a business that does not yet exist at anything close to that scale in today’s market.

But the market had priced in something more. Broadcom shares were up over 20% year-to-date entering the session. The stock was trading at all-time highs. At that price, the bar for what constitutes a positive earnings event had moved well above what most companies could ever deliver. A 143% year-on-year increase in AI chip revenue was not enough, because the market had already decided that acceleration was the baseline. Holding the full-year forecast was read not as discipline but as a ceiling signal.

For long-term investors, Broadcom’s week illustrates the new rules of the AI trade with unusual clarity. In the first phase of the AI investment cycle, beating estimates was sufficient to send stocks higher. In the current phase, beating estimates is necessary but not sufficient. Guidance must also accelerate beyond what analysts have already modelled. When Broadcom said $16 billion for Q3 and analysts expected $17.2 billion, the interpretation was not that Q3 will still be extraordinary. The interpretation was that the acceleration story had a ceiling the market had not previously priced.

The 46 analysts who maintain a strong buy consensus on Broadcom with an average 12-month price target of $497.74 have not changed their fundamental view. The business remains exceptional. What changed this week is not the business. It is the expectations’ environment that surrounds it.

The Jobs Report That Changed Everything

The May jobs report showed 172,000 positions added. The consensus estimate was 85,000. April was revised up to 179,000. The unemployment rate held at 4.3% for the fourth consecutive month. By any conventional measure of labour market health, these are strong numbers.

The market’s response: the Nasdaq fell 4.18%, its worst session since April 2025. The 10-year Treasury yield surged above 4.5%. The 30-year crossed 5% for the first time in over a year. The VIX surged 34% to above 20.

This is good-news-is-bad-news at its most extreme. And the mechanism is straightforward: a labour market this strong, combined with core PCE at 3.3% and headline inflation expected to reach 4.2% in May, removes any remaining argument for rate cuts and puts rate hikes firmly back on the table heading into the June 16 to 17 FOMC meeting.

The report had one genuine note of complexity. Average hourly earnings rose just 3.4% year-on-year, the lowest since 2021. That is below inflation, meaning real wages remain negative. The labour market is adding jobs, but those jobs are not paying enough to outrun prices. For workers, that combination is straightforwardly bad. For the Fed, it is a partial complication: if wage growth is not accelerating, services inflation may not be as entrenched as the headline jobs number suggests. Goldman Sachs cited this nuance in its post-report note, saying the data increases the risk of a longer pause but does not make rate hikes a certainty.

Cleveland Fed President Beth Hammack was less equivocal. She said it may soon be appropriate for the Fed to act to address the growing risks of persistently elevated inflation. Morgan Stanley’s Ellen Zentner called the report strong from every angle and said the Fed remains watching and waiting, focused on the inflation side of its mandate. A watching and waiting Fed with a strong jobs market and accelerating inflation is a Fed that is running out of reasons to stay still.

The June 16 to 17 FOMC meeting will be Kevin Warsh’s first as chair. He will have May’s NFP, Wednesday’s CPI, Thursday’s PPI, and April’s core PCE all in hand. The data corridor between now and that meeting is the most consequential two-week stretch of the current cycle. History suggests that a new Fed chair can affect market trends, and that it may take a few meetings for investors to calibrate how hawkish or dovish Warsh’s language will be relative to Powell’s. Either way, the jobs report has made sure the first meeting matters.

May CPI Wednesday: The Number That Sets the Summer

Wednesday, June 10, is the most important single data release of the summer. May CPI drops at 8:30 a.m. ET, and what it shows will determine the tenor of the June 16 to 17 FOMC meeting, the trajectory of bond yields through July, and whether the equity market’s repricing this week is the beginning of a larger reset or a short-term buying opportunity.

Headline CPI is expected to come in at 4.2% year-on-year, up from 3.8% in April. If confirmed, that would be the highest headline reading since early 2023. The energy shock from the Iran conflict accounts for a significant portion of that acceleration. Gasoline prices have remained elevated through May, and the full pass-through of April’s oil price spike has not yet fully registered in monthly price data.

Core CPI is the number that matters most for the Fed. The trend over recent months has been consistent: March’s core was 2.6% and April’s core was 2.8%. If May prints at 3.0% or above, it would represent clear evidence that inflation is broadening beyond energy and into the structural categories; shelter, services, transportation, that the Fed finds most difficult to address through rate policy without inducing a recession.

The timing makes this print unusually consequential. The Fed does not meet again until June 16 to 17. The FOMC will have May CPI, May PPI, and April PCE all available when it sits down to deliberate. Warsh’s first meeting as chair will be shaped almost entirely by the data that arrives in the next seven days. If core CPI accelerates and PPI follows, the probability of a rate hike at that meeting or at the one that follows in late July becomes difficult to dismiss.

For long-term investors, the CPI print does not change the underlying business cases of the companies discussed in this issue. Broadcom’s custom AI chip business, CrowdStrike’s cybersecurity platform, Micron’s HBM manufacturing advantage — none of these are materially affected by whether core CPI prints at 2.8% or 3.1% on Wednesday. What does change is the discount rate that the market applies to future earnings, the cost of capital for companies financing growth, and the sentiment environment in which all of these businesses trade. In that sense, Wednesday’s number matters not because it changes the fundamentals, but because it determines how long investors are willing to wait for those fundamentals to be fully rewarded.

When Good News Becomes Bad News: CrowdStrike and Micron

CrowdStrike reported one of the strongest quarters in its history on Wednesday, June 3. Revenue of $1.39 billion, up 26% year-on-year, beating estimates. Adjusted EPS of $1.10 against a $1.07 forecast. Record net new ARR of $256 million, up 32%. Record free cash flow of $468 million. A raised full-year guidance. A 4-for-1 stock split announcement. GAAP net income turned positive for the first time at $27.8 million, compared to a loss of $104.3 million a year earlier.

By any conventional scorecard, it was a clean, dominant quarter. The stock fell 11% in after-hours trading.

Micron crossed $1 trillion in market capitalisation last Tuesday on a 19% single-session gain. By Friday, it had given back 13% of that in a single day, pulled down entirely by the macro selloff triggered by the jobs report. No company-specific news. No earnings miss. No guidance change. Just a strong jobs print pushing Treasury yields higher, which pushed technology valuations lower, which sent a stock that had risen 19% in a week down 13%.

Both stories illustrate the same underlying dynamic: the AI trade has entered a regime in which valuation is the determining variable, not quality. CrowdStrike’s shares were up nearly 65% year-to-date entering Wednesday’s session. The stock was trading at a forward price-to-earnings ratio of approximately 154 times. At that multiple, billing growth of 18%, below analyst expectations even as revenue grew 26%, was enough to trigger a re-rating. The market is not asking whether CrowdStrike is a great business. It is asking whether it is a great business at $747 per share. On Wednesday night, the answer was no.

Micron’s situation is different but related. The fundamental thesis is intact: HBM supply is sold out through 2026, AI GPU demand is compounding, and Micron is the only US-based manufacturer of advanced memory. None of that changed on Friday. What changed is that a strong jobs report pushed the 30-year Treasury yield above 5%, which mechanically increases the discount rate applied to future earnings, which reduces the present value of long-duration growth stocks. Micron, up nearly 850% over the prior 18 months, carries a significant amount of future growth in its price. When discount rates move, that future growth becomes less valuable in today’s terms.

For long-term investors, the lesson from both CrowdStrike and Micron this week is not that the AI trade is over. The businesses are excellent. The structural demand drivers are real. The lesson is about the difference between owning a great business and owning a great business at the right price. In a regime where good news becomes bad news, where record results send stocks down 11% and macro surprises compress valuations irrespective of fundamentals, the margin of safety in your entry price is the only variable fully within your control.

CrowdStrike’s 4-for-1 stock split, effective July 2, will increase retail accessibility and liquidity. The business case has not weakened. Micron’s HBM advantage has not narrowed. What has happened is that the market has used a week of external pressure to reprice both stocks towards a more rational relationship between price and expected future cash flows. Whether this week’s levels represent that rational relationship is the question worth sitting with before acting in either direction.

Closing Thoughts

What the Winning Streak Ending Actually Means

Nine consecutive weeks of gains ended on Friday with the Nasdaq’s worst session since April 2025. The VIX is above 20 for the first time since April. The 30-year Treasury yield is above 5%. Rate hike odds have moved from 45% to something that is no longer being dismissed as a tail risk.

It would be easy to frame this as a turning point. Markets reversed. The AI trade stumbled. The jobs data changed the calculus. But the more accurate framing is simpler: the market was pricing a set of assumptions that required a specific path for inflation, rates, and AI revenue acceleration. This week, all three of those assumptions were tested. Broadcom’s guidance held flat when the market needed it to accelerate. The jobs report removed the last argument for near-term rate cuts. And CrowdStrike’s billings growth disappointed even as revenue accelerated.

None of these are existential events for the companies involved or for the broader AI investment thesis. They are recalibrations. The market is adjusting its expectations towards something more consistent with the actual trajectory of the data. That process is uncomfortable when it happens quickly, but it is healthy over time. Valuations that outrun earnings growth eventually come back. When they do, the businesses underneath them are often better than they were when the premium was at its peak.

Wednesday’s CPI print will tell us whether this week’s repricing is the beginning of something larger or a sharp but contained reset. Either way, the long-term edge is not built by predicting which way that print lands. It is built by owning businesses that can compound through the uncertainty that surrounds it.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Week of May 25 to 29, 2026

Fifty-two consecutive weeks. One full year of publishing without missing a single issue. One year of markets, macro, earnings, geopolitics, and the ongoing work of separating signal from noise. Thank you for reading every week. The closing of this issue reflects on the year. But first, here is what happened and what is coming.

This week the S&P 500 closed its ninth consecutive week of gains. Dell reported the fastest revenue growth since its return to public markets, up 88% year-on-year. Core PCE rose to 3.3%, its highest level in nearly three years. And the market, somehow, set new records on the same day. A year in, the contradictions that define this market are as sharp as they have ever been.

Market Overview

Nine Weeks. A Record Month. A Market That Ignores Everything.

The S&P 500 closed its ninth consecutive week of gains on Friday, its longest winning streak since 2004. The Dow crossed 51,000 for the first time ever, closing at 51,054. The Nasdaq finished May up 8%, its best monthly performance since January 2023. For the month as a whole, the S&P 500 rose approximately 6%, delivering the strongest May performance in more than a decade. Sell in May? Not this year.

The week opened cautiously after markets were closed Monday for Memorial Day. When they reopened Tuesday, fresh Iran peace rumours pushed stocks higher. The S&P 500 gained 0.61% to 7,519.12, a fresh record. The Nasdaq jumped 1.19% to 26,656.18, also a record. Micron Technology was the session’s defining moment, jumping 19% and crossing $1 trillion in market capitalisation for the first time, driven by a UBS upgrade citing AI demand changing Micron’s earnings trajectory from cyclical to structural.

Wednesday brought a flicker of reality. Chip stocks pulled back on fresh Iran headlines that turned out to be unverified. The Dow set a new record, but the Nasdaq and S&P 500 were little changed. FOMC minutes confirmed the committee voted unanimously to hold rates at 3.50% to 3.75%, noting inflation remains elevated and that Middle East developments are contributing to a high level of uncertainty. No language about hikes, but no language about cuts either.

