ASTERA LABS, INC. (NASDAQ: ALAB)

Current Price: ~$289 | Market Cap: ~$$50.22B | Shares Outstanding: ~173.49 million

Table of Contents

1. Company Overview

2. Financial Performance — Most Recent Results

3. Forward Guidance

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

17. Entry Strategy

Company Overview

Astera Labs designs semiconductor connectivity products that sit between AI accelerators and the wider data center infrastructure.

Historically, the company has focused on PCIe retimers and CXL memory connectivity controllers. These products help GPUs, CPUs, memory and other components communicate efficiently inside increasingly complex AI systems.

Astera’s next phase of growth is more ambitious. The company is moving up the infrastructure stack into higher value and higher ASP products, particularly fabric switches through its Scorpio X Series.

This shift is strategically important.

If successful, Astera could evolve from a component supplier into an important connectivity and switching platform for both scale up and scale out AI clusters. That would materially expand its addressable market and potentially support a much higher long term earnings profile.

The key question is whether Astera can make this transition quickly enough to justify its current valuation.

That question has become more important because the largest GPU vendors are increasingly pursuing vertically integrated networking and interconnect solutions of their own. Nvidia’s proprietary NVLink strategy and broader platform integration create a direct challenge to the assumption that third party connectivity vendors will continue to capture the highest value layers of AI infrastructure.

Financial Performance — Most Recent Results

Q2 was an exceptionally strong quarter.

Revenue reached $392.4M, up 104% year over year from $191.9M. Non GAAP EPS came in at $0.80, beating the $0.69 consensus estimate by approximately 15.9%.

GAAP net income reached $153.1M, compared with $51.2M a year earlier. That represents approximately 199% year over year growth.

Importantly, this is not a situation where headline earnings need to be heavily adjusted because of a large one time gain, impairment or other accounting distortion.

The underlying operating performance is genuinely strong.

The combination of 104% revenue growth, expanding profitability and a 15.9% EPS beat shows that demand for Astera’s products remains exceptionally strong.

Forward Guidance

Management expects Q3 2026 revenue of $540M to $560M.

At the midpoint, that represents approximately 40% sequential growth from Q2.

The primary driver is the ramp of the Scorpio X Series switching platform.

Management also expects:

Non GAAP gross margin: approximately 72%

Non GAAP EPS: $1.16 to $1.21

The EPS guidance is particularly notable because it was well above the approximately $0.81 consensus estimate that existed before the results.

However, there is an important data quality issue with some third party FY2027 and FY2028 consensus estimates.

One source cited FY2027 consensus EPS of $2.71, described as a 47.3% increase over FY2026. That implies a FY2026 EPS base of approximately $1.84.

That number does not make sense against Astera’s actual results.

Q1 EPS was approximately $0.62.

Q2 EPS was $0.80.

Q3 guidance is $1.16 to $1.21.

Those three quarters alone imply approximately $2.58 to $2.63 of EPS, already well above the implied FY2026 consensus base of $1.84.

Because of this inconsistency, this analysis does not use the third party consensus figure as its valuation anchor.

Instead, it builds a bottom up estimate.

Assuming Q4 sequential growth slows to approximately 20% to 25% as the Scorpio ramp matures, FY2026 non GAAP EPS is estimated at approximately $4.05.

That is a self derived estimate rather than a sourced consensus number.

For FY2027, assuming growth moderates further to approximately 45% year over year as AI infrastructure spending normalizes and competition increases, FY2027 EPS is estimated at approximately $5.87.

This $5.87 estimate is therefore the primary earnings anchor used throughout the valuation analysis.

Balance Sheet & Capital Allocation

Astera has a very strong balance sheet.

The company holds approximately $1.18B of cash against only $41.85M of debt.

That leaves approximately $1.14B of net cash, equivalent to roughly $6.67 per share.

This is an important strength.

Astera is aggressively investing in its next generation product portfolio, but it does not need to rely heavily on external financing to fund that expansion.

The balance sheet therefore provides substantial flexibility as the company moves from PCIe connectivity products toward higher value switching infrastructure.

No material share buyback or dividend program was identified in the sources reviewed.

That is consistent with Astera’s position as a high growth company that is still prioritizing reinvestment and financial flexibility over returning capital to shareholders.

Recent Developments Since the Prior Quarter

August 4, 2026

Astera reported Q2 2026 results.

Revenue reached $392.4M, up 104% year over year. Non GAAP EPS was $0.80, beating expectations by 15.9%. GAAP net income reached $153.1M, up 199% year over year.

Management also issued strong Q3 guidance, driven primarily by the Scorpio X Series ramp.

2026

Nvidia announced a major partnership with Intel to integrate Nvidia’s proprietary NVLink networking technology with Intel’s x86 data center CPUs.

This is strategically important for Astera because NVLink represents a proprietary alternative interconnect architecture controlled by the dominant GPU vendor.

The development raises a direct question about how much of the future AI connectivity market will remain available to independent PCIe and switching suppliers.

2026

AMD and Oracle announced plans to deploy 50,000 AMD GPUs into Oracle data centers beginning in the second half of 2026.

This creates another question around the long term addressable market for Astera’s PCIe based connectivity products, particularly if large accelerator deployments increasingly adopt platform specific interconnect architectures.

Ongoing through 2026

Astera’s stock has fallen from its 52 week and all time high of $499.48 to approximately $289.47.

That represents a drawdown of approximately 42%.

The decline has occurred alongside the competitive developments above, a broader semiconductor sector pullback, reported institutional selling and continued concerns about Astera’s valuation.

2026

Sell side sentiment remains relatively constructive.

Bank of America issued a Street high price target of $460, while broader analyst targets reviewed in this analysis range from approximately $333 to $388.77.

This creates an important divergence between analyst expectations and actual share price performance.

Key Risk Factors

1. Nvidia’s NVLink strategy could challenge Astera’s addressable market

The most important risk is not generic semiconductor competition.

It is the possibility that Nvidia increasingly controls the connectivity layer of AI infrastructure through proprietary technologies such as NVLink.

The Nvidia and Intel integration is therefore strategically significant.

If hyperscalers increasingly adopt Nvidia controlled interconnect architectures, the opportunity for independent connectivity vendors could shrink precisely in the highest value portion of the market Astera is targeting with Scorpio.

2. AMD and Oracle could accelerate platform specific architectures

The AMD and Oracle 50,000 GPU deployment creates a second competitive signal.

Large scale AI infrastructure could increasingly be built around tightly integrated accelerator, networking and data center architectures.

If that happens, third party PCIe retimers and switches could capture a smaller portion of the total infrastructure spend.

3. Valuation remains demanding

At approximately $289.47, Astera trades at roughly 49x this analysis’s FY2027 EPS estimate of $5.87.

That is still a very demanding valuation.

The 42% decline from the highs has therefore not automatically made the stock inexpensive.

4. Third party earnings estimates are inconsistent

The FY2027 consensus EPS data reviewed contains material inconsistencies.

That makes valuation particularly sensitive to which earnings estimate is used.

For this reason, this analysis relies on its own bottom up earnings build rather than blindly applying a reported consensus number.

5. Extreme share price volatility

Astera’s 52 week trading range spans approximately $97.89 to $499.48.

That represents more than a five times difference between the low and high.

The magnitude of the movement demonstrates how sensitive the stock is to changes in expectations around AI infrastructure growth, competition and valuation.

DECISION TREE

Level 1 — Macro & Cost of Capital Filter

Read more

CONSTELLATION ENERGY CORPORATION (NASDAQ: CEG)

Current Price: ~$279.52 | Market Cap: ~$99B | Shares Outstanding: ~354.3M

Table of Contents

1. Company Overview

2. Financial Performance — Most Recent Results

3. Forward Guidance

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

17. Entry Strategy

    ---

Company Overview

Constellation Energy is the largest operator of nuclear power generation in the United States. Its nuclear fleet achieved a 93% capacity factor in the latest quarter.

The company expanded significantly through its January 7, 2026 acquisition of Calpine for $21.835B in cash and stock. The deal added natural gas and renewable generation, increased scale, and strengthened Constellation’s ability to provide firm power across different generation sources.

The strategy is increasingly focused on becoming a preferred supplier of firm, clean, 24/7 power to AI data centers and other large corporate customers through long term power purchase agreements.

The most prominent example is the 20 year, 835 MW Microsoft PPA supporting the restart of the former Three Mile Island Unit 1, now renamed the Crane Clean Energy Center.

Financial Performance — Most Recent Results

Q2 2026 was strong.

Revenue increased 23% YoY to $7.504B, while adjusted EPS reached $2.55, $0.64 higher than the prior year.

The gap between GAAP EPS of $1.42 and adjusted EPS of $2.55 is substantial, but it is disclosed and largely reflects acquisition related accounting and mark to market effects from Calpine. These include purchase accounting amortization, integration costs, and hedging volatility.

For this analysis, adjusted EPS is the more useful valuation measure because it is the metric management uses to guide investors and the Street uses to model the business.