Thursday brought the week’s defining data. April core PCE rose to 3.3% year-on-year, the highest reading in nearly three years, up from 3.2% in March. The Fed’s preferred inflation gauge is now running 1.3 percentage points above its 2% target with no clear sign of deceleration. Markets fell initially on the print, then recovered. The Nasdaq and S&P 500 set new records despite the highest PCE reading in three years.

Friday closed the week and the month on a high. Dell surged nearly 30% after reporting revenue of $43.84 billion, up 88% year-on-year, against a consensus of $35.43 billion. The Dow gained 0.72%, closing above 51,000 for the first time. May closed as the best month of the year by a significant margin.

Week Ahead

June 1 to 5: Jobs, Broadcom, and the Start of Summer

Monday June 1

HPE reports after the close. Revenue expected at $9.82 billion, up 28.7% year-on-year, with networking having grown 152% in Q1 on the back of AI infrastructure deployments. Its results will be the first major read of the week on enterprise AI demand. Manufacturing PMI and Construction Spending also drop.

Tuesday June 2

JOLTS job openings data lands, the first labour market read ahead of Friday’s NFP. Palo Alto Networks reports, giving a further read on enterprise cybersecurity demand alongside HPE’s networking commentary from the previous evening.

Wednesday June 3

Broadcom reports after the close. It is outperforming both Nvidia and the S&P 500 year-to-date, up over 22%, driven by its AI networking chip business and the VMware integration delivering ahead of schedule. Wall Street expects earnings per share of $2.40 on revenue of $22.11 billion. CrowdStrike also reports on the same day after the close. Together, the two reports will provide the most comprehensive single-day read on AI infrastructure and cybersecurity demand of the quarter. ADP private payrolls data drops in the morning, the last major preview of Friday’s NFP. Medtronic and Costco also report.

Thursday June 4

Lululemon, DocuSign, and Ciena all report. The ISM Services PMI drops alongside weekly jobless claims, completing the economic picture heading into Friday. Lululemon’s results add a premium consumer data point to the picture already painted by Walmart’s basket-size decline and Costco’s member spending commentary.

Friday June 5

The May Employment Situation report drops at 8:30 a.m. ET. This is the first jobs print that Chair Kevin Warsh will have in hand ahead of the June FOMC meeting on June 16 to 17. A strong print combined with core PCE at 3.3% makes a rate hike conversation at June’s meeting almost unavoidable. A weaker print opens a small window for patience. Either way, this number sets the tone for the rest of summer.

Dell and the AI Infrastructure Trade Nobody Saw Coming

Dell reported revenue of $43.84 billion for the quarter ended May 1, 2026. The consensus estimate was $35.43 billion. That is not a beat. That is a different order of magnitude entirely.

Revenue grew 88% year-on-year, the fastest pace since Dell returned to public markets in 2018. Adjusted earnings per share of $4.86 came in against a consensus of $2.94. The stock surged nearly 30% in extended trading on Thursday, finishing Friday at an all-time high, up over 220% for the year.

The engine behind these numbers is AI server demand. Dell’s Infrastructure Solutions Group grew well over 100% year-on-year. The company exited the quarter with an AI server backlog of over $51 billion, up from $43 billion at the start of the year and $18.4 billion just six months ago. CEO Michael Dell said demand continues to exceed supply and that the company expects to exit 2026 with a strong backlog. AI-optimised server sales are expected to approximately double this year to around $50 billion.

Dell is not a company most investors associate with the AI revolution. It does not manufacture chips. It does not run cloud services at hyperscaler scale. But it does something enormously valuable: it assembles the servers, integrates the components, and sells the completed infrastructure to the enterprises and governments building AI systems. It is the system integrator for the AI era, sitting between Nvidia’s GPUs, Micron’s memory, and the enterprise customers who need both delivered, installed, and working.

This is the other side of the AI trade. While investors crowded into Nvidia and the hyperscalers, Dell was quietly building one of the most direct exposures to the AI infrastructure build-out available in public markets. The backlog of $51 billion provides earnings visibility that is unusual for a hardware business. For long-term investors, Dell’s results reinforce a principle that runs through this entire earnings season: the AI infrastructure trade is broader than the index concentration suggests. Dell’s 88% revenue growth is the most dramatic illustration of that yet.

Core PCE at 3.3%: What the Fed’s Own Measure Is Saying

April core PCE rose to 3.3% year-on-year, the highest reading in nearly three years, up from 3.2% in March. This is the Federal Reserve’s preferred inflation gauge, and it is now running 1.3 percentage points above the Fed’s 2% target with no clear sign of deceleration.

The monthly reading offered a marginal improvement: core PCE rose 0.2% in April, down from 0.3% in March. In isolation that would be encouraging. In context, it is insufficient. The Fed requires several sustained months of sub-0.2% monthly readings before a rate-cut conversation becomes credible. One month at 0.2% does not move that needle.

What the data confirmed is that inflation is becoming more broad-based, not less. Analysts who looked beyond the headline numbers found that trimmed mean measures of PCE showed clear acceleration since the first quarter of 2025. The energy shock from the Iran conflict is no longer doing all the work. Underlying inflation is rising independently of oil prices.

Americans spent more in April, but rising prices consumed nearly all of those gains. Real personal spending barely grew. Real paychecks shrank for the second consecutive month. The personal saving rate declined. The consumer is spending to maintain living standards, not to expand them. That is a fundamentally different dynamic than the consumption-driven growth of 2024.

The FOMC minutes released Wednesday confirmed the Fed is watching all of this carefully. Chair Kevin Warsh inherits a situation where patience is the only available option, and even patience has limits. The June FOMC meeting on June 16 to 17 will be Warsh’s first as chair. He will have May’s NFP, May’s CPI, and April’s PCE all in hand. If the jobs report on Friday is strong and May’s CPI shows continued acceleration, the rate hike conversation at June’s meeting will be unavoidable. The market is not yet pricing that scenario fully.

Micron Crosses $1 Trillion: The Memory Story Arrives

Micron Technology crossed $1 trillion in market capitalisation on Tuesday, driven by a UBS upgrade that raised its price target to $1,625 and called Micron an AI giant in the making. The stock jumped 19% in a single session, one of the largest single-day moves for a mega-cap company in recent memory.

The UBS thesis is straightforward: AI demand is changing the nature of Micron’s business from a cyclical memory manufacturer to a structural beneficiary of one of the largest infrastructure build-outs in history. The argument rests on High Bandwidth Memory, the specialised memory bonded directly to AI GPUs that every Nvidia and AMD accelerator requires. Micron is the only US-based manufacturer of advanced HBM. The entire global supply is sold out through the end of 2026. HBM4, Micron’s latest generation, carries pricing up over 50% on its predecessor.

UBS sees Micron potentially becoming the seventh-largest US company by market cap by the end of 2026. That would represent one of the most dramatic rises in corporate market capitalisation in the history of public markets. Analysts project earnings growth of over 300% in the current fiscal year.

What makes Micron’s story structurally compelling beyond the near-term numbers is the HBM architecture itself. Each successive generation of AI GPU requires significantly more HBM than the last. Nvidia’s Blackwell architecture requires substantially more HBM per chip than Hopper. Rubin, the generation after Blackwell, will require more still. The demand curve for high-bandwidth memory is tied directly to the AI model complexity curve, and that curve is pointing in one direction. The memory supercycle has now produced its first trillion-dollar company. It will not be the last milestone it generates.

Fifty-Two Issues: What a Year in Markets Taught Us

The Publishing Year in Numbers:

The S&P 500 delivered a total return of approximately 29% over the period. The index hit an all-time high of 7,519 this week. Core PCE inflation ended the year at 3.3%, its highest level in nearly three years.

When Issue 1 went out, the dominant conversation was about rate cuts, a softening labour market, and whether the AI trade was real or speculative. The answer, one year later, is definitive on all three counts.

Rate cuts did not come. They went further away. The labour market did not soften. It surprised consistently to the upside, quarter after quarter, defying every consensus forecast. And the AI trade turned out to be one of the most significant structural shifts in corporate earnings since the internet era, generating revenue and profit growth that exceeded even the most optimistic projections made twelve months ago.

The year produced extraordinary results. Nvidia delivered $81.6 billion in a single quarter. Dell grew revenue 88% year-on-year. Micron crossed $1 trillion in market cap. Google Cloud grew 63%. AWS grew 28%. The hyperscalers collectively committed $690 billion in AI infrastructure capex for 2026. Every significant prediction made at the start of the year about AI demand was too conservative.

But the year also produced complications that nobody fully anticipated. An Iran conflict disrupted the Strait of Hormuz in late February and pushed Brent crude above $100. Inflation, which had been declining, reversed course sharply. Core PCE ended the year at 3.3%, the highest in three years. Real wages turned negative. A new Federal Reserve chair arrived with no easy options. The geopolitical backdrop became more fragile, not less, as the year progressed.

What the year taught, more than any individual data point or earnings result, is that markets can hold contradictions for longer than seems rational. Record equity highs and rising existential risks coexisted for months. Rate hike odds went from zero to 45% in six weeks while the S&P 500 continued to set records. Dell grew 88% while Walmart’s customers were editing their baskets. Both things were true simultaneously. The market priced both simultaneously. Understanding that contradictions can persist is not the same as ignoring them. It is the beginning of investing clearly.

The other lesson the year taught is simpler. Showing up every week, doing the work, and building a perspective grounded in evidence rather than prediction, that is the practice. Not every issue identified the right trade. Not every analysis proved correct. But the discipline of doing it consistently, of looking at what actually happened rather than what was expected to happen, compounds over time. That is what this newsletter is. That is what the next year will be.

Thank you for reading every issue. The edge is built one week at a time. On to year two.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

The Week of Sharp Reversals

Markets opened Monday under pressure as the 10-year Treasury yield hit its highest level in a year. Tech stocks led the losses. The Nasdaq fell 0.76%, the S&P 500 shed 0.29%, and the Dow barely held flat. Bond yields, not earnings, were setting the tone at the start of the week.

Wednesday snapped the losing streak. Stocks staged a strong recovery as risk appetite returned ahead of Nvidia’s results. The Nasdaq gained 1.54%, the S&P 500 rose 1.08%, and the Dow retook the 50,000 level. Technology and consumer discretionary led, up more than 2% each. Energy, staples, and healthcare were the only sectors to decline.

Nvidia reported after Wednesday’s close. Revenue came in at $81.6 billion, up 85% year-on-year and ahead of the $79.19 billion consensus. Adjusted EPS of $1.87 beat the $1.77 estimate. Data centre revenue reached $75.2 billion, up 92%. Nvidia raised its quarterly dividend from $0.01 to $0.25 per share and announced an $80 billion share repurchase authorisation. By any historical standard, it was an extraordinary result.

Thursday reversed sharply anyway. The S&P 500 and Dow both shed 0.5% while the Nasdaq 100 dropped 0.7%. Nvidia shares hovered near flat as investors processed Q2 guidance of $89.18 to $92.82 billion, which, while strong, landed below the upper range of analyst estimates that had been building for weeks. Walmart reported the same morning: revenue up 7.3% to $177.8 billion, eCommerce up 26% globally, but the stock fell 6.43% in its worst single-day drop in years as fuel costs hit operating income and forward guidance disappointed.

Wednesday also changed the IPO landscape entirely. SpaceX filed its public S-1 with the SEC at a $1.75 trillion valuation. Within the same 24-hour window, OpenAI announced it was preparing to confidentially file for a US IPO, targeting a September listing above $1 trillion. With Anthropic targeting an October listing, the AI IPO summer has officially begun.

For the week, the S&P 500 ended roughly flat. The shape of the week said more than the closing numbers. Record results, record filings, and a market increasingly unwilling to reward anything less than perfection.

Week Ahead

Monday, May 25

US markets are closed for Memorial Day. Investors head into the holiday weekend with the S&P 500 just below its all-time high of 7,501, rate hike odds sitting at 45%, and the most important economic data of the month arriving across the back half of the week.

Tuesday, May 26

The May Consumer Confidence Index drops from the Conference Board, the first major sentiment read since the summit disappointment and Walmart’s cautious guidance. AutoZone and Zscaler report. Durable Goods Orders for April also land, giving a forward-looking read on business investment intentions.

Wednesday, May 27

FOMC minutes from the May meeting drop at 2 p.m. This is the first detailed public window into how the Fed under Kevin Warsh is framing the inflation versus growth dilemma. Any language suggesting rate hikes are being actively considered rather than merely discussed would move markets significantly. Salesforce, Marvell Technology, and Snowflake all report, giving a further read on enterprise AI software demand.

Thursday, May 28

The second estimate of Q1 2026 GDP drops alongside initial jobless claims. If the GDP revision is downward and claims are rising, it creates a stagflation signal at exactly the moment the Fed minutes are being digested. Dell, Costco, Dollar Tree, Best Buy, and MongoDB all report. Costco’s commentary on member spending and private label penetration will add another dimension to the Walmart consumer read from earlier in the week.

Friday, May 29

The April PCE inflation print lands at 8:30 a.m. ET. PCE is the Federal Reserve’s preferred inflation measure, and this reading will either confirm or challenge the CPI picture from two weeks ago. Core PCE is the number that matters most. If it accelerates beyond the March reading, the rate hike conversation becomes a rate hike reality. Personal income and spending data land simultaneously, completing the picture of whether the consumer is holding or beginning to crack under the weight of inflation and elevated energy prices.

Nvidia Beat. The Market Did Not Care. Here Is Why.

Revenue up 85% year-on-year. Data Centre up 92%. Adjusted EPS of $1.87, beating the $1.77 estimate. An $80 billion buyback. A dividend increase from $0.01 to $0.25 per share. By almost any historical measure, Nvidia’s Q1 fiscal 2027 result was extraordinary.

The stock barely moved.

This is one of the most important concepts in long-term investing, and Nvidia’s week illustrated it with unusual clarity. A great business and a great stock are not the same thing. The price you pay determines the return you receive. And when expectations outrun even the best businesses, a genuine beat can feel like a miss.

Here is what happened. Through the quarter, Wall Street analysts revised their Q2 guidance estimates upward repeatedly, chasing Nvidia’s momentum. By the time Nvidia reported, the consensus Q2 revenue estimate had drifted well above $86 billion. Nvidia guided Q2 revenue of $89.18 to $92.82 billion, a range that beats the average estimate but whose midpoint landed below the high-end expectations that had become common in analyst models. In a company with a more modest track record, that guidance would have been celebrated. For Nvidia, in May 2026, it was read as deceleration.

Jensen Huang called what is happening “the largest infrastructure expansion in human history.” The data supports that framing. Data Centre revenue of $75.2 billion represents a business larger than most Fortune 500 companies, growing at 92% year-on-year. Blackwell is in the heart of its production ramp. Vera Rubin samples are already in customer hands. Sovereign AI tripled to over $30 billion in the fiscal year. Hyperscaler capex has been revised up to approximately $725 billion for 2026.

The bull case remains as intact as it has ever been. The question for long-term investors is a different one: at what price is that bull case already reflected? Nvidia’s market cap implies a set of expectations that even a record-breaking quarter barely satisfies. That is not an argument to sell. It is an argument to understand precisely what you are paying for and why, and to hold with the conviction that comes from that clarity rather than the momentum that preceded it.

The most durable long-term positions are held by investors who know the difference between a stock moving because a business is compounding and a stock moving because expectations are compounding. With Nvidia, right now, both are happening simultaneously. The discipline is in knowing which one you are exposed to.

Walmart Told the Real Consumer Story

Revenue up 7.3% to $177.8 billion. Global eCommerce up 26%. US comp sales up 4.1%. On the surface, Walmart delivered another strong quarter. Look one layer deeper and the picture changes.

Transactions were up 6.2%. Average ticket was down 2.2%.

That combination is the clearest single data point of the week. Customers are visiting Walmart more often but spending less per trip. They are not abandoning the store. They are editing the basket. Buying necessities. Cutting discretionary items. Trading branded products for private label. This is the behaviour of a consumer under genuine financial pressure, not a consumer in crisis, but one who is making real adjustments to the way they spend.

The operating margin told the same story from the supply side. Operating income rose just 5%, hit by 250 basis points of headwind from higher fuel costs in distribution and fulfilment. Walmart is absorbing the energy shock through its logistics network, and that absorption is compressing profitability even as revenue grows. The company maintained its full-year guidance unchanged, but the Q2 adjusted EPS outlook of $0.72 to $0.74 was below what analysts had hoped for, and the stock fell 6.43% on the day, its worst single session in years.

The more important signal is what Walmart’s results say about the broader consumer. Walmart has been gaining share among higher-income households in recent years as its merchandise quality has improved and its eCommerce and grocery business has expanded. If even higher-income consumers are now trading down, visiting more frequently but spending less per visit, it confirms that inflation is reaching further up the income distribution than the aggregate data suggests.

For context: food at home prices rose 0.7% in a single month in April, the biggest monthly gain since August 2022. Gasoline remains above $4.50 nationally. Real wages are negative for the first time in three years. These are not abstractions. They show up in the basket size of every Walmart shopper in the country. And Walmart’s results confirmed they are showing up right now.

For long-term investors, Walmart is one of the most instructive businesses to watch in an inflationary environment. It has genuine pricing power, a massive logistics network, and an eCommerce operation growing at 26% globally. These are structural strengths that do not disappear in a difficult consumer environment. But the margin compression from fuel costs, and the average ticket decline that signals consumer stress, are real near-term headwinds. Understanding the difference between cyclical pressure and structural impairment is the work. Walmart looks like the former.

The $2.75 Trillion IPO Summer

On a single Wednesday in May, SpaceX filed its public S-1 with the SEC and OpenAI announced it was preparing to confidentially file for a US initial public offering. Anthropic is targeting an October listing. Together, these three companies represent a wave of new equity supply with no modern precedent.

SpaceX, which merged with Elon Musk’s xAI in February 2026, disclosed $18.67 billion in consolidated 2025 revenue in its S-1. The combined entity is targeting a valuation of approximately $1.75 trillion, which would make it the largest IPO in history by a significant margin. Goldman Sachs is leading the offering, which could price as early as June and raise up to $75 billion. Starlink generates most of the company’s growth and profitability. xAI, which includes the Grok AI model, adds an AI narrative but also a meaningful loss contribution.

OpenAI’s confidential filing, announced the same day, comes after the company fended off Elon Musk’s legal challenge in Oakland and resolved a major governance dispute that had clouded its path to the public markets. The company is targeting a September listing at a valuation above $1 trillion. Revenue grew 225% to approximately $13 billion in 2025, with an annualised run rate above $20 billion at year end. OpenAI does not expect to turn a profit until 2030. It is working with Goldman Sachs and Morgan Stanley on the prospectus.

Anthropic is targeting an October listing at approximately $900 billion. The three companies combined represent potential new equity supply approaching or exceeding $135 billion, a scale with little modern precedent and a direct test of public market appetite for pre-profit AI businesses at extraordinary valuations.

The question for long-term investors is not whether these companies are important. They clearly are. The question is how to price a business rationally when the revenue is growing at triple digits, the profits are years away, and the competitive landscape is shifting monthly. OpenAI’s price-to-sales ratio, based on its $1 trillion target valuation and $13 billion in 2025 revenue, is approximately 77 times. SpaceX, based on its $1.75 trillion valuation and $18.67 billion in revenue, is approximately 94 times. These are not traditional valuation frameworks. They are bets on the size of the markets these companies will eventually serve.

That bet may be correct. But the discipline required to make it intelligently, rather than reactively, is exactly what long-term investing demands. The IPO summer of 2026 will test that discipline for investors.

The Week That Defines the Second Half

The week of May 25 to 29 is a shortened one. Markets close Monday for Memorial Day. What arrives across four trading days could define the direction of markets for the rest of the summer.

The FOMC minutes drop Wednesday afternoon. This is the first detailed public window into how the Federal Reserve under Kevin Warsh is framing the dilemma it faces: a labour market that refuses to soften, inflation that is spreading beyond energy, and a dual mandate that is pulling in opposite directions. The language in those minutes will matter as much as any individual data point. If the committee is using the word “hike” rather than “hold,” markets will reprice quickly.

Thursday brings the second estimate of Q1 2026 GDP. The first estimate came in at 2.1% annualised growth, slightly below expectations. A downward revision alongside rising inflation would be the clearest stagflation signal of the cycle so far, the combination that puts the Fed in its most uncomfortable position. Rising prices and slowing growth simultaneously, with no clean policy response.

Friday closes the week with the April PCE inflation print, the Federal Reserve’s preferred inflation gauge. Core PCE in March came in at 2.6% year-on-year. Any acceleration from that level, and in particular any reading above 2.8%, would validate the rate hike bets that have been building since the CPI print two weeks ago. Personal income and spending data land at the same time, completing the picture of whether the consumer is absorbing or beginning to crack under the pressure of elevated energy costs and declining real wages.

The earnings calendar is also meaningful. Costco’s commentary on member spending trends and private label penetration will add another dimension to Walmart’s consumer read. Salesforce and Marvell will give a further read on enterprise AI software and semiconductor demand. Dell will provide insight into the PC and server refresh cycle that is increasingly being driven by AI workload requirements.

For long-term investors, the week ahead is not primarily about any single data point. It is about the picture that emerges when all of them are placed next to each other: inflation, growth, consumer, and Fed in the same five-day window. The last time all four were simultaneously in focus was late 2022, and the market’s response to that convergence was decisive. This time the starting conditions are different, earnings are stronger, the labour market is more resilient, and the AI investment cycle is injecting genuine demand into the economy, but the tension is real. The week ahead is where the second half of 2026 begins to take shape.

Closing Thoughts

Nvidia delivered the best result in its history. The stock barely moved. Walmart delivered a solid revenue quarter. The stock had its worst day in years. SpaceX and OpenAI filed for IPOs at combined valuations approaching $3 trillion. The market absorbed it and kept going.

What the week taught long-term investors is something that does not change regardless of the macro environment: the relationship between price and value is the only variable that ultimately determines returns. Nvidia is an extraordinary business. Whether it is an extraordinary investment at $3.5 trillion depends on what you believe about the next decade of AI infrastructure spending. Walmart is a resilient, well-run retailer. Whether Thursday’s selloff was a buying opportunity depends on whether you read the basket size decline as cyclical consumer pressure or structural trade-down.