Importantly, the higher adjusted EPS was not simply an accounting benefit. It reflected operational contributions from Calpine, stronger PJM capacity pricing, and commercial gains.

Forward Guidance

Following the Q2 beat, management raised FY2026 adjusted operating EPS guidance to $11.50 to $12.50, from $11.00 to $12.00 previously.

At the midpoint, this implies approximately 28% EPS growth versus FY2025 adjusted EPS of $9.39.

Management also signed approximately 920 MW of new long term nuclear contracts during Q2, with an average duration of 18.5 years.

This is important because it provides tangible evidence that demand for firm, clean power is translating into long duration contracted revenue rather than remaining an AI data center narrative.

Street estimates imply approximately 26% additional EPS growth in FY2027, with normalized EPS around $15.12.

The market is therefore assuming that Constellation can sustain high double digit earnings growth for a second consecutive year.

Balance Sheet & Capital Allocation

The Calpine acquisition materially changed Constellation’s balance sheet.

Long term debt increased from $7.250B at year end 2025 to $19.111B as of June 30, 2026.

Total assets increased to $98.253B, including $11.107B of goodwill. Total liabilities increased to $65.931B, compared with $42.396B at year end 2025.

Shareholders’ equity increased to $31.977B, partly reflecting approximately 50 million new shares issued as stock consideration, valued at $17.603B.

The transaction also included $4.342B of cash and $12.551B of assumed Calpine debt.

Constellation pays a quarterly dividend of $0.43 per share, up 10% YoY. However, the roughly 0.5% dividend yield is well below the utility sector average of approximately 3.6%.

That reflects the market’s positioning of Constellation as a growth utility rather than an income stock.

Capital allocation is currently focused on the Calpine transaction, organic nuclear investment, and long term PPA growth rather than aggressive share buybacks.

The key point is simple: the balance sheet is materially more leveraged than before Calpine. The additional debt is a real cost of pursuing the growth strategy and needs to be considered alongside the earnings opportunity.

Recent Developments Since the Prior Quarter

January 7, 2026
Completed the $21.835B Calpine acquisition, using cash and stock. The transaction added natural gas and renewable generation while materially increasing scale, debt, and share count.

Ongoing through 2026
Constellation is targeting a 2027 restart of the Crane Clean Energy Center, supported by the 20 year Microsoft PPA. The company accelerated the expected restart timeline by roughly one year.

Mid 2026
PJM disclosed that transmission upgrades needed for full grid deliverability from Crane, including 765 kV and 500 kV projects, are not expected to be completed until December 2030. A Constellation executive has acknowledged that full reconnection could potentially extend into 2031.

This creates a significant timing gap between the targeted 2027 physical restart and the point at which the grid may be capable of fully delivering the electricity.

August 6, 2026
Q2 results showed revenue growth of 23% YoY, adjusted EPS of $2.55, and higher FY2026 EPS guidance of $11.50 to $12.50. Management also highlighted approximately 920 MW of new long duration nuclear contracts signed during the quarter.

August 2026
JPMorgan lowered its price target while maintaining an Overweight rating, reflecting a more balanced risk and reward outlook following the stock’s earlier rally.

The broader utility sector has also faced pressure from higher bond yields and softer electricity pricing sentiment. This has affected peers including NextEra, Duke, and Southern, suggesting that not all of Constellation’s decline is company specific.

Key Risk Factors

1. Crane grid deliverability risk

This is the most important unresolved issue.

Constellation is targeting a 2027 physical restart, while PJM’s disclosed transmission timeline points toward December 2030 at the earliest, with full deliverability potentially extending into 2031.

The key question is whether the Microsoft PPA can deliver the expected firm power on the original timetable. If not, the economics or timing of the agreement could eventually require adjustment.

2. PJM capacity market price caps

FERC approved capacity price caps of approximately $325 to $333 per MW day, materially below the estimated $530+ scarcity price that an uncapped market could potentially produce.

Constellation has a large cleared capacity position of 18,875 MW for the 2028/2029 auction, but regulatory price ceilings limit how much of the underlying scarcity value it can capture.

3. Higher leverage following Calpine

Long term debt increased from $7.25B to $19.111B in roughly six months.

The acquisition provides greater scale and earnings potential, but it also introduces substantially higher financial risk and creates additional integration requirements.

4. Interest rate and utility valuation risk

Constellation has fallen more than 51% from its 52 week high, alongside broader weakness across utilities.

Higher bond yields can pressure utility valuations because investors have more attractive alternatives in fixed income and apply higher discount rates to future cash flows.

Therefore, some of the stock’s decline could persist even if the company executes well.

5. Regulatory risk around large loads

PJM’s proposed rules covering co located generation and large load interconnection, including Reliability Backstop Procurement and Interim Resource Adequacy Service, remain under FERC review.

The outcome will influence how Constellation structures and prices future AI data center power agreements.

DECISION TREE

Level 1 — Macro & Cost of Capital Filter

Read more

What Actually Happened

Nvidia Finally Broke the “Sell the Good News” Pattern

Nvidia finally delivered the clean beat and hold that this memo has been building toward for a month.

Fiscal Q2 revenue came in at $96.2 billion, up 106% year over year, with EPS of $2.22. The stock jumped 8.7% the following session, while October quarter guidance pointed to another 89% year over year growth.

More importantly, the reaction spread across the entire AI complex.

Broadcom rose 4.5%, SK Hynix gained 2%, Intel gained 4%, and the Nasdaq rose 1.57% that day.

This broke the pattern that had dominated August, where strong results and positive announcements were repeatedly sold.

The broader week was still volatile. Nvidia entered earnings after a seven session losing streak, its longest since October 2022. Ahead of the print, Intel was down 5%, AMD 4%, and TSMC 3% on the Monday.

The strong Nvidia reaction was enough to pull the week into positive territory:

• S&P 500: +0.5%
• Nasdaq: +0.9%
• Dow: +0.5%

That made it the market’s first winning week in three, despite a weaker Friday where the S&P fell 0.25% and the Nasdaq fell 0.52%.

Nvidia Opens a New AI Financing Channel

Nvidia also announced a major financing initiative with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR.

The group plans to mobilize more than $500 billion of third party capital for AI infrastructure.

The important point is the structure.

The financing is designed to remain off Nvidia’s balance sheet, with Nvidia compute capacity itself being used as collateral for debt issued through special purpose vehicles.

This potentially creates a new way to fund the enormous capital requirements of the AI buildout without relying entirely on hyperscaler balance sheets.

Memory Costs Are Starting to Show Up in Margins

Nvidia’s October quarter gross margin guidance came in at 74%, described as slightly below expectations.

Management partly attributed the pressure to early memory cost inflation.

This is becoming a broader theme.

Amazon had already flagged higher memory costs as a contributor to its rising capex requirements three weeks ago. Nvidia’s latest guidance provides a second data point suggesting that rising DRAM and HBM costs are beginning to flow through the economics of the AI infrastructure chain.

Weekend reports also suggested that some Nvidia customers could face price increases of up to 15%.

The Fed Reintroduces the Rates Risk

New Fed Chair Kevin Warsh delivered his first Jackson Hole speech on Friday.

His message was notably hawkish. He described underlying inflation trends as “concerning” and said the Fed “has more work to do”, while declining to provide forward guidance.

Markets interpreted the comments as hawkish.

September rate hike odds moved above 50% for the first time this cycle.

The 30 year Treasury yield remained around 5.20% to 5.21% at the end of the week.

That is stable for now, but Warsh’s tone puts the modest stabilization following the earlier 5.31% peak at risk.

Why It Happened

Nvidia Was the Catalyst the Market Needed

Nvidia’s clean beat and hold, and the broader rally that followed it, provide the clearest evidence yet that August’s “sell the good news” pattern was primarily about crowding and positioning rather than deteriorating fundamentals.

For several weeks, investors had been selling strong earnings and positive announcements.

Nvidia was the biggest and most anticipated test.

When it delivered strong results and the stock rallied decisively, the market finally had confirmation that strong fundamentals could still overcome positioning.

The reaction was also broad.

Unlike the earlier AMD and SK Hynix earnings reactions, the relief was not limited to one company. It spread across the semiconductor and AI infrastructure complex.

AI Financing Is Evolving

The Nvidia financing platform is also a direct response to the capex concerns tracked throughout August.

The market has increasingly questioned:

How much can hyperscalers spend?

How much debt can they carry?

How quickly will massive AI investments generate returns?

Oracle’s backlog versus cash flow gap and Meta’s rising capex requirements highlighted the problem.

The new financing structure suggests the industry is beginning to find another answer.

Instead of putting all the debt on hyperscaler balance sheets, outside institutional capital can participate directly in financing AI infrastructure.

That validates continued investor appetite for the buildout, while also pushing more leverage into private credit and insurance capital markets.

The most interesting element is the collateral.

Compute itself is becoming part of the financing structure.

That is potentially significant, but it also introduces a new source of systemic risk.

Rates Risk Has Become More Direct

Warsh’s Jackson Hole comments bring the rates risk tracked throughout this memo back into focus.

Previously, the concern was largely coming from bond market movements.

Now there is a clear policy signal from the Fed Chair himself.

Because Warsh has declined to provide forward guidance, every future inflation, employment and growth data point becomes more important.

That increases uncertainty for the entire AI complex.

What Changed

1. The “Sell the Good News” Pattern Finally Broke

Nvidia’s earnings provided the confirmation this memo has been waiting for since early August.

Strong news can still drive the complex higher when the fundamental result is strong enough.

Crowding risk has not disappeared, but it is no longer overwhelming fundamentals.

2. Memory Inflation Is Becoming a Real Margin Issue

Amazon’s capex commentary and Nvidia’s margin guidance now provide two separate data points.

Rising DRAM and HBM costs need to be incorporated into models for hyperscalers, chipmakers and other AI infrastructure companies.

This is no longer just a theoretical risk.

3. AI Financing Is Entering a New Phase

The financing model is moving beyond hyperscaler and neocloud balance sheets.

A purpose built financing platform involving six major institutions could mobilize more than $500 billion of third party capital.

Using compute as collateral creates a new funding channel, but also a new potential channel for systemic risk.

4. Rates Risk Has a Face and a Name

The rates risk is no longer simply about Treasury market volatility.

Fed Chair Kevin Warsh has now explicitly highlighted inflation concerns and suggested the Fed still has work to do.

His refusal to provide forward guidance means the market faces greater uncertainty heading into September.

What Matters Next (2–4 Weeks)

The September FOMC Is Now the Main Macro Catalyst

The September FOMC meeting is now the most important near term event for the AI complex.

Rate hike odds have moved above 50%, while the 30 year Treasury yield remains around 5.20% to 5.21%.

If long yields continue rising, they could become a meaningful headwind even after Nvidia’s strong earnings result.

Watch for Nvidia Price Increases

Reports suggest some Nvidia customers could face price increases of up to 15%.

The key question is whether customers can absorb those increases.

If Nvidia successfully passes higher memory costs through to customers, that would support the broader memory inflation thesis.

If customers resist, margins across the buying side of the AI infrastructure chain could come under pressure.

Watch the $500 Billion Financing Platform

The next question is whether Nvidia’s financing announcement moves from MOUs to actual capital deployment.

Pay particular attention to:

• Terms

• Pricing

• Investor participation

• Take up rate

If institutional capital actually begins flowing, the initiative could materially change the debate around how the AI infrastructure buildout is funded.

If the agreements remain largely aspirational, the near term impact will be much smaller.

Test Whether “Sell the Good News” Is Really Over

Nvidia has broken the pattern, but it needs to remain broken.

The next earnings reports from AMD, other compute companies and hyperscalers will provide the test.

If strong results continue to produce positive reactions, August’s selling can increasingly be viewed as a positioning reset.

If strong results are sold again, Nvidia may have been a one off exception.

What the Market May Be Mispricing

The Relief Rally May Be More Fragile Than It Looks

The rally rests on one major clean print after a month of positioning driven selling.

A hawkish September Fed decision could quickly reopen the selling dynamic across the complex.

The Financing Initiative May Be More Important Than It Looks

The market may be discounting the Nvidia financing announcement because the agreements are still primarily at the MOU stage.

But if even a fraction of the $500 billion mobilizes on schedule, it could materially change the capex funding debate that has weighed on hyperscalers and neoclouds.

What To Do About It

Rebuild Compute Exposure, But Keep the Rates Hedge

The defensive posture from last week can now be partially reversed.

Nvidia has delivered the clean beat and hold that the market was waiting for, so it makes sense to rebuild selective exposure across:

• Nvidia

• AMD

• Broadcom

• Memory names

The key is position sizing.

The fundamental catalyst has turned positive, but the rates risk remains very much alive following Warsh’s Jackson Hole comments and the greater than 50% September hike odds.

Treat AI Financing as Both Positive and Risk

The Nvidia financing platform is structurally positive for AI infrastructure because it creates another source of capital for the buildout.

But compute backed financing is also a new potential source of systemic risk.

The opportunity and the risk should therefore be tracked together.

Keep Memory Costs in Every Model

Memory inflation is now a confirmed recurring theme rather than a one off data point.

Rising DRAM and HBM costs could affect:

• Nvidia margins

• Hyperscaler capex

• AI infrastructure economics

• Customer pricing

This should remain a core assumption going into September.

Shift Hedges Toward Rates

The compute specific catalyst has now resolved positively.

The remaining major risk is increasingly macro.

That means downside protection should shift toward rate hike risk around the September FOMC, rather than focusing primarily on another fundamental disappointment from Nvidia.

Portfolio Implications & Action Plan

Final Thoughts

Nvidia’s clean beat and hold finally broke the “sell the good news” pattern that dominated August and lifted the broader AI complex.

But two risks have simultaneously become more important.

Memory costs are rising.

Rates are becoming more hawkish.

The next phase should therefore be selective rather than indiscriminate.

Rebuild exposure to Nvidia, AMD, Broadcom and memory, while maintaining protection against a September Fed surprise.

At the same time, watch Nvidia’s $500 billion financing platform closely. If those commitments convert into real capital deployment, it could become one of the most important structural developments in AI infrastructure financing this year.

The fundamental AI story remains intact.

The question now is whether capital costs and financing structures can keep pace with the scale of the buildout.

Disclaimer

This report is provided solely for informational and educational purposes only. It represents the personal research notes, analysis, and observations of the author and does not constitute investment advice, financial advice, a recommendation to buy, sell, or hold any securities, or an offer to solicit any transaction in any securities or financial instruments.

I have a friend who is a developer. He barely uses AI in his daily life. No ChatGPT, Claude or Gemini.

Another friend is a senior developer. His AI usage is mostly limited to vibe coding the occasional side project and a weather forecast app.

A third senior developer uses AI mainly to summarise a few news stories every morning.

These are smart, technically capable people. They understand what AI can do.

Yet, in their personal lives, AI is still optional.

That made me wonder: if even developers are not deeply integrating AI into their daily lives, what does that say about the broader consumer market?

The Gap Between AI Hype and Daily Reality

Most of us have experienced something similar.

We tried ChatGPT or another AI tool. At first, it felt remarkable.

Then the novelty wore off.

Eventually, AI became useful for a few things:

Summarising information.
Rewriting emails.
Planning meals or trips.
Researching products.
Generating ideas.

Useful? Absolutely.

Essential? For most consumers, probably not.

We expected AI to transform everyday life in the same way smartphones did. Instead, for many people, it has settled into a collection of useful but occasional tools.

That distinction matters.

Because it suggests that the biggest long term economic opportunity may not be consumer subscriptions.

It may be enterprise.

In this piece, I want to explore why consumer AI may have a natural ceiling, why enterprise adoption has a much stronger economic model, and where the next phase of AI growth is likely to come from.

1. Consumer AI Has a Natural Ceiling

The Problem: Most Personal AI Tasks Are Occasional

Look at how people actually use AI outside work.

You might ask it about the weather.

Summarise some news.

Rewrite an email.

Suggest a recipe.

Plan a holiday.

Research a product.

These are useful tasks, but they are usually sporadic rather than continuous.

They do not become a daily operating system in the way that email, messaging, calendars or search engines did.

The difference is simple.

Those products solved problems that appeared constantly throughout the day.

AI often solves problems that appear only occasionally.

Useful Does Not Always Mean Worth Paying For

This creates a monetisation problem.

For many consumers, the free version is already good enough. I have seen people who use three to four AI applications for free.

If you only use AI a few times a week, paying $20 per month for marginally better performance can feel unnecessary.

The initial excitement fades.

The question changes from:

“This is amazing. What can it do?”

to:

“Do I actually need this?”

For a large proportion of consumers, the answer may be:

Not really.

This is not a failure of AI technology.

It is simply human behaviour.

Most people do not have enough high value, repetitive personal tasks to justify constant AI assistance.

When the pain is occasional, willingness to change habits and willingness to pay both remain limited.

2. Enterprise AI Works Because the Economics Are Different

Inside a Company, AI Is Not Just a Convenience

The dynamic changes completely once AI enters the workplace.

Companies are not asking:

“Is this interesting?”

They are asking:

“Does this save money, generate revenue or improve productivity?”

That creates a fundamentally different adoption model.

AI can be used for:

• Faster software development

• Automated customer service

• Research and knowledge synthesis

• Document processing

• Sales and marketing workflows

• Process automation

• Agentic workflows that complete multi step tasks

The key difference is repetition.

A consumer may summarise five articles a week.

A company may process thousands of documents every day.

A consumer may occasionally rewrite an email.

A customer service organisation may handle millions of interactions every year.

The economics compound quickly.

ROI Creates a Stronger Reason to Adopt

Companies do not need every employee to love AI.

They simply need the numbers to work.

If AI can reduce the time required to complete a task, increase output or reduce headcount growth, the return on investment becomes measurable.