These are not questions that resolve in a single week. They are the ongoing work of long-term investing. And that work requires the kind of clarity that does not come from watching prices move but from understanding what you own, why you own it, and what would have to be true for your thesis to be wrong.

The week ahead will add more data to that picture. The week after that will add more still. The edge is built by accumulating that understanding consistently, one issue at a time.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Market Overview

Seven Weeks. Then a Reality Check.

The S&P 500 rose to a record 7,501.24 on Thursday, breaking above 7,500 for the first time in history. The Dow crossed 50,000, its first time above that level since February. The Nasdaq hit a fresh record at 26,635.22. For a brief moment, this was the best-looking market in years.

Then Friday happened.

The Trump-Xi summit in Beijing ended without any substantive agreement on Iran or the Strait of Hormuz. Markets had been pricing in a breakthrough. They didn’t get one. The S&P 500 shed 1.24% to close at 7,408.50. The Nasdaq dropped 1.54%. The Dow fell 537 points. Semiconductor stocks bore the brunt. Intel fell over 6%, AMD and Micron lost more than 5%, Nvidia dropped 4.4%. For the week overall, the major indexes ended roughly flat, but the shape of the week told a more complicated story.

The defining data point came Tuesday morning. April CPI came in at 3.8% year-on-year, the highest since May 2023. Core CPI rose 0.4% for the month, the sharpest monthly increase since January 2025, and 2.8% annually. Energy accounted for over 40% of the headline gain, but the more important detail is that inflation is no longer confined to energy. Shelter, transportation, and apparel are all accelerating. For the first time in three years, real average hourly wages turned negative. Workers are earning less in purchasing power terms than they were a year ago.

Also this week, the Senate confirmed Kevin Warsh as the new Federal Reserve Chair. He inherits a central bank with no good options. Rate hike odds on futures markets jumped to 45% by Friday’s close, up from essentially zero a month ago. Rate cuts, which were the dominant market narrative heading into 2026, are no longer on the table.

Geopolitically, the Iran conflict remains unresolved. The Strait of Hormuz remains disrupted. Brent crude finished the week above $104. The summit produced statements, not solutions. And with Nvidia reporting on Wednesday and the full wave of consumer earnings arriving Thursday, the week ahead will test whether the AI trade and the consumer can hold up simultaneously against a macro backdrop that is becoming harder to ignore.

Week Ahead

May 18 to 22: Nvidia, the Consumer, and the Fed

Monday, May 18.

The NAHB Housing Market Index for May is the main data release, giving an early read on builder confidence heading into the spring selling season. Baidu reports, worth watching for signals on Chinese AI adoption and enterprise cloud demand in the wake of the Beijing summit.

Tuesday, May 19.

Home Depot reports before the open. Its results will be one of the clearest early signals on whether $4.50 gasoline and 3.8% inflation are visibly pulling back big-ticket discretionary spending. Pending Home Sales for April also drop, alongside ADP weekly employment data.

Toll Brothers and Cava Group also report.

Wednesday, May 20.

Nvidia reports after the close. Wall Street expects earnings per share of $1.78 on revenue of $78.98 billion, nearly double the year-ago figures. Jensen Huang returns from Beijing, having just accompanied the President of the United States on a state visit to China, and his commentary on the China chip sales question, the Blackwell demand pipeline, and the path to $1 trillion in combined revenue through 2027 will set the tone for the entire AI trade heading into summer. At 2 p.m., FOMC minutes from the last meeting drop, the first public window into internal Fed deliberations under new Chair Kevin Warsh.

Also reporting: Intuit, TJX, Lowe’s, Target, and Analog Devices.

Thursday, May 21.

Walmart reports before the open. Wall Street expects earnings per share of 66 cents on revenue of $174.62 billion, up 6.5% year-on-year. As the world’s largest retailer, Walmart’s commentary on spending trends, private label penetration, and whether higher-income consumers are trading down will be the most important consumer data point of the quarter. Housing Starts, Initial Claims, the Philadelphia Fed Index, and preliminary PMI data for both manufacturing and services all land the same morning.

Deere, Workday, Ross Stores, and Ralph Lauren also report.

Friday, May 22.

The final University of Michigan consumer sentiment reading for May drops at 10 a.m. After weeks of inflation data, a summit that disappointed, and rising rate hike odds, this number will tell us whether the consumer is beginning to feel the weight of it all or whether confidence is holding despite everything.

Macro

Inflation Is No Longer Just an Energy Story

The April CPI print was bad. What made it worse is where the pressure is coming from.

Headline inflation at 3.8% year-on-year was expected. Energy costs have been rising since the Iran conflict began in late February, and the gasoline index is now up 28.4% annually. That part of the story was already priced in. What markets were not fully prepared for was the breadth of the acceleration underneath it.

Core CPI, which strips out food and energy precisely to identify structural inflation, rose 0.4% in a single month. That is the sharpest monthly increase since January 2025. Shelter costs rose 3.3% year-on-year. Transportation services rose 4.3%. Apparel rose 4.2%. Services inflation, the category most sensitive to wage pressures and the hardest for the Fed to bring down, rose 3.3% broadly.

The most uncomfortable data point for workers: real average hourly wages fell 0.5% for the month and 0.3% annually. For the first time in three years, Americans’ pay is no longer keeping pace with prices. The cost of living is rising faster than the wages used to meet it.

Economists and commentators have begun drawing the comparison to the 1970s OPEC oil embargo openly. The parallel is not exact, but the mechanism is similar. An external energy shock pushes headline inflation higher. That shock spreads into non-energy categories through transportation costs and supply chain pressures. Workers demand higher wages to compensate. Those wages feed back into services inflation. The Fed faces a choice between fighting inflation and protecting growth and finds it cannot do both simultaneously.

Kevin Warsh inherits a Federal Reserve at exactly this inflection point. Rate hike odds jumped to 45% on Friday from essentially zero a month ago. The FOMC minutes dropping Wednesday will be the first public window into how the Fed is thinking about this under new leadership. The answer will matter, not just for rate expectations, but for the valuation of every long-duration asset in the market.

For long-term investors, the lesson from every inflationary period in modern history is consistent. Companies with genuine pricing power, the ability to pass rising costs to customers without losing volume, compound through inflation. Companies without it get squeezed. The retailers reporting next week will tell us, in real time, which category they fall into.

Geopolitics

The Summit That Did Not Solve the Problem

Trump and Xi met in Beijing for two days. The optics were extraordinary. The US President arrived with Jensen Huang, Elon Musk, and Tim Cook in tow, a delegation that spoke as loudly as anything said in the meeting rooms. The venue was Zhongnanhai, China’s leadership compound. Trump called it “a great couple of days”.

What came out of it: a trade council, an investment council, some agricultural market access agreements, a statement that the Strait of Hormuz should remain open, and an invitation for Xi to visit the White House in September.

What did not come out of it: any substantive commitment on Iran, any mechanism for reopening the Strait of Hormuz to normal commercial traffic, any resolution on Taiwan’s $11 billion arms package, or any framework for addressing the underlying conflict that has been disrupting global energy markets since February.

Xi told Trump that China wants to help negotiate an end to the war and the reopening of the Strait. China’s foreign ministry called the summit “historical.” Trump said he had made “fantastic trade deals.” But when pressed on whether Xi made any commitments to pressure Iran, Trump said “we don’t need favors” and suggested “a little cleanup work” may still be required. That is not the language of a breakthrough.

Markets had been pricing in something more decisive. The S&P 500’s sharp selloff on Friday and the 45% rate hike odds that appeared in futures markets by the close reflected a recalibration. The summit was constructive but not conclusive. The Iran conflict remains unresolved. The Strait of Hormuz remains disrupted. Brent crude remains above $100. And inflation, as Tuesday’s CPI print confirmed, is no longer just a geopolitical artefact. It is becoming structural.

The Taiwan question added another layer of unresolved tension. Xi warned Trump that disagreements over the self-governed island could lead to clashes and conflicts if the $11 billion arms package moves forward. Trump said he had not decided whether to proceed. That ambiguity is a signal in itself and a risk premium that markets have not yet fully priced.

For long-term investors, geopolitical risk is always hardest to price because it does not resolve on a schedule. The summit was a step, not an answer. The energy shock that is driving inflation, compressing consumer purchasing power, and paralysing the Fed remains intact. That is the context for everything that happens next week.

Earnings

Nvidia Wednesday: The Number That Defines the AI Era

On Wednesday, May 20, after the closing bell, Nvidia reports. Wall Street expects earnings per share of $1.78 on revenue of $78.98 billion. A year ago, the company reported 81 cents per share on $44.06 billion in revenue. If the estimates are right, Nvidia will have nearly doubled its revenue year on year, one of the most extraordinary growth trajectories at this scale in the history of the stock market.

This is not just an earnings report. It is a check on the entire AI investment thesis.

The $690 billion in combined AI capex committed by the four hyperscalers this year ultimately flows through Nvidia’s order books. Every percentage point of cloud revenue growth that Google, Amazon, and Microsoft reported last month depends, in part, on the compute that Nvidia manufactures. The AI trade is, at its core, a bet that demand for Nvidia’s chips will continue to exceed supply for long enough to justify the extraordinary valuations placed on the entire ecosystem around it.

Jensen Huang returns from Beijing having just stood alongside the President of the United States at a state visit to China, a moment that crystallises how central Nvidia has become not just to the technology sector but to global geopolitics. The China chip sales question will be central to the conference call. Any signal of easing on export controls would be a significant positive for the revenue outlook.

At Nvidia’s GTC conference in March, Huang updated the company’s forecast to exceed approximately $1 trillion in combined Blackwell and Rubin revenue through calendar 2027. Analysts have raised their targets accordingly. The question on Wednesday is not whether Nvidia is dominant. It is whether the guidance extends the runway or introduces the first signs of deceleration.

The FOMC minutes, also dropping Wednesday afternoon, add a second layer of market-moving potential to the day. Together, Nvidia earnings and the Fed minutes will tell us whether the AI tailwind and the monetary policy headwind are on a collision course or whether both can coexist for another quarter.

Consumer

The Consumer Is the Last Domino

Everything in this market eventually rests on the consumer. And this week, the consumer gets its report card.

Walmart, Home Depot, Target, TJX, Lowe’s, and Ross Stores all report between Tuesday and Thursday. Combined, they represent the most comprehensive real-time read on whether 3.8% inflation and $4.50 gasoline are visibly changing how Americans spend money. That question matters not just for the retail sector but for the entire economic outlook. Consumer spending accounts for more than two-thirds of US GDP.

The early signals are mixed. April CPI data showed food at home prices rising 0.7% in a single month, the biggest monthly gain since August 2022. Real wages are negative for the first time in three years. The national average gasoline price topped $4.50 for the first time since 2022 earlier this month. These are the conditions under which consumers historically begin to pull back, trading down to private-label products, cutting discretionary categories, and shifting spending towards necessities.

Walmart is the most important report of the week. The world’s largest retailer has been gaining share among higher-income consumers in recent years, driven by improving merchandise quality and the convenience of its expanding e-commerce and grocery business. If higher-income consumers are now trading down to Walmart in meaningful numbers, it is a signal that inflation is reaching further up the income distribution than the headline data suggests. Management commentary on private label penetration, grocery comp trends, and whether customers are changing the size or frequency of their baskets will be the data points to watch.