Once that happens, adoption can scale rapidly.

• More employees.

• More API usage.

• More workflows.

• More agents.

• More infrastructure.

That is very different from trying to convince millions of individual consumers to pay a monthly subscription for occasional convenience.

Enterprise AI becomes embedded in the operating model.

And once a workflow becomes dependent on AI, switching costs and usage intensity can increase significantly.

3. Enterprise Usage Compounds. Consumer Usage Often Does Not.

Consumer AI Is Usually Episodic

Most consumer usage follows a relatively simple pattern.

• A question.

• An answer.

• Task completed.

• The interaction ends.

Even regular users may only generate a limited amount of activity every day.

The economic ceiling is therefore naturally constrained by:

1. How often consumers need help

2. How much they are willing to pay

3. How easily they can switch between free alternatives

There is also a ceiling on how much AI assistance most people actually want in their personal lives.

Not every task needs an AI agent.

Sometimes it is simply faster to do the task yourself.

Enterprise AI Can Expand Within the Same Customer

Enterprise usage works differently.

A company might start with one use case.

Perhaps coding assistance.

Then expand into customer support.

Then research.

Then internal knowledge management.

Then autonomous workflows.

Then company wide agents.

One customer can therefore grow significantly over time without the company needing to acquire another customer.

Usage compounds within the organisation.

This is particularly important for AI companies because the revenue model increasingly extends beyond simple software seats.

1. There are APIs.

2. Inference.

3. Training.

4. Cloud infrastructure.

5. Data centres.

6. Specialised hardware.

7. Agentic workloads.

The more work AI performs, the more computing infrastructure it requires.

That creates a much deeper economic stack than a consumer paying $20 per month for a chatbot.

4. Consumer AI Was the On Ramp, Not the Final Destination

Consumers Built Awareness

Consumer AI was still incredibly important.

ChatGPT and similar products did something remarkable.

They made AI accessible to everyone.

Millions of people experimented with the technology without needing an IT department, procurement process or enterprise software implementation.

That created awareness at a scale that traditional enterprise software could never have achieved.

Many employees discovered useful applications themselves.

Then they started bringing those tools into the workplace.

This created a powerful bottom up adoption model.

Consumer experimentation became enterprise demand.

But the Larger Economics Sit Inside Companies

The long term opportunity, however, increasingly appears to be enterprise.

High value workflows are more sticky.

ROI can justify significant spending.

One deployment can expand across thousands of employees.

AI can become embedded inside mission critical processes.

And as AI moves from answering questions to actually performing work, the amount of computing required per customer can rise dramatically.

That is where the broader AI ecosystem benefits.

The model providers.

The cloud companies.

The semiconductor companies.

The networking infrastructure.

The data centres.

The power suppliers.

Consumer AI helped create the market.

Enterprise AI is where the economic engine becomes much larger.

5. The Real Divide: Convenience vs Productivity

The difference can be summarised simply.

Consumer AI improves convenience

It helps you:

• Write faster

• Search faster

• Summarise information

• Generate ideas

• Plan things

These are valuable improvements.

But for most people, they are not essential enough to create unlimited usage or unlimited willingness to pay.

Enterprise AI changes the economics of work

It can help a company:

• Produce more with the same workforce

• Automate repetitive tasks

• Reduce operating costs

• Increase the speed of decision making

• Improve customer service

• Run workflows continuously

That creates a much stronger economic incentive.

A consumer might ask:

“Is this worth $20 a month?”

A company might ask:

“If this saves us $20 million, how quickly can we deploy it?”

That is a completely different growth curve.

The Bigger Investment Implication

This distinction also matters for investors.

The strongest long term opportunities may not necessarily be the companies with the largest consumer user bases.

The real value could increasingly concentrate around the companies enabling enterprise AI adoption.

That includes:

AI Infrastructure

The chips, networking, cloud and data centre infrastructure required to train and run increasingly complex AI workloads.

Enterprise AI Platforms

Companies building AI directly into business workflows where usage can expand across entire organisations.

Industry Specific AI

AI designed for sectors such as healthcare, finance, law, manufacturing and software development, where domain knowledge and workflow integration can create stronger switching costs.

AI Agents

The next major step may be systems that do not simply assist humans but complete increasingly complex tasks with limited supervision.

This is where AI usage could become much more intensive.

The shift from AI as a tool to AI as a worker is economically far more significant.

The Bottom Line

My three developer friends may not represent everyone.

But I do not think they are unusual.

They illustrate something important about consumer AI.

Even people who understand the technology can find it useful without making it central to their daily lives.

That suggests there may be a natural ceiling to consumer usage.

AI does not need to become irrelevant for this to be true.

It can remain extremely useful while still being mostly optional.

Enterprise is different.

The value is measurable.

The usage can compound.

The budgets are larger.

And once AI becomes embedded into workflows, the cost of removing it can become significant.

Consumer AI was the opening act.

It introduced the technology to the world.

But the next phase of growth is likely to come from companies using AI not simply to make life more convenient, but to fundamentally change how work gets done.

That is where the biggest economic opportunity may be.

The consumer AI story created the excitement. The enterprise AI story is where the compounding begins.

Where do you see the biggest gap between consumer AI hype and how people actually use it in daily life or at work?

What Actually Happened

1. Rates became the dominant force

This week reversed the …

Read more

CEREBRAS SYSTEMS INC. (NASDAQ: CBRS)

Current Price: ~$221.85 | Market Cap: ~$52.7B | Shares Outstanding: ~237.56M

Table of Contents

1. Company Overview

2. Financial Performance — Most Recent Results

3. Forward Guidance

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

Company Overview

Cerebras is an AI computing company that builds wafer scale systems designed to compete with traditional GPU based infrastructure.

Its core technology is the Wafer Scale Engine, which powers the CS 3 and the newly launched CS 4. The company sells its systems directly to customers and also operates a cloud service focused on high speed AI inference.

The cloud business is growing faster and is becoming an increasingly important part of the revenue mix. Management is positioning Cerebras primarily around AI inference, where it claims significant speed advantages over conventional GPU systems.

The business is also highly concentrated. Rather than serving a broad enterprise customer base, Cerebras relies heavily on a small number of very large customers, particularly G42 and MBZUAI in Abu Dhabi, and increasingly OpenAI, following the Master Relationship Agreement signed in December 2025.

The investment case therefore rests on two things:

1. Whether Cerebras can establish a durable performance advantage in AI inference

2. Whether it can turn that advantage into a much broader customer base

   

Financial Performance — Most Recent Results

Revenue growth is strong

The headline number is compelling. Q2 2026 core revenue reached $209.9M, up 103% year over year.

Cloud and inference revenue was even stronger, growing approximately 281% to 287% year over year, making it the fastest growing part of the business.

Core gross margin also improved significantly to 40.6%, up approximately 940 basis points year over year.

GAAP revenue is being distorted by the OpenAI agreement

Cerebras has an unusual accounting issue that makes GAAP revenue less representative of underlying demand.

Under the December 2025 Master Relationship Agreement with OpenAI, Cerebras issued OpenAI warrants to purchase Cerebras shares. Under ASC 606, these warrants are treated as consideration paid to a customer and therefore reduce reported revenue as OpenAI purchases computing capacity.

In Q2 2026:

• Core revenue was $209.9M

• GAAP revenue was $180.1M

• The difference was $29.8M

• This included $44.3M of non cash warrant asset amortization

• This was partially offset by $14.5M of pass through revenue removed

Management expects the warrant amortization to continue reducing GAAP revenue every quarter through October 2031.

The Q2 net loss needs context

Cerebras reported a $450.5M GAAP net loss in Q2.

However, this does not represent a comparable $450.5M cash operating loss. The result was heavily affected by the non cash warrant amortization and IPO related stock based compensation.

For valuation purposes, this analysis therefore uses core revenue rather than GAAP revenue, because GAAP revenue is being mechanically reduced by a disclosed accounting schedule.

That does not mean the warrant arrangement should be ignored. Cerebras carried approximately $1.128B of warrant related assets on its balance sheet at June 30, 2026. The accounting treatment is non cash, but the scale of the arrangement is economically important because it shows how much of the OpenAI relationship was structured around equity rather than cash.

Forward Guidance

Management raised its full year 2026 core revenue guidance following the Q2 results.

The full year guidance was raised for the second consecutive time following the IPO.

However, the Q3 guide deserves attention.

Q3 core revenue of $214M to $216M would represent only a modest sequential increase from Q2’s $209.9M. That is a significant slowdown from the much stronger sequential growth seen between Q1 and Q2.

The key question is therefore whether Q2 represented the beginning of a sustained acceleration or simply a particularly strong quarter.

The backlog is enormous, but concentration matters

Cerebras reported approximately $25.4B of remaining performance obligations, up from $24.6B at FY2025 year end.

That is roughly 29 times the midpoint of FY2026 revenue guidance.

On the surface, this suggests extraordinary long term revenue visibility.

However, there is an important caveat.