Home Depot faces a different dynamic. Big-ticket home improvement spending is one of the most rate-sensitive and inflation-sensitive categories in the economy. With mortgage rates still elevated and home equity drawdowns slowing, the appetite for major renovation projects has been compressing. Home Depot’s results will tell us whether that compression is accelerating and whether the housing market slowdown is beginning to ripple into the home improvement supply chain.

For long-term investors, the retailer earnings wave is not just a sector-specific story. It is the ground-level confirmation test for the macro thesis. If inflation is hurting consumers, it will show up here first, in margins, in unit volumes, in management tone. If the consumer is holding, it will show up here too. Either way, this week answers a question that has been building for months.

Closing Thoughts

The Market That Holds Two Things at Once

Records on Thursday. A 500-point drop on Friday. Both happened in the same week. Neither cancels the other out.

The market’s ability to hold two things simultaneously, all-time highs and rising existential risks, is both its greatest strength and its most persistent source of confusion for investors trying to make sense of it. The S&P 500 broke 7,500 for the first time in history. Rate hike odds hit 45%. A geopolitical summit delivered photo opportunities instead of solutions. And the most important earnings report of the quarter is still three days away.

This is the market as it actually is, not as it is easy to describe. The job of a long-term investor is not to resolve that tension on any given week but to stay positioned in businesses that can compound through it regardless of which way it resolves in the short term.

The week ahead will answer several questions at once. Whether Nvidia’s guidance extends the AI runway. Whether the consumer is beginning to crack under inflation pressure. Whether the Fed’s new leadership is more hawkish than markets currently price. Any one of those outcomes could reset the narrative for the rest of the quarter. All three arriving in the same week is unusual and worth paying attention to.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

This week the market continued to ignore every reason it shouldn’t be at record highs and push higher anyway. Six consecutive weeks of gains. New all-time highs on the S&P 500 and Nasdaq. And underneath the surface, a set of stories that long-term investors need to understand before the next chapter of this market is written.

Market Overview

Six weeks in a row.

The S&P 500 gained 2.3% for the week, closing at a record 7,398.93. The Nasdaq surged 4.5% to 26,247.08, also a record, marking the longest winning streak for both indexes since 2024. The Dow lagged, adding just 0.2% to settle at 49,609.

The defining moment came Friday morning. April nonfarm payrolls came in at 115,000, nearly double the 65,000 consensus estimate, with unemployment holding steady at 4.3%. Job gains were driven by healthcare, transportation, and retail. Federal government employment declined. The labour market, in short, is not breaking.

That single number simultaneously relieved growth fears and reignited the rate hike conversation. Back-to-back months of stronger-than-expected jobs data have shifted the probability distribution. Rate cuts are off the table. Rate hikes are being actively discussed. The Fed’s press conference language from its last meeting, with three hawkish dissents against the forward guidance, now reads differently in light of Friday’s print.

AMD was the stock of the week, surging 20% on the back of a decisive earnings beat and crossing $700 billion in market cap for the first time. Memory stocks were the sector story; the Roundhill Memory ETF gained nearly 30% in a single week. The S&P 500 technology sector rose over 35% since April 27, one of the most concentrated and rapid sector moves in recent history.

Geopolitically, the Iran conflict continues to anchor Brent crude above $110 and drive gasoline prices nationally above $4.50 for the first time since 2022. The US-China summit scheduled for May 14–15 has become an informal market deadline; both sides have been urging Iran towards diplomatic resolution ahead of the meeting. How that plays out next week will matter.

Week Ahead

Monday, May 11.

Existing Home Sales data for April lands at 10 a.m., giving a read on whether elevated mortgage rates and high energy costs are cooling the housing market. Simon Property Group, Mosaic, and Constellation Energy report. The NFIB Small Business Optimism Index also drops, an often-overlooked gauge of ground-level economic sentiment.

Tuesday, May 12.

April CPI lands at 8:30 a.m. ET. Headline inflation is expected to come in at 0.6% month-on-month and 3.7% year-on-year, driven by the energy shock from the Iran conflict. The core reading, which strips out food and energy, is the number that matters most. March core CPI was 2.6% year-on-year. Any meaningful acceleration from that level will intensify rate hike expectations and potentially rattle a market that has been remarkably calm about inflation risk. Alibaba and Sea Limited also report useful reads on Chinese consumer health ahead of the summit.

Wednesday, May 13

Producer Price Index data alongside Cisco’s earnings. PPI gives a forward-looking read on consumer price pressures, if input costs are rising, CPI follows. Cisco’s results will test whether enterprise AI spending is translating into network infrastructure demand. Analysts expect earnings per share of $1.04 on revenue of $15.6 billion.

Thursday, May 14

The US-China summit between President Trump and President Xi begins on May 14–15. Both leaders have been publicly urging Iran towards a diplomatic resolution ahead of the meeting, making this summit an informal deadline for the conflict. Retail sales data for April also drops, the clearest read yet on whether $4.50 gasoline is visibly hurting consumer spending. Applied Materials report; a direct window into semiconductor equipment demand and AI infrastructure buildout velocity.

Friday, May 15

Industrial Production and Capacity Utilisation for April drop at 9:15 a.m. The Empire State Manufacturing Survey is also released. These provide a broader read on whether the energy shock is beginning to show up in factory activity and output.

AMD and the Other AI Trade

The AI chip story has had one name for two years. This week it got a second one.

AMD reported first-quarter revenue of $10.3 billion, with non-GAAP earnings per share of $1.37, beating the $1.25 estimate by nearly 10%. Data centre revenue reached $5.8 billion, making it the primary driver of both revenue and earnings growth for the first time in the company’s history. That is a structural milestone, not a quarterly beat.

The stock surged 20% on the week. AMD crossed $700 billion in market cap. It is now up close to 90% over the past month.

What the AMD results tell us is that AI infrastructure spending is broadening. Nvidia remains the dominant GPU supplier. But the AI build-out requires an entire ecosystem of chips, and AMD is building its position in that ecosystem methodically. The Instinct MI450 accelerator. AMD’s most advanced chip to date positions it to compete for workloads that don’t require Nvidia’s most premium hardware. That is a very large addressable market.

For long-term investors, the AMD story illustrates a principle worth remembering. The first wave of any transformational technology is often captured by one or two dominant players. The second wave tends to reward the companies building credible alternatives. AMD is the most credible alternative in AI compute. The market is beginning to realise that the AI trade is not a single-name story, and AMD’s results just confirmed it.

The Memory Supercycle

The Roundhill Memory ETF, dedicated exclusively to memory and storage chips, gained nearly 30% in a single week.

Micron hit an all-time high, up over 120% year-to-date. The reason is structural, and most investors are still missing it.

High Bandwidth Memory, or HBM, is the specialised memory that gets bonded directly to AI GPUs. Every Nvidia and AMD accelerator requires it. Every hyperscaler data centre is full of it. And right now, the entire global supply of HBM is sold out through the end of 2026. Micron began shipping HBM4 in April, the newest generation, with bandwidth exceeding 2.8 terabytes per second and pricing up over 50% on the prior generation.

The numbers behind the supercycle are significant. Global DRAM revenue is forecast to surge 51% year-over-year in 2026, driven almost entirely by the AI demand wave. The HBM market specifically is projected to grow from $35 billion in 2025 to $100 billion by 2028, larger than the entire DRAM market was in 2024. Micron, SK Hynix, and Samsung have collectively redirected their most advanced manufacturing capacity to HBM production, creating a shortage in general-purpose memory that is pushing prices higher across every category of electronics.

Micron is the only US-based manufacturer of advanced memory. That gives it a unique strategic position as hyperscalers, particularly Microsoft and Amazon, increasingly seek supply chain resilience and domestic sourcing options.

The memory supercycle is not a short-term trade. The structural drivers, more HBM required per GPU with every generation, hyperscaler capex accelerating, and AI model complexity compounding, are multi-year in nature. The companies with the manufacturing capacity, the technology, and the customer relationships to serve that demand are in a rare position. This is what a structural tailwind looks like at the beginning, not the end, of a cycle.

CoreWeave: The First Crack in the AI Infrastructure Story?

CoreWeave reported first-quarter revenue of $2.1 billion, more than double the $982 million from a year earlier. By almost any measure, that is extraordinary growth.

And yet the stock fell on Friday.

The reason: CoreWeave guided second-quarter revenue of $2.45–$2.6 billion, meaningfully below the $2.69 billion consensus estimate. Simultaneously, the company raised its full-year capital expenditure estimate to $31–$35 billion, a figure that implies significant cash burn before the revenue growth closes the gap.

Set alongside the OpenAI revenue miss reported last week, missed monthly targets, concerns from its CFO about funding its own compute commitments, CoreWeave’s guidance miss is now the second consecutive data point suggesting that AI infrastructure monetisation is taking longer than the spending curves imply.

This is worth sitting with honestly, because it doesn’t resolve cleanly in either direction.

The bull case remains intact. CoreWeave’s revenue backlog stood at $66.8 billion at the end of 2025, more than four times where it began the year. The demand for AI cloud compute is real and accelerating. First-quarter revenues doubled year-on-year. Management guided for a revenue of $8.7 billion for the full year, a number that, if achieved, would represent one of the fastest revenue ramps in cloud computing history.

The bear case is also real. OpenAI, CoreWeave’s largest reported customer is itself struggling to grow revenue fast enough to justify its own compute commitments. If OpenAI’s growth disappoints, CoreWeave’s backlog becomes less certain. The capital expenditure required to build out the infrastructure is enormous. Negative free cash flow at this scale requires either continued access to capital markets or revenue acceleration that hasn’t fully materialised yet.

For long-term investors, CoreWeave is not a verdict on AI; it is a reminder that infrastructure build-outs move faster than the monetisation that follows them. That has always been true. The question is whether the revenue eventually arrives in proportion to the spending. The evidence so far says yes, just more slowly than the most optimistic projections assumed.

The Fed’s Impossible Position

The Federal Reserve is caught in a trap it has not faced in four decades.

The labour market is strong. April added 115,000 jobs, nearly double what economists expected, following March’s 178,000, which was itself far above consensus. Unemployment holds at 4.3%. By the Fed’s own historical framework, these are healthy numbers.

Inflation is rising. March headline CPI came in at 3.3% year-on-year, the highest reading in two years, driven almost entirely by the energy shock from the Iran conflict. Gasoline has topped $4.50 a gallon nationally for the first time since 2022. April CPI, due Tuesday, is expected to show further acceleration to 3.7% year-on-year.

The Fed’s dual mandate, maximum employment and price stability, is pulling in opposite directions simultaneously. Cutting rates would ease any pressure on the labour market but risk stoking inflation that is already running hot. Raising rates would address inflation but risks tipping an economy that is still growing into something more fragile.

Ray Dalio called it stagflation outright this week. Apollo’s chief economist flagged similar risks. The Fed itself, at its last meeting held rates steady at 3.50%–3.75% with three hawkish dissents. The market has moved from pricing in cuts earlier this year to now pricing in no movement at all. Rate hike bets are beginning to appear in futures markets for the first time in this cycle.

The US-China summit on May 14–15 has become an important variable in this equation. Both sides have been urging Iran towards a diplomatic resolution. If a peace deal, or even a credible ceasefire, emerges from the summit, energy prices could fall meaningfully, headline inflation could ease, and the Fed’s path becomes clearer. If the Strait of Hormuz remains closed, the energy shock extends, inflation stays elevated, and the Fed’s dilemma deepens.

For long-term investors, the lesson from every previous episode of elevated inflation and strong growth is the same. Corporate earnings, which are compounding at 28% year-on-year, are the best anchor in an environment where monetary policy is paralysed. Companies with pricing power, strong balance sheets, and structural demand advantages can navigate this. Companies that require cheap money to fund their growth cannot.

That distinction is what matters right now. Not the rate decision. Not the CPI print. But which businesses are durable enough to compound through uncertainty, and which ones are not.

Closing Thoughts

Six weeks of consecutive gains. Record highs. A labour market defying expectations. And an inflation picture that is getting more complicated by the week.

The market’s ability to look through bad news and price in good news has been extraordinary. But the test is coming. Tuesday’s CPI print, the US-China summit, and the energy market’s response to whatever emerges from that meeting will collectively tell us whether the conditions underpinning this rally remain intact.

What hasn’t changed, and what won’t change regardless of how those events unfold, is the long-term investment case for the businesses at the centre of this issue. AMD is widening the AI chip market. Micron is the only US manufacturer sitting at the heart of a structural supercycle. The Fed’s dilemma will resolve itself, one way or another, over quarters and years.

Long-term edge is not built in a single week. It is built by staying positioned in businesses that can compound through uncertainty and having the patience not to confuse short-term noise for long-term signal.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Welcome to Issue 48 of The Long Term Edge, your weekly guide to compounding over 7+ years.

This week delivered the most consequential earnings season of 2026. The Magnificent Seven reported. The Fed held. April closed as the best month for equities since 2020. And underneath all of it, a set of quieter stories told us more about the real state of the market.

Market Overview

April ended the way long-term investors dream about.

The S&P 500 gained 10.4% for the month, its best performance since November 2020. The Nasdaq rose 15.3%, its best month since April 2020. The Dow added 7.1%, its strongest month since November 2024. All three indexes closed the week at or near all-time highs, with the S&P 500 breaking above 7,200 for the first time ever.

The week itself was more nuanced. The S&P 500 ended Friday at 7,230, up 0.29% on the day. The Nasdaq closed at 25,114, up 0.89%, also a record. The Dow slipped 152 points to 49,499.

What drove it was earnings, specifically, the volume and quality of beats. Analysts now see S&P 500 first-quarter earnings growth at 27.8%, the fastest pace since Q4 2021. Of the companies that have reported, 84% beat EPS estimates, above both the five-year and ten-year averages.

The macro picture was more complicated. The Fed held rates steady at 3.50%–3.75%, with three dissents against forward guidance language, interpreted as a hawkish hold. The Iran conflict continued to disrupt Strait of Hormuz shipping, pushing input prices to a four-year high. Oil remained elevated above $110 for much of the week before pulling back on Friday after reports Iran had submitted a new peace proposal through Pakistani mediators.

The tension between record equity highs and a still-live geopolitical conflict defines the market mood heading into May. So far, earnings are winning that argument.

Week Ahead

Monday May 4

Opens with volatility in focus.

Avis Budget and Hertz are set up for another sharp week based on options positioning. The Fed’s Senior Loan Officer Opinion Survey drops, tightening lending standards would be an early warning signal for financial stocks and the broader economy. IBM’s four-day Think conference begins, centred on agentic AI.

Palantir, On Semiconductor, Tyson Foods, and Pinterest all report. Devon and Coterra Energy shareholders vote on their pending merger.

Tuesday May 5

Data and earnings day of the week.

AMD reports after the close, the most direct read on AI chip demand outside Nvidia. Shopify, Pfizer, Arista Networks, and Duke Energy also report. New Home Sales data lands at 10 a.m. Fed Vice Chair Michelle Bowman speaks at the Washington Financial Symposium, any rate commentary will move markets.

Wednesday May 6

Disney headlines the morning, with analysts focused on streaming momentum, the ESPN-NFL deal, and the FCC investigation of ABC.

Uber, DoorDash, AppLovin, Arm Holdings, and Novo Nordisk all report.

is one of the most-watched prints of the week, its guidance will shape the ad-tech narrative for Q2. Anthropic holds its Code with Claude event in San Francisco. Fed speakers Goolsbee and Musalem both speak.

Thursday May 7

Airbnb, McDonald’s, Shell, and CoreWeave report.

CoreWeave deserves particular attention, as one of the most exposed companies to OpenAI’s compute spending, its results will either validate or complicate last week’s revenue miss story.

Citi holds its Investor Day, expected to be a strategy reset with a clearer roadmap on returns and capital deployment.

Friday May 8

The April jobs report drops at 8:30 a.m. ET. Economists expect just 49,000 job additions, a sharp slowdown from March’s 178,000, with unemployment holding at 4.3%. A print near or below that number reignites rate cut speculation. A surprise to the upside complicates the Fed’s already-hawkish hold. Either way, this number moves markets.

The Broader Economy Is Actually Holding Up

The mega-cap earnings dominated the headlines. But the more important story for long-term investors was happening one column to the right.

Nine of eleven S&P 500 sectors reported year-over-year earnings growth in Q1 2026, with seven of those nine delivering double-digit growth. That breadth matters. A market driven by five companies is fragile. A market where industrials, healthcare, payments, pharma, and consumer names are all beating estimates is something else entirely.

The names tell the story. Caterpillar, Eli Lilly, Mastercard, Merck, and Chipotle all topped forecasts, with fundamental strength showing up across industrial machinery, healthcare, payments, and consumer chains. Merck tightened its full-year sales guidance while raising its profit outlook, a signal of margin confidence, not defensiveness. Eli Lilly continued to demonstrate that weight-loss drug demand is not slowing.

There is one cautionary note worth watching. Mastercard reported net revenue of $8.4 billion, up 16% year-over-year, beating estimates, but disclosed a deceleration in cross-border transaction growth in April. Cross-border travel volume grew just 2% in April through the 28th, down sharply from 8% in Q1, with the Iran conflict putting clear pressure on international travel patterns.

That is worth monitoring heading into Q2. When Mastercard’s cross-border data slows, it is one of the earliest real-time reads on whether consumers are pulling back on discretionary spending. It hasn’t broken down, but it is no longer accelerating.

The takeaway for long-term investors: the foundation underneath this market is broader than the index concentration suggests. That is a healthier setup. But the Mastercard signal is a reminder to stay alert to what the consumer is actually doing, not just what the survey data says.

Caterpillar and the Hidden AI Trade

The most interesting non-tech story of the week didn’t come from a software company.

Caterpillar reported first-quarter sales and revenues of $17.4 billion, up 22% year-over-year, with adjusted earnings per share of $5.54, an increase of 30%. The headline beat was significant. But the number that really matters is the backlog.

Caterpillar’s backlog reached a record $63 billion, up 79% year-over-year, fuelled by broad-based demand across all three primary segments, with Power & Energy, where sales jumped 32%, driven directly by data centre power generation equipment.

Read that again. The company best known for bulldozers and mining equipment is now one of the most direct beneficiaries of the AI infrastructure build-out, because data centres need power, and power needs engines, turbines, and generators at massive scale.

Morgan Stanley upgraded Caterpillar from Underweight to Equal Weight on May 1, raising its target from $430 to $915, citing the record backlog and hyperscaler capital expenditure growth.

This is the kind of connection that long-term investors should be building into their thinking. The $690 billion being spent on AI infrastructure this year doesn’t just flow to Nvidia and the cloud providers. It flows to the companies supplying power, cooling, construction, and logistics for the physical infrastructure underneath it all. Caterpillar is one of those companies. It won’t be the last.

Management raised its full-year 2026 revenue growth outlook to low double digits, significantly above the 7% previously projected. With a record backlog providing visibility well into 2027, this is not a one-quarter story. It is a multi-year structural demand shift hiding in plain sight inside an industrial company most investors haven’t looked at since the last commodity cycle.

OpenAI’s Uncomfortable Week

While the hyperscalers were celebrating blowout earnings, the company at the centre of the entire AI thesis was having a very different week.

The Wall Street Journal reported that OpenAI missed its internal goal of one billion weekly active users and its target revenue for the year, with its CFO raising concerns about the company’s ability to fund its computing contracts if growth doesn’t accelerate. OpenAI has missed multiple monthly revenue targets in 2026, in part due to Anthropic’s growing success in coding and enterprise markets.

Simultaneously, the Musk versus Altman trial opened in Oakland. Musk’s lawyers alleged OpenAI had effectively stolen a charity by converting from a nonprofit to a profit-seeking entity. The market reacted immediately, Oracle fell roughly 5%, CoreWeave dropped 7%, and SoftBank sank nearly 10% as investors reassessed the commercial orbit around OpenAI.

The tension here is important and underappreciated. OpenAI has committed to $250 billion in Azure spending through 2032, a number that becomes significantly harder to justify if monthly revenue keeps slipping. That is a commitment made on the assumption of a growth trajectory that is not currently materialising.

OpenAI disputed the reporting, with Altman and the CFO calling it “ridiculous.” Analysts pointed out that revenue is still growing rapidly, likely doubling year-over-year, even if falling short of stretch targets.

But the question it raises is the right one for anyone thinking about AI as a long-term investment thesis. The hyperscalers are spending hundreds of billions on the assumption that demand will follow. The 63% Google Cloud growth and 28% AWS acceleration both suggest it is. OpenAI’s stumble is a data point, not a verdict. But it is the first meaningful evidence that converting AI hype into sustainable revenue is harder, and slower than the spending curves imply.

The $650 Billion Question

Four companies. One week. A combined capital expenditure commitment that is difficult to comprehend.

Amazon has committed approximately $200 billion in capital expenditure for 2026. Alphabet raised its guidance to $180–$190 billion. Meta lifted its range to $125–$145 billion. Microsoft raised its forecast to $190 billion. Combined, that is roughly $690 billion, nearly double 2025 levels, flowing into AI compute, data centres, and networking in a single year.

The earnings results that accompanied these figures were, by any historical standard, extraordinary. Alphabet’s profit rose 81%, with Google Cloud growing 63% to $20 billion. AWS grew 28% to $37.59 billion, its fastest growth rate in more than three years. Microsoft’s total AI business hit a $37 billion annualised revenue run rate, up 123% year-over-year.

The bull case is clear: these companies are not spending speculatively. They are spending because enterprise demand is outpacing supply, cloud backlogs are growing, and AI is beginning to generate real, compounding revenue at scale.

The bear case is also clear. Analysts flagged that the doubling of capex leaves limited free cash flow in 2026 with uncertain return on investment. Amazon is facing negative free cash flow of close to $17 billion this year, according to Morgan Stanley estimates. JPMorgan downgraded Meta to neutral, citing a challenging path to generating returns on its heavy capex forecast.

The honest answer is that both things are true simultaneously. The revenue is real. The returns are unproven. The spending is accelerating faster than the revenue can currently justify, and every CEO on these calls knows it.

When Zuckerberg was asked directly what signs he’s watching to confirm Meta is on a healthy path to a return on its AI investments, his answer pointed to model quality and product progress, not a financial return timeline. That response didn’t soothe investors.