If much of that $25.4B backlog comes from the same small group of customers already responsible for most of Cerebras’s revenue, then the backlog provides less diversification than the headline number suggests.

The key question is therefore not simply how large is the backlog, but who is behind it.

Balance Sheet & Capital Allocation

Cerebras has a very strong cash position following its IPO.

As of June 30, 2026:

• Cash, cash equivalents and restricted cash: $7.43B

• Investments: $1.18B

• Combined cash and investments: $8.61B

• OpenAI working capital loan: $918.2M

• Approximate net cash: $7.69B

The company also has an undrawn revolving credit facility.

Strong liquidity, but concentrated financing

The balance sheet provides Cerebras with a substantial financial cushion.

However, there is an unusual structural issue: OpenAI is simultaneously a major customer, equity counterparty and lender.

The relationship therefore affects three areas of the business:

1. Revenue

2. Equity economics through the warrants

3. Financing through the $918.2M working capital loan

This creates a level of counterparty concentration that goes beyond ordinary customer concentration.

Cerebras does not pay a dividend and does not currently have a buyback program.

Capital allocation is focused entirely on growth, capacity expansion and infrastructure investment, which is consistent with a company only a few months removed from its IPO.

Recent Developments Since the Prior Quarter

May 14, 2026: IPO

Cerebras priced its IPO at $185 per share, selling 30 million Class A shares and raising approximately $5.55B.

The stock opened at $350 and closed its first day at $311.07, a 68% gain.

June 23, 2026: First public quarterly results

Cerebras reported its first quarterly results as a public company for Q1 2026.

August 12 to 13, 2026: Q2 results

Q2 core revenue reached $209.9M, up 103% year over year and ahead of the approximately $194M analyst estimate.

Management also raised FY2026 core revenue guidance to $880M to $890M.

Despite the beat and raised guidance, the stock fell approximately 15% to 17% in the following sessions.

Mid August 2026: CS 4 launch

Cerebras launched the CS 4, which the company describes as the fastest AI accelerator in the industry.

Despite the positive product announcement, shares fell another approximately 12.7% around the announcement period.

Sell side reaction was mixed. Wedbush and Morgan Stanley raised their targets, while Citi lowered its target.

That split reaction is important because it shows the market is not automatically rewarding positive company news.

Key Risk Factors

1. Extreme customer concentration

Cerebras has one of the most concentrated customer bases in the semiconductor and AI infrastructure universe.

In 2025:

• G42 represented 24% of revenue

• MBZUAI represented 62%

• Together they represented 86% of revenue

• G42 had represented 85% of revenue in 2024

Both are Abu Dhabi based entities and are treated as related parties under ASC 850.

This means a change in even one major relationship could have a material or potentially existential impact on near term revenue.

2. OpenAI creates a new concentration risk

OpenAI is becoming strategically important while also creating a complicated financial relationship.

Cerebras has:

• A major customer relationship with OpenAI

• A $918.2M working capital loan from OpenAI

• Warrants issued to OpenAI

• A GAAP revenue reduction mechanism running through October 2031

OpenAI is therefore not simply another customer. It is increasingly intertwined with Cerebras’s revenue, equity structure and financing.

3. Growth is already beginning to decelerate

Q3 core revenue guidance of $214M to $216M represents only a modest sequential increase from Q2.

That does not invalidate the growth story, but it means investors should not simply extrapolate the 103% Q2 growth rate indefinitely.

4. Extreme share price volatility

Cerebras has already experienced unusually large moves in a very short period.

The stock:

• IPO priced at $185

• Closed its first day at $311.07

• Reached an all time high of $386.34

• Currently trades around $221.85

That is approximately a 43% decline from the high.

The stock also fell approximately 15% to 17% following the Q2 beat and another 12.7% around the CS 4 launch.

The market is therefore demanding extremely strong execution.

5. Competitive risk

Cerebras’s core differentiation is its wafer scale architecture and claimed inference speed advantage.

That advantage is not guaranteed to remain large.

Nvidia and other specialized AI infrastructure providers can respond through:

• Better GPU performance

• Lower pricing

• Improved inference software

• More efficient architectures

If the performance gap narrows, Cerebras’s fastest growing business, cloud inference, could face significant pressure.

DECISION TREE

Level 1 — Macro & Cost of Capital Filter

Read more

As Generative AI matures, the next major technology wave is already taking shap…

Read more

What Actually Happened

1. Softer Inflation Extended the Rates Tailwind

The macro …

Read more

TERADYNE, INC. (NASDAQ: TER)

Current Price: ~$395 | Market Cap: ~$57-$58B | Shares Outstanding: ~156.6M basic / ~157.7M diluted

Table of Contents

1. Company Overview

2. Financial Performance — Most Recent Results

3. Forward Guidance

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

Company Overview

Teradyne is one of the two dominant players in automated test equipment, or ATE, for semiconductors, alongside Advantest. Together, the two companies form a highly concentrated market.

Teradyne operates across three main segments:

Semiconductor Test: Test equipment for SoCs, memory, and increasingly AI accelerators, GPUs, and ASICs.

Product Test: System level testing, including defense and aerospace applications.

Robotics: Universal Robots collaborative arms and MiR autonomous mobile robots. The business is increasingly being positioned around AI driven factory automation.

Management’s long term strategy is built around a $6B revenue and $9.50 to $11.00 non GAAP EPS target. The strategy has two main growth drivers:

• Rising semiconductor test intensity as AI computing becomes more complex
• Expansion of Robotics into AI driven automation for electronics manufacturing and logistics

Management is targeting approximately 46% share of a $13B compute test TAM, with AI related semiconductor demand expected to be the primary growth driver.

Financial Performance — Most Recent Results

The financial results show a genuine acceleration rather than a gradual improvement.

Q2 2026 was Teradyne’s second consecutive record revenue quarter. Revenue increased 104% YoY, while GAAP EPS increased more than 300%, from $0.49 in Q2 2025 to $2.38 in Q2 2026.

Management said that AI driven demand accounted for more than 60% of Q2 sales.

The quality of the earnings result is also relatively strong. GAAP EPS of $2.38 was close to non GAAP EPS of $2.47, suggesting there was no significant one time gain or accounting distortion that needs to be adjusted away.

For this analysis, GAAP and non GAAP earnings can therefore be considered broadly comparable for the current quarter, while management guidance will continue to be assessed on its stated non GAAP basis.

Forward Guidance

For Q3 2026, based on guidance issued July 28, 2026, management expects:

• Revenue of $1.2B to $1.3B
• Non GAAP EPS of $1.85 to $2.15

The guidance implies a sequential decline in EPS from Q2’s $2.47, despite revenue being expected to remain at a similar level.

Management has attributed the decline primarily to business mix and operating expense timing, rather than a slowdown in demand. This is an important distinction, but it should still be treated as a specific reported data point rather than something to smooth over.

Full year 2026 revenue guidance implies approximately 60% growth at the midpoint.

Based on FY2025 revenue of $3.19B, this implies approximately $5.1B of FY2026 revenue.

The outlook for FY2027 is much less certain. The Street’s average FY2027 EPS estimate is approximately $9.78, with estimates ranging from $6.96 to $12.04.

This wide range highlights significant disagreement among analysts about how durable the current AI driven semiconductor test surge will be.

Management has also reiterated its longer term $6B revenue and $9.50 to $11.00 non GAAP EPS target, previously framed around 2028.

Importantly, the FY2026 revenue guidance is already tracking toward or ahead of the revenue component of that longer term target, showing how quickly the current AI driven cycle has accelerated Teradyne’s growth trajectory.

Balance Sheet & Capital Allocation

Teradyne has a conservative balance sheet relative to many AI and semiconductor infrastructure companies.

The company holds approximately $322M to $350M of cash and equivalents against only $200M to $283M of total debt, leaving it with a net cash position. Shareholder equity is approximately $2.8B, implying a debt to equity ratio of only around 7.2%.

The balance sheet has also remained strong while Teradyne has continued to invest and return capital. The company repaid $200M of short term debt during the first half of 2026, while also funding acquisitions, share buybacks, and dividends. Cash still increased during the period.

Teradyne also pays a modest quarterly dividend, although the yield is well below 1% at the current elevated share price. The company maintains an active share repurchase program as well.

Overall, capital allocation is focused on self funded growth with modest shareholder returns. This is materially different from the growth at any cost financing model sometimes seen among AI infrastructure suppliers.

This is a genuine financial strength for Teradyne.

Recent Developments Since the Prior Quarter

June 2026 — Teradyne Robotics showcased next generation AI driven automation platforms at Automate 2026, extending the company’s AI automation strategy beyond semiconductor testing and into factory floor robotics.

Robotics revenue increased 33% YoY to $100M in Q2, while revenue from the electronics manufacturing and semiconductor end markets increased 50% sequentially.

July 28, 2026 — Teradyne reported its Q2 2026 results, delivering record quarterly revenue of $1.329B, up 104% YoY.

GAAP EPS was $2.38, while non GAAP EPS was $2.47. Both exceeded the high end of the company’s previous guidance.

Management also disclosed that Teradyne had completed correlation with a second hyperscale AI customer and had shipped its first merchant GPU order.

Late July 2026 — Several sell side firms raised their price targets following the Q2 results:

• Morgan Stanley: $397, up from $387
• JPMorgan: $465, up from $450
• UBS: $500, up from $440, maintaining a Buy rating

All three firms raised their targets shortly after the results. None reduced their targets.

Early August 2026 — Northland Securities raised its earnings estimates again following the Q2 earnings beat and stronger Q3 guidance.

Q3 2026 results — Teradyne has not yet reported Q3 results as of this analysis. The latest forward guidance is $1.2B to $1.3B of revenue and $1.85 to $2.15 of non GAAP EPS.

The Q3 report is expected in late October 2026, although the exact date has not yet been formally announced.

Key Risk Factors

1. Customer concentration. Teradyne’s five largest direct customers accounted for 44.0% of 2025 revenue, while one specified customer accounted for 19.0%.

   This is unusually high concentration for a company with a market capitalization of approximately $57B. A spending pause by one major hyperscaler or a shift toward an internal testing strategy by a large customer could therefore have a material impact on Teradyne’s consolidated results.

2. Q3 guidance points to margin pressure. Q3 2026 non GAAP EPS guidance of $1.85 to $2.15 is meaningfully below Q2’s $2.47 actual EPS, despite revenue guidance remaining at a similar level.

   Management attributes this primarily to mix and operating expense timing. Nevertheless, it is an important data point because it shows that margins may not improve smoothly even during a strong semiconductor up cycle.

3. Semiconductor capital equipment remains cyclical. Teradyne’s share price has ranged from $103.00 to $487.91 over the past 52 weeks, representing more than a 4.7x trough to peak move.

   Semiconductor test equipment demand has historically been highly cyclical. Strong capacity investment can be followed by periods of inventory digestion and lower capital expenditure.

   The current AI driven cycle may be structurally stronger than previous cycles, but it does not eliminate the risk of another sharp downturn.

4. Execution risk against the long term target. Teradyne’s $6B revenue and $9.50 to $11.00 EPS target depends on several assumptions continuing to play out:

   • WFE capital expenditure increasing from approximately $140B in 2025 to $250B by 2028 to 2030
   • Continued increases in semiconductor test intensity
   • Successful market share gains in compute testing
   • Continued competitive performance against Advantest

   AI infrastructure spending can remain strong while one or more of these assumptions still fall short. The long term target is therefore achievable only if Teradyne executes across multiple variables simultaneously.

5. US China trade and export control exposure. As a semiconductor capital equipment supplier, Teradyne remains exposed to changes in US export controls and semiconductor trade policy involving China.

   This is a common risk across the semiconductor equipment industry, but it remains particularly relevant for Teradyne given its exposure to major leading edge chipmakers and the strategic importance of AI semiconductors.

DECISION TREE

Level 1 — Macro & Cost of Capital Filter

Read more

PALANTIR TECHNOLOGIES INC. (NASDAQ: PLTR)

Current Price: ~$172 | Market Cap: ~$440B | Diluted Shares Outstanding: ~2.57B

Table of Contents

1. Financial Performance — Most Recent Results

2. Forward Guidance

3. Balance Sheet & Capital Allocation

4. Recent Developments Since the Prior Quarter

5. Key Risk Factors

6. Decision Tree

7. Financial Health & Unit Economics

8. Valuation, Scenario Mapping & Expected Value

9. Technical & Execution Confluence

10. Final Recommendation

11. Expected Value Summary

12. Next Actions / Monitoring Items

    ---

Financial Performance — Most Recent Results

Q2 2026 was another exceptional quarter, but the more important development was where the growth came from.

Revenue reached $1.935B, up 92.8% YoY and 18% sequentially, while GAAP operating income reached approximately $912M, representing a 47% operating margin. GAAP net income reached $1.062B, equivalent to a roughly 55% net margin.

The commercial business is rapidly becoming the centre of the investment thesis.

U.S. Commercial revenue increased 149% YoY to $764M, while total commercial revenue reached $945M, up approximately 110%. U.S. Government revenue also grew an impressive 90% YoY to $809M. This matters because Palantir is no longer asking investors to believe that AIP can eventually create a large commercial business. The commercial business is already approaching the size of the government business.

U.S. commercial customer count also increased to 653, while U.S. commercial remaining deal value reached approximately $6.24B, up 124% YoY. Quarterly U.S. commercial contract value exceeded $2B, providing additional evidence that the current growth rate is being supported by new bookings rather than simply accounting timing.

The other important development is that government growth has not slowed while commercial growth accelerated. U.S. Government revenue grew 90% YoY, meaning Palantir is currently benefiting from two powerful growth engines rather than having to replace a slowing government franchise with commercial AIP.

GAAP profitability is also increasingly difficult to dismiss as a distorted headline. The company generated more than $1B of GAAP net income in a single quarter while still growing revenue close to 100%.

The main accounting caveat remains stock based compensation.

Palantir continues to report significantly higher adjusted profitability than GAAP profitability because adjusted metrics exclude stock based compensation and related costs. SBC therefore remains a genuine economic cost to shareholders because it transfers part of the company’s economic value to employees through equity.

The correct conclusion is not that Palantir’s profits are fake. They are clearly real. The conclusion is that GAAP earnings and per share economics should remain the primary valuation anchor, rather than adjusted earnings alone.

Updated verdict: EXCELLENT. The fundamental quality of the business has improved further, particularly because commercial AIP is now demonstrating real scale while government growth remains exceptionally strong.

Forward Guidance

Following the Q2 results, management raised full year 2026 revenue guidance to approximately $8.15B to $8.158B, implying roughly 82% YoY growth.

Q3 revenue guidance is approximately $2.160B to $2.164B, implying another quarter of more than 80% YoY growth.

Management also raised expectations for U.S. Commercial revenue to more than $3.42B for 2026, up from the previous expectation of approximately $3.22B.

This is important because the guidance now looks less like a heroic long term assumption and more like a continuation of an already accelerating business.

The key question has therefore changed.

The old question was:

Can Palantir convert AIP enthusiasm into actual commercial revenue?

Q2 largely answered that question.

The new question is:

How long can Palantir sustain exceptionally high commercial growth as its revenue base becomes much larger?

That is a considerably harder question.

The company has now demonstrated several consecutive quarters of beating and raising expectations, and Q2 provided another major upside revision. The immediate risk is therefore not a collapse in demand. It is growth normalization.

Even if Palantir grows revenue by approximately 82% in 2026, maintaining 80% growth becomes mathematically difficult as the revenue base approaches $10B. The market will eventually shift from rewarding absolute growth to evaluating the durability of that growth.

Street expectations for FY2027 remain around the low $11B revenue range, implying a substantial deceleration from 2026 but still an extremely strong growth rate.

That deceleration is likely to become the central valuation debate over the next 12 to 18 months.

Balance Sheet & Capital Allocation

Palantir continues to have one of the strongest balance sheets in the software industry.

The company has approximately $2B of cash and equivalents and more than $7B of marketable securities, giving it roughly $9B to $10B of highly liquid financial assets.

There is effectively no debt burden and no meaningful leverage risk.

This provides considerable strategic flexibility.

Palantir can continue investing aggressively in R&D, sales capacity, international expansion, AI infrastructure and new products without relying on external financing.

The balance sheet also provides downside protection during periods of market stress because the company is not dependent on debt markets to fund growth.

The less obvious issue remains dilution.

Stock based compensation is substantial, and unrecognized equity compensation represents future dilution that investors should continue to monitor. This does not invalidate the business model, but it means investors should focus on GAAP earnings per diluted share and diluted share count, rather than simply looking at absolute net income or adjusted free cash flow.

The balance sheet is therefore excellent, but it should not be described as completely shareholder neutral.

Recent Developments Since the Prior Quarter

August 3, 2026

Palantir reported another exceptional quarter.

Revenue reached $1.935B, up 92.8% YoY. GAAP net income exceeded $1B. U.S. Commercial revenue increased 149% to $764M, while U.S. Government revenue increased 90% to $809M. Full year guidance was raised again.

Early August 2026

The market responded aggressively to the earnings report. Shares gained roughly 40% over the following week, including a particularly strong move immediately after the earnings release. The stock had previously experienced a roughly 30% drawdown from its 2026 highs, making the reversal a useful reminder of how aggressively PLTR can reprice when expectations change.

Commercial AIP momentum

U.S. Commercial revenue is now $764M per quarter, compared with $809M for U.S. Government revenue.

The gap between the two businesses has therefore narrowed dramatically.

If current trends continue, commercial revenue could become the company’s largest revenue category within the next few quarters.

This is arguably the most important development in the entire Q2 report.

Commercial contract momentum

U.S. Commercial remaining deal value increased to approximately $6.24B, while quarterly U.S. commercial contract value exceeded $2B.