For long-term investors, the question isn’t whether AI is real. It clearly is. The question is whether the companies spending the most are building durable competitive advantages, or whether they are participating in an infrastructure arms race that ultimately commoditises the very thing they’re racing to own.

That answer will take years to emerge. The spending decisions are being made right now.

Closing Thoughts

April 2026 will be remembered as the month the AI investment thesis went from speculative to structural.

The earnings were not marginal beats. They were blowouts, and they were broad. From Caterpillar’s record backlog to Google Cloud’s 63% growth to Apple’s best March quarter ever, the evidence that AI is generating real economic activity across the full supply chain is now substantial.

But the OpenAI story is worth sitting with. Not because it changes the thesis, it doesn’t, but because it clarifies what the thesis actually requires. AI spending at this scale can only be justified by revenue that compounds fast enough to close the gap. Right now, for the hyperscalers, it is. For the companies built on top of that infrastructure, it is proving harder and slower.

Long-term edge isn’t found in predicting which quarter the returns materialise. It’s found in owning the businesses with the structural advantages to still be standing, and compounding when they do.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

Equities pushed higher even as oil surged and geopolitical tensions escalated, a signal that markets are increasingly willing to look through short-term noise. Meanwhile, the biggest earnings week of the quarter arrives all at once, with mega-cap tech, the Federal Reserve, and a historic Berkshire Hathaway annual meeting all converging in the same seven days.

Market Overview

Wall Street closed the week sending mixed signals, but the underlying message is far clearer than the headlines suggest.

The S&P 500 and Nasdaq pushed to fresh highs, even as geopolitical tensions in the Middle East intensified and oil prices surged. That divergence matters. Markets are increasingly willing to look through short-term uncertainty rather than react to it.

Under the Trump administration, the U.S. extended ceasefire efforts with Iran while maintaining pressure through military positioning. At the same time, tensions escalated in the Strait of Hormuz, one of the world’s most critical oil routes following reported attacks on commercial vessels.

Oil responded sharply. Crude rose over 13% to around $94, while Brent climbed above $105.

Normally, that would rattle equities.

This time, it didn’t.

That’s your signal.

Markets are prioritising earnings resilience and economic strength over geopolitical noise. Retail sales surprised to the upside, business inventories rose, and investor sentiment improved, even if still historically low.

For the week:

• The S&P 500 gained 0.6%

• The Nasdaq rose 1.5%

• The Dow slipped 0.4%.

This isn’t a broad rally. It’s a selective one driven by strength in tech, AI, and dominant platforms.

Week Ahead

Monday, 27th April

Volatility is building in pockets of the market, particularly in heavily shorted names and stocks showing extreme momentum. Names like Avis Budget Group and Hertz Global Holdings are set up for sharp moves reflecting how positioning, not fundamentals, is driving short-term action.

Earnings from Verizon and Domino’s Pizza will provide early signals on consumer strength. And any weekend developments in the Middle East could ripple through energy markets from the open.

Tuesday, 28th April

Market-sensitive names including Visa, Coca-Cola, Starbucks, and Spotify report. This is where we start to see how the consumer is actually behaving, not just what surveys suggest.

It’s also a day where structural shifts may quietly show up. The opening of large-scale gaming infrastructure in New York could begin to reshape regional revenue dynamics for MGM Resorts and Caesars Entertainment.

Wednesday, 29th April

This is where the market will likely decide direction.

Earnings from Microsoft, Amazon, Meta Platforms, and Alphabet hit all at once, arguably the most important cluster of reports this quarter. Layered on top is the Federal Reserve decision, followed by a press conference from Jerome Powell.

Markets won’t just react to the rate decision. They’ll react to tone, language, and forward guidance.

Thursday, 30th April

Apple earnings will be released.

Alongside it, major names across healthcare, energy, and industrials report, giving a broader read on the economy beyond tech. Core PCE inflation data, the Fed’s preferred measure, will provide one of the clearest signals yet on whether inflation is stabilising. Commentary from Christine Lagarde will also be watched closely, reinforcing how interconnected global policy expectations have become.

Friday, 1st May

Energy giants Exxon Mobil and Chevron report, giving direct insight into how rising oil prices are translating into earnings. Global data from EV deliveries to Macau gaming revenue will provide a broader read on demand across key sectors.

Saturday, 2nd May

Berkshire Hathaway holds its first annual meeting without Warren Buffett in his traditional role. All eyes will be on Greg Abel, his tone, his capital allocation signals, and how he frames the road ahead.

It marks the end of an era. And a reminder that even the greatest long-term stories eventually evolve.

What I’m Watching

The most important shift right now isn’t getting enough attention.

Capital is concentrating.

Investors are crowding into dominant platforms, particularly those tied to AI, cloud infrastructure, and scalable ecosystems. These aren’t just companies participating in growth. They are enabling it.

Alongside this, a second shift is emerging: the battle for attention.

Engagement is becoming the most valuable currency in the digital economy, and the companies winning that battle are pulling further ahead. Consider the numbers: the average adult in the U.S. now spends over seven hours per day consuming digital content. That figure hasn’t plateaued; it’s still climbing. The platforms capturing the largest and most loyal share of those hours are compounding their advantage in ways that don’t show up cleanly in quarterly earnings but do show up in long-run pricing power, advertiser dependency, and user switching costs.

This is why the convergence between streaming, social, and short-form video matters so much right now. When Netflix moves into vertical video and Meta doubles down on Reels, they aren’t chasing trends. They are fighting for the same finite resource: your time. And whoever wins that fight consistently across devices, formats, and moments in the day wins the monetisation battle that follows.

This isn’t just a tech trend. It’s a behavioural one, and it’s reshaping how capital gets allocated across the entire market.

Intel Rebound, Real or Temporary?

Intel is showing signs of life.

The company reported better-than-expected earnings for Q1 2026, with revenue rising 7% to $13.6 billion, Data Centre & AI revenue jumping 22% to $5.1 billion, and Foundry revenue growing 16% to $5.4 billion, signalling progress in a segment that matters enormously for the long-term thesis.

The narrative is shifting. Intel is positioning the CPU as a foundational component of the AI era, a counterpoint to the GPU-dominated story that has defined the past two years.

But execution still matters.

Here is what the bull case actually requires: Intel needs to demonstrate that its foundry business can compete for external customers at scale, not just manufacture chips for its own products. That is a fundamentally different business model, with different economics, different relationships, and a different set of risks. The question is whether that ambition survives contact with the timeline.

Competitors like TSMC remain significantly ahead in manufacturing capability. With 14A not expected until 2028, this remains a long-duration turnaround story, not a confirmed comeback. The encouraging results are real. But one good quarter does not close a multi-year execution gap.

Intel Revenue by Segment: Q1 2025 vs. Q1 2026

Netflix vs. Social Media

Netflix is adapting to a changing world of attention.

Its move into vertical, mobile-first video is a direct response to platforms like TikTok and Meta. The goal is to capture short bursts of engagement, the in-between moments that now dominate modern content consumption.

At the same time, Netflix is expanding into creator content, video podcasts, and live events.

The takeaway is clear: streaming and social media are converging. The winner won’t be the platform with the most content. It will be the platform that owns engagement.

Apple’s Next Chapter

Apple is returning to its core identity.

Leadership is signalling a renewed focus on product excellence, the kind of obsession with detail, experience, and craftsmanship that has always defined the company at its best.

Apple’s moat has never been innovation alone. It’s been execution, from Apple Silicon to the tightly integrated hardware and software ecosystem that makes switching genuinely costly.

The next phase of growth will come from deepening that advantage, not from reinventing it.

Amazon and Anthropic, Owning the AI Infrastructure Layer

Amazon is making a massive long-term bet.

Its expanding partnership with Anthropic, including up to $20 billion in investment, positions it at the centre of the AI infrastructure race. Anthropic is scaling rapidly, with a revenue run rate exceeding $30 billion and plans to spend over $100 billion on AI infrastructure over the next decade.

Much of that flows through Amazon’s AWS.

That’s the real story. The investment is significant, but the more durable advantage is structural. Every dollar Anthropic spends on compute, storage, and model training runs through AWS infrastructure. As Anthropic grows, so does the demand it generates for Amazon’s cloud business, independent of whether the AI models themselves succeed commercially. Amazon has engineered itself into the position of landlord for one of the fastest-growing tenants in the technology sector.

This is a playbook Amazon has run before, with third-party sellers on its marketplace, with media studios on Prime Video, with startups on AWS. Build the platform. Set the terms. Collect the toll. What’s different this time is the scale of the underlying opportunity, and the speed at which it is compounding.

This isn’t just an investment; it’s a demand engine. Amazon is doing what it has always done best: build the platform and let others scale on top of it.

Final Thoughts

This market isn’t rewarding everything.

It’s rewarding positioning.

The biggest mistake investors can make right now is assuming broad participation will return quickly. What we’re seeing instead is a tightening group of winners, companies with scale, execution, and structural advantages that are difficult to replicate.

That’s where the edge is.

Not in predicting headlines. But in owning the businesses that don’t depend on them.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.

This week’s market narrative is increasingly defined by divergence. Bank earnings are revealing the true health of the economy, major tech firms continue to double down on AI infrastructure, and streaming platforms like Netflix are quietly evolving into advertising giants.

Underneath the surface, capital is becoming more selective. Investors are rewarding durable cash flows, scalable platforms, and companies with the balance sheets to invest aggressively through uncertainty. As we move deeper into earnings season, the signals coming from banks, big tech, and digital platforms will help shape where the next phase of market leadership emerges.

In this issue:

• Why Netflix’s advertising business could become a multi-billion dollar revenue engine
• What bank earnings are revealing about hidden credit risks in the economy
• And why Amazon’s satellite ambitions may be the beginning of a new connectivity arms race

If you find this issue valuable, consider sharing The Long Term Edge with one investor who thinks long term.

Market Overview (13th to 17th April)

Wall Street posted strong weekly gains, pushing toward new highs as geopolitical tensions temporarily eased and investors rotated back into risk assets.

A developing ceasefire in the Middle East improved market sentiment after weeks of volatility. Iran declared the Strait of Hormuz open for the ceasefire period, though negotiations remain ongoing and the US naval blockade on Iranian ports remains in place. Both sides are expected to return to Pakistan next week for a second round of talks.

The most important macro data point of the week was below-expectation PPI readings, suggesting upstream inflation pressures may be moderating, which removes one of the key arguments for the Fed to hold rates higher for longer.

For the week:

• S&P 500 gained 4.5%
• Nasdaq Composite rose 6.8%
• Dow Jones Industrial Average added 3.2%

Technology and AI-linked companies once again led the rally, reflecting continued investor confidence in long-term infrastructure spending tied to artificial intelligence.

Week Ahead (April 20 – 24)

The coming week brings several earnings releases and macro events that are directly relevant to the themes we track.

Tuesday, April 21

March Retail Sales report. One of the most important consumer data points of the quarter, arriving at a moment when the consumer health debate is unresolved. Capital One also reports, providing an early read from a lender with meaningful lower-income exposure.

Wednesday, April 22

Tesla reports. The earnings call will be closely watched for updates on robotaxi deployment, the Optimus humanoid robot program, and AI infrastructure capital spending. Also: Google Cloud Next begins, where Alphabet is expected to showcase new generative AI tools and enterprise cloud capabilities, a direct read on whether AI investment is translating into enterprise revenue.

Thursday, April 23

Intel reports. The earnings call will provide insight into data-center demand, AI computing infrastructure, and the progress of the US foundry buildout, the bet that went from being ridiculed to being geopolitically essential in the span of two years. Texas Instruments also reports Wednesday, providing a broader read on semiconductor demand outside the AI supply chain.

Friday, April 24

University of Michigan Consumer Sentiment Survey. A secondary indicator, but worth watching given the gap that has opened between sentiment data and actual spending behaviour this year.

The primary focus this week and next is the hyperscaler earnings cycle. Microsoft, Alphabet, Meta, and Amazon will collectively tell us whether the AI capex story is generating real revenue, or whether the market is still pricing a promise rather than a result.

Netflix: The Advertising Business Is the Story

Netflix’s advertising business is on track to generate $3 billion in annual revenue by 2026. That’s the number long-term investors should focus on from this week’s earnings, not the Q2 margin guidance that briefly rattled markets.

The company is in genuine transition. For years the primary metric was subscriber growth. Management is now prioritising revenue per user, pricing power, and advertising monetisation. That’s a fundamentally different and more durable business model.

The evidence is accumulating. Subscription prices were raised earlier this year to $8.99 for the ad-supported plan and $19.99 for the premium tier. Subscriber growth continued. Engagement held. A company that can raise prices without losing customers is demonstrating exactly the kind of pricing power that compounds over a decade.

Netflix is also building its own ad-technology stack rather than relying on third-party infrastructure, a move that will take years to fully monetise but that will ultimately give the company better margins and deeper relationships with global brands than a platform-dependent model would allow.

The Q1 headline numbers: earnings of $1.23 per share on $12.25 billion in revenue. However, results included a $2.8 billion breakup fee from the terminated Warner Bros. Discovery agreement. Excluding that one-time payment, adjusted earnings were closer to $0.58 per share, which better reflects underlying performance. Second-quarter operating margin guidance of 32.6% came in below the prior year’s 34.1%, reflecting heavier content spending scheduled for mid-year.

That margin dip is real but contextual. Netflix projects full-year 2026 revenue between $50.7 billion and $51.7 billion, implying the company still sees significant long-term growth in the global streaming market. International subscriber expansion and operating margin recovery in the second half of the year are the two metrics worth tracking.

One notable development: co-founder Reed Hastings is stepping down as chairman and will not seek reelection to the board. Hastings built a company that navigated the pivot from DVD rental to streaming, created a global content machine, and survived the 2022 subscriber collapse to emerge with a stronger, more profitable business. What he’s leaving behind is a company in genuine transition, and whether the next generation of leadership can execute the advertising pivot is the real long-term question for Netflix investors.

Amazon’s Acquisition of Globalstar: Strategic Logic, Asymmetric Risk

Amazon is expanding its satellite ambitions with a $10.8 billion acquisition of GlobalStar, giving its Project Kuiper network access to satellites, spectrum, and direct-to-device technology; the capability that allows satellites to connect directly to smartphones beyond the reach of cell towers.

The deal also secures a significant new customer: Apple, which will use Amazon’s LEO network to power emergency texting features on iPhones and Apple Watches, extending GlobalStar’s existing partnership with Apple.

The strategic rationale is straightforward. The industry is converging toward a world where the same provider bundles broadband and mobile satellite connectivity. Telcos, cable companies, and now satellite providers are all racing toward the same destination. Amazon felt it needed to match what SpaceX is building with Starlink rather than remain exposed as a broadband-only satellite operator.

However the risk profile of this acquisition is more complex than the market’s 3% positive reaction suggested.

When SpaceX acquired EchoStar spectrum for roughly $20 billion, they bought spectrum that can be redeployed on terrestrial towers and sold back to AT&T, Verizon, or T-Mobile if the satellite market develops more slowly than expected. Their downside is protected.

GlobalStar’s spectrum is far less useful on terrestrial networks. If the direct-to-device market doesn’t develop into a meaningful consumer business, and there are genuine reasons to be cautious here, Amazon has limited ability to recycle the asset. They need this market to work. That’s a fundamentally different risk profile than SpaceX’s position.

The direct-to-device market itself remains unproven at scale. Starlink’s broadband service has 11 million subscribers and roughly $11 billion in revenue, that’s a proven business. The Starlink-T-Mobile direct-to-device partnership, by contrast, has generated less than $100 million over several years. The service only works outdoors. Data speeds are a fraction of what terrestrial networks deliver indoors. And the historical precedent is worth keeping in mind: in the 1990s, Iridium and GlobalStar both projected tens of millions of satellite phone subscribers. They ended up with a few hundred thousand and went bankrupt.

The technology is better now. The devices are smarter. The service is being embedded rather than requiring a separate purchase. But the fundamental user behaviour question, whether people will pay a monthly fee for a service they rarely need — has not yet been answered at scale.

None of this makes the Amazon acquisition wrong. Project Kuiper already has more than 200 satellites in orbit and regulatory approval to launch more than 7,000 broadband satellites, the broadband side of the business has a credible path. The Apple partnership provides real near-term revenue. And being present in a converging industry is strategically rational even when the outcome is uncertain.

But investors buying Amazon specifically on the back of this deal should understand that this is a bet with limited downside protection — not a repositioning with an exit option attached.

Semiconductor Earnings: The Scale of the Buildout

Recent earnings from semiconductor companies continue to reinforce the scale of global investment in AI infrastructure.

ASML raised its long-term outlook after posting €2.76 billion in profit on €8.77 billion in revenue, and now expects 2026 revenue between €36 billion and €40 billion, reflecting sustained demand for the advanced lithography machines used to manufacture cutting-edge AI chips. There is effectively one company in the world that can produce these machines. ASML is it.

Broadcom extended its partnership with Meta Platforms to supply custom AI chips through 2029, a multi-year commitment that reflects how deeply the hyperscalers are embedding custom silicon into their infrastructure strategies.

Bank of America estimates the largest hyperscale cloud companies spent $166 billion on AI infrastructure in Q1 alone, putting total spending on track to reach $750 billion this year and potentially $902 billion by 2029.

To put that in context: the entire US interstate highway system cost roughly $500 billion in today’s dollars and took decades to build. The AI infrastructure buildout is happening at a pace that has no real historical precedent.

The question for long-term investors is not whether the spending cycle is real, it clearly is. It is which companies in the supply chain have the structural position to capture the most durable portion of it. The coming hyperscaler earnings calls will be the most important signal yet on whether that investment is beginning to generate the revenue growth that justifies it.

Bank Earnings: Credit Is Still Fine, For Now

The headline from bank earnings was strong trading revenues. Equities trading rose roughly 27% on average across the six major banks, fuelled by war-driven volatility. Goldman, JPMorgan, Wells Fargo, Citigroup, Bank of America, and Morgan Stanley all beat expectations.

That’s the story most people covered. The more important story is in the credit data.

Non-accruing loans, those more than 90 days delinquent, where the bank stops booking interest income and starts reserving for losses, remain near long-term averages. JPMorgan reported non-accruals up 11% year-over-year but down 3% sequentially. Bank of America saw non-accruals decline 4% year-over-year and flat sequentially. These are not the numbers of a system under stress.

Banks also disclosed their exposure to private credit markets this quarter. JPMorgan has $50 billion in exposure, Wells Fargo $36 billion, and Citigroup $22 billion. In the context of the overall size of these institutions these numbers are manageable, and many of the loans carry significant subordination protection underneath.

On individual bank results: JPMorgan had the strongest overall quarter with a 23% return on tangible common equity and fixed income trading up 21% year-over-year. Morgan Stanley delivered what I’d call a best-in-class quarter, 27% return on tangible common equity, the highest in the industry, with strength across trading, advisory, and wealth management. Goldman beat on earnings per share but saw fixed income trading disappoint, down 10% year-over-year while every other major bank reported strength in that category. Wells Fargo had a difficult quarter, missing on net interest income with a 13 basis point sequential decline in net interest margin, a larger move than it looks for a lending-centric bank where net interest margin is the primary earnings driver. Citigroup delivered solid operating leverage with 14% revenue growth versus 7% expense growth. Bank of America was solid and notably improved its return on tangible common equity from around 14% to 16%, a meaningful step for a bank that has been stuck at the lower end of the industry return profile for several years.

We have now had 17 years of benign credit since the financial crisis. Multiple credit cycles have been called. None have materialised at scale. The banks have the broadest and most granular view of the economy. When their credit data is benign, that matters more than any sentiment survey or economist forecast.

The signal to watch over the next two weeks is consumer finance company earnings, lenders with deeper exposure to lower-income households than JPMorgan or Bank of America. That’s where any broadening of stress would show up first, and it’s the clearest test of whether current delinquency trends are contained or spreading.

What I am Watching

With bank earnings largely reported and credit data holding benign, the next meaningful read on consumer health comes from consumer finance companies reporting over the next two weeks. Unlike JPMorgan and Bank of America, whose credit card books skew affluent, lenders with deeper exposure to lower-income households will tell us whether the stress showing up in auto and credit card delinquency data is contained or broadening. That’s the data point most worth waiting for before drawing firm conclusions about the consumer.

On the labor market, the April jobs report lands in early May and carries more weight than usual. Job losses have appeared in six alternate months since June of last year. If April follows that pattern, the plateau in payroll employment becomes harder to dismiss as noise. Watch also for any revision to the March number, which came in at 178,000 but fits a trend of initial prints being revised lower.

In AI and semiconductors, the hyperscaler earnings season is the main event. Microsoft, Alphabet, Meta, and Amazon all report this week and next. The market is looking for two things: confirmation that capex commitments are holding and early evidence that AI-driven revenue is beginning to materialise at scale. Any softening in forward capex guidance, even modest, will hit semiconductor stocks hard given how much of the bull case rests on sustained infrastructure spend. Conversely, strong cloud growth numbers from AWS or Azure would reinforce the case that the compute investment is already generating returns.

On private credit, the benign bank credit data this week provided temporary relief to the sector. But the underlying question, whether software-exposed private credit portfolios are facing structural impairment rather than cyclical stress, won’t be answered by bank earnings alone. Watch for disclosures from the larger business development companies and private credit managers reporting over the next few weeks. That’s where the real exposure sits.

Closing Thoughts

The divergence that opened this issue is the theme that will define the next phase of earnings season.

Capital is becoming more selective, rewarding businesses with durable cash flows, scalable platforms, and the balance sheets to invest through uncertainty. Netflix is demonstrating that with its advertising pivot. The banks are demonstrating that with credit quality that has held despite 17 years of predicted crises. And the semiconductor supply chain is demonstrating it with earnings that reflect a spending cycle unlike anything the industry has seen before.

The risks are real. Geopolitical uncertainty hasn’t resolved. The direct-to-device satellite market is unproven. The SaaS sector faces structural pressure that won’t be fixed by one good quarter. And the AI capex cycle eventually needs to show revenue that justifies the scale of investment.

But long-term investors who stay focused on structural trends rather than short-term volatility are in the right place. The businesses building genuine competitive advantages today, in AI infrastructure, digital advertising, satellite connectivity, and financial services, are the ones most likely to reward patience over the next seven years.

Clarity compounds. Stay long-term.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Always do your own research or consult a licensed advisor.