This suggests that the commercial acceleration is being supported by a meaningful increase in customer commitments rather than simply a temporary revenue recognition effect.

Sovereign AI positioning

Palantir is increasingly positioning its technology around sovereign and controlled AI deployment, allowing enterprises and governments to retain control over their data, models and infrastructure.

This is strategically important as AI adoption moves from experimentation into production.

Large enterprises and governments may be reluctant to place sensitive operational data entirely inside the infrastructure of a handful of hyperscalers or frontier AI companies.

Palantir’s ability to integrate AI directly into an organization’s existing operational systems could therefore become an increasingly important differentiator.

Key Risk Factors

1. Growth normalization

The largest risk is no longer whether Palantir can grow.

It is whether it can continue growing at a rate high enough to justify its valuation.

An 80% growth rate is exceptional. Maintaining 40% to 50% growth for several years would still be exceptional. But the valuation requires something closer to the second outcome than a normal software growth trajectory.

2. Valuation compression

The business has become stronger, but the stock has also become more expensive following the post earnings rally.

At roughly $172, investors are paying a substantial premium for future earnings that are still several years away.

This creates asymmetric sensitivity to any disappointment.

A company can report excellent results and still produce poor shareholder returns if the valuation multiple contracts faster than earnings grow.

3. Commercial AIP competition

Palantir increasingly competes for commercial AI spending against hyperscalers, large language model providers, enterprise software companies and internal AI teams.

The company’s advantage is its ability to integrate AI into complex operational environments.

The risk is that increasingly capable foundation models and AI development tools eventually make parts of Palantir’s software stack easier to replicate.

4. Government concentration

Despite the rapid commercial expansion, government remains a major component of the business.

U.S. Government revenue reached $809M in Q2 and total government revenue was approximately $990M.

Government procurement, appropriations, contract renewals and political priorities therefore remain important variables.

5. U.S. concentration

U.S. revenue is now more than 80% of total revenue.

The acceleration in the United States is a major strength, but it also highlights that Palantir has not yet demonstrated the same level of commercial penetration internationally.

International expansion therefore remains both a major opportunity and an important test of the long term TAM.

6. Stock based compensation

SBC remains a meaningful economic cost.

The investment case should therefore focus on GAAP earnings per diluted share and long term dilution rather than relying exclusively on adjusted earnings or adjusted free cash flow.

7. Insider selling

Continued executive stock sales remain worth monitoring.

Insider selling does not automatically indicate deteriorating fundamentals because sales can reflect diversification, taxes and predetermined 10b5b1 plans.

However, persistent selling at very high valuations is still relevant to the supply and demand dynamics of the stock.

DECISION TREE

Financial Health & Unit Economics

Flow metrics: EXCEPTIONAL.

Q2 2026 produced approximately a 47% GAAP operating margin and a 55% GAAP net margin while revenue grew almost 93%.

This is an extraordinary combination.

Most companies must choose between rapid growth and high profitability during their scaling phase. Palantir is currently demonstrating both.

The company’s Rule of 40 profile is therefore far beyond normal software standards.

Balance sheet: EXCELLENT.

Palantir has approximately $9B to $10B of cash and marketable securities and effectively no debt.

There is no meaningful balance sheet constraint on future investment.

Customer economics: IMPROVING.

The most encouraging development is not simply revenue growth.

It is the combination of:

U.S. Commercial revenue growth of 149%

Commercial customer expansion

Remaining deal value growth of 124%

Quarterly contract value above $2B

Rapid growth in commercial deployments

This suggests AIP is increasingly moving from pilot projects into production deployments.

ROIC versus WACC: STRONGLY POSITIVE.

Palantir is generating extraordinary profits relative to its invested capital, while requiring relatively little debt or external capital to support growth.

The business therefore appears to be generating returns substantially above its cost of capital.

The question is not whether Palantir currently creates economic value.

It clearly does.

The question is how much of that future economic value is already embedded in the stock price.

Verdict: PASS.

The fundamental business quality is among the strongest in the software universe.

The only meaningful financial caveat is dilution.

Valuation, Scenario Mapping & Expected Value

Read more

SPACE EXPLORATION TECHNOLOGIES CORP (NASDAQ: SPCX)

Current Price: ~$130 | Market Cap: ~$1.5T | Shares Outstanding: ~11.3B

Table of Contents

1. Company Overview

2. Financial Performance — Most Recent Results

3. Forward Guidance

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

Company Overview

SpaceX, legally known as Space Exploration Technologies Corp, operates three disclosed segments:

Space: Orbital launch services through Falcon 9, Falcon Heavy, and Starship, which remains in development. SpaceX conducts more orbital launches annually than any other provider.

Connectivity: Starlink, its satellite broadband business, with more than 9,600 satellites deployed and 10.3 million subscribers as of the S 1 disclosure date.

Artificial Intelligence: Grok, X, and related data center infrastructure.

The combination of these three businesses is important to the investment story. SpaceX is no longer a pure play space company. Its equity story now depends on execution across three very different competitive markets: space launch, satellite connectivity, and artificial intelligence.

Financial Performance — Most Recent Results

The key takeaway is unusual: revenue is growing extremely quickly while SpaceX remains loss making under GAAP.

Revenue increased 92% YoY in Q2 2026, its first reported quarter as a public company. The result also exceeded analyst estimates of $6.72B to $6.81B cited across sources.

At the same time, the company remains loss making at both the operating and net income levels. Adjusted EBITDA has moved sharply into positive territory, reaching $3.5B in Q2 2026, up 191% YoY.

This is not simply the result of a one time gain or a difference between GAAP and non GAAP accounting. SpaceX is still spending heavily on future growth. Capital expenditure alone was reported at $18.4B in Q2 2026.

The company is therefore investing ahead of its current revenue base. Revenue is already very large in absolute terms, but it has not yet caught up with the level of spending required to build out the business.

Starlink is currently the main financial engine. Connectivity generated $11.387B of revenue in FY2025, representing approximately 61% of total company revenue. The segment also generated $4.423B of operating income.

This means the Connectivity business was already solidly profitable even though SpaceX as a whole reported an operating loss. This implies that the Space and/or AI segments are currently loss making and are weighing on the consolidated result.

For this analysis, GAAP operating income and net income should remain the primary measures of profitability. Adjusted EBITDA is useful as a supplementary indicator, but it should not be treated as the headline measure given the company’s limited public reporting history and relatively new non GAAP framework.

Forward Guidance

On the Q2 2026 earnings call, CFO Bret Johnsen said SpaceX is on pace to reach $100 billion in annualized recurring revenue by the end of 2026.

This is an extremely aggressive target.

Q2 2026 revenue was $7.8B, which implies an annualized quarterly run rate of approximately $31B. Reaching $100B of annualized recurring revenue by year end would therefore require the company to more than triple its current quarterly run rate within just two quarters.

This is also SpaceX’s first quarter of guidance as a public company. There is therefore no historical record of the company consistently meeting, exceeding, or missing its guidance.

The $100B ARR target should consequently be treated as a single management assertion rather than a proven guidance range. There is not yet enough public history to assess how reliably management sets or achieves its targets.

Balance Sheet & Capital Allocation

SpaceX’s balance sheet reflects two major transactions completed around its IPO: a nearly $86B equity raise, including the underwriters’ option, followed by a $25B debt raise less than two weeks later.

The company now holds approximately $100.0B in cash and equivalents against roughly $39.7B of total debt, giving it a net cash position of approximately $60.3B.

Total shareholder equity is approximately $41.6B, while total assets are around $102.1B. This implies a debt to equity ratio of approximately 31%. While SpaceX has a very large cash balance, it is meaningfully more leveraged than a typical debt free hypergrowth technology IPO.

The bigger issue is cash generation. Trailing twelve month free cash flow is approximately negative $32.5B, based on operating cash flow of roughly $9.9B and capital expenditure of approximately $42.4B.

At the current spending rate, the $100B cash balance represents a multi year runway rather than an indefinite cash cushion. If capital expenditure continues at its current pace, SpaceX will eventually need either significantly higher cash generation, further debt, additional equity, or some combination of the three.

There is currently no dividend. Capital allocation remains focused entirely on growth, including:

• Starlink satellite deployment
• Starship development
• AI and data center infrastructure
• Other expansion initiatives across the company’s businesses

The current strategy is therefore focused on reinvestment and expansion rather than returning capital to shareholders.

Recent Developments Since the Prior Quarter

June 12, 2026 — SpaceX priced its IPO at $135 per share, making it the largest IPO in history at an implied valuation of approximately $1.8T. The shares opened around $150 and closed their first trading day at $160.95, a 19% gain. The stock subsequently reached an all time high of $225.64 within days.

June 23, 2026 — SpaceX raised $25B in debt, less than two weeks after the IPO. This materially changed the company’s balance sheet from a largely equity funded structure to one carrying meaningful leverage alongside its large cash position.

August 4, 2026 — SpaceX reported its first quarterly results as a public company. Revenue reached $7.8B, up 92% YoY and ahead of consensus. The net loss narrowed to $541M, while adjusted EBITDA increased 191% YoY. CFO Bret Johnsen also guided toward $100B of annualized recurring revenue by the end of 2026. Despite the strong results, the stock fell approximately 8% in extended trading as investors focused on the approaching lockup expiration.

August 6, 2026 — The first tranche of the IPO lockup expired. Up to approximately 911.5 million insider shares became eligible for sale. This represented 20% of the restricted shares and was tied to the second trading day after the Q2 earnings release.

At the time, the pre unlock public float was reported to be below 280.1 million shares. The potential release of more than 900 million additional shares therefore represented a multi fold increase in tradable supply within the same week.

Ongoing through 2027 — Additional shares will become eligible for sale through a series of staggered lockup releases. The full 180 day lockup expires on December 8, 2026.

The largest single tranche is expected to become eligible on June 14, 2027, involving approximately 6.4 billion shares, including shares held by CEO Elon Musk.

Key Risk Factors

1. Imminent and staggered insider share unlocks. The August 6, 2026 unlock made up to approximately 911.5 million shares eligible for sale, compared with a pre unlock float of less than 280 million shares. Further releases will continue through the December 8, 2026 full lockup expiration, followed by the approximately 6.4 billion share tranche on June 14, 2027. This creates a clearly defined and already partially realized supply overhang that is separate from the company’s underlying operating performance.

2. Governance concentration and Musk’s time allocation. SpaceX faces significant key person and governance risk. Reported consideration of a dual class share structure could preserve founder control, while CEO Elon Musk is simultaneously involved with Tesla, xAI, X, Neuralink, and The Boring Company. This creates both governance concentration and potential management bandwidth risks.

3. Heavy and ongoing cash burn. Trailing twelve month free cash flow is approximately negative $32.5B, compared with approximately $100B of cash. If capital expenditure remains above $18B per quarter, the current cash balance provides a multi year runway rather than an indefinite one. Without a significant improvement in cash generation, SpaceX may eventually require additional capital or a substantial revenue and profitability inflection.

4. AI and Grok specific regulatory and litigation exposure. Grok has faced legal scrutiny in multiple jurisdictions relating to its content moderation policies. This creates a new category of regulatory and litigation risk that is separate from SpaceX’s existing space related exposure, including FAA oversight of Starship launch cadence, environmental issues, and debris related incidents.

5. Spectrum and competitive risk to Starlink. Starlink faces ongoing spectrum coordination disputes involving Amazon’s Project Kuiper and European satellite operators. An adverse decision by the FCC, ITU, or European regulators could restrict Starlink’s ability to expand its network capacity. This risk is particularly important because Connectivity generated approximately 61% of FY2025 revenue and the majority of segment level profitability. Any regulatory constraint on Starlink therefore has the potential to directly affect the company’s most important current earnings engine.

6. No multi year public company track record. SpaceX has only one quarterly earnings report as a public company. There is therefore no established history of management meeting or missing guidance, no long term public disclosure pattern, and limited evidence for assessing normal execution volatility.

This makes it difficult to determine whether future deviations from guidance would represent normal operating fluctuations or a genuine change in the company’s growth trajectory. As a result, there is currently a structural limit to how much confidence investors can place in SpaceX’s forward guidance.

DECISION TREE

Level 1 — Macro & Cost of Capital Filter

Read more

What Actually Happened

The AI infrastructure complex had its best week since Apr…

Read more

Warren Buffett’s famous rule: “Be fearful when others are greedy, and greedy wh…

Read more

SK HYNIX INC. (NASDAQ: SKHY / KRX: 000660)

Current Price: $150.75 | Market Cap: ~$1.04 trillion (ADR-priced) | Shares Outstanding: 728.87 million common shares

Table of Contents

1. Company Overview

2. Financial Performance — Most Recent Results (KRX filings, KRW)

3. Forward Guidance

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

17. Entry/Risk framework

Company Overview

SK hynix is the world’s #2 DRAM and #2 NAND flash memory maker, behind Samsung. More importantly for the current AI cycle, it is the clear #1 supplier of High Bandwidth Memory (HBM), the specialized stacked DRAM used alongside AI accelerator GPUs.

The company began mass shipments of HBM4 in Q2 2026, reinforcing its position as the technology leader in the highest margin segment of the memory market.

On July 10, 2026, SK hynix completed a dual listing on Nasdaq under ticker SKHY through American Depositary Receipts. The offering was reported at roughly $29.4 billion, making it the largest ADR offering in history. Each 10 ADRs represent one KRX common share. The underlying common shares have traded on the Korea Exchange since 1996 under ticker 000660.

A key geopolitical consideration is SK hynix’s manufacturing footprint. Roughly 30% to 40% of its DRAM and NAND output is manufactured in China, creating direct exposure to potential US export controls and other China related restrictions.

Financial Performance — Most Recent Results (KRX filings, KRW)

The Q2 2026 net profit of ₩93.92 trillion requires an important adjustment. Approximately ₩63.27 trillion came from a one time investment gain related to the completion of SK hynix’s sale of its stake in Japanese NAND maker Kioxia, which it had held since 2018. There was also approximately ₩1.15 trillion of FX gains.

Together, these items pushed pretax profit to ₩122.7 trillion, compared with operating profit of ₩60.54 trillion.

This Kioxia related gain is a genuine, filed one time item rather than recurring earnings. It also explains why some data providers show SK hynix trading at an artificially low trailing P/E of around 6.9x.

For valuation purposes, this analysis therefore focuses on operating profit rather than Kioxia inflated net income. A one time investment gain should not be treated as sustainable earnings power.

Importantly, removing the Kioxia gain does not weaken the underlying operating story. Q2 operating profit of ₩60.54 trillion was exceptional, with a 76% operating margin, an all time company record.

The margin expansion reflects genuine pricing power in AI memory rather than an accounting benefit:

• DRAM ASP increased approximately 30% QoQ

• NAND ASP increased by the mid 50% range QoQ

• Operating margin increased another 4 percentage points QoQ

This confirms that SK hynix is benefiting from exceptionally strong pricing across both DRAM and NAND, with HBM at the center of the current cycle.

Forward Guidance

SK hynix does not provide formal quarter ahead revenue or EPS guidance in the same way many US listed companies do. However, management commentary following the Q2 2026 results points to continued strong underlying demand.

For 2026, management expects:

• DRAM demand growth: approximately mid 20% YoY

• NAND demand growth: approximately high teens YoY

• Long term supply agreements: signed with approximately 10 customers, providing greater visibility into mid to long term volumes

• FY2026 capital expenditure: raised to ₩35 trillion+, compared with ₩29 trillion in FY2025

The increase in capex is significant. SK hynix is committing to substantially higher capital intensity at a point when memory pricing is arguably approaching a cyclical peak.

That creates an important tension in the investment case: record profitability is encouraging capacity expansion, but higher capacity eventually creates the risk of another memory supply cycle.

There is also an early warning signal in the Q2 results. SK hynix generated ₩60.54 trillion of operating profit, but this was below analyst consensus of approximately ₩64 trillion.

That shortfall matters despite the record result. It represents the first meaningful sign in this cycle that market expectations may be running ahead of even exceptionally strong underlying results.

The key question from here is therefore not whether SK hynix is performing well. It clearly is. The question is how much of the exceptional HBM and memory pricing environment is already reflected in expectations and the stock price, and how sustainable those economics will be as additional industry capacity comes online.

Balance Sheet & Capital Allocation

SK hynix ended Q2 2026 with approximately ₩88 trillion of cash and equivalents against ₩18.6 trillion of total debt, giving it around ₩69.4 trillion of net cash.

This is a genuinely strong balance sheet with low leverage. It gives the company the financial capacity to fund its ₩35 trillion+ FY2026 capex program largely through operating cash flow rather than relying on additional debt.

SK hynix also announced a new 2025 to 2027 shareholder return program:

• Fixed annual dividend increased 25% to ₩1,500 per share, from ₩1,200

• Fixed cash dividends are expected to total approximately ₩1 trillion per year

• A variable dividend component will be based on 50% of accumulated free cash flow

• No specific share buyback commitment or timetable has been disclosed

At an estimated ₩50 trillion of FY2026 operating cash flow and more than ₩35 trillion of capex, free cash flow could be around ₩15 trillion. That leaves only a modest amount available for the variable dividend component in the near term.

The overall capital allocation profile is therefore clear: SK hynix is primarily a capital intensive growth company funding its own expansion, rather than an income focused investment.

Recent Developments Since the Prior Quarter

July 10, 2026: SKHY ADR began trading on Nasdaq at a reference price of approximately $166 per ADR. The listing was reported as the largest ADR offering in history, at roughly $29.4 billion.

July 16, 2026: SK hynix and SanDisk fell approximately 7%, while Micron declined around 5%, after news that China’s CXMT was preparing an $8.6 billion IPO triggered renewed concerns about future competition in the memory market.

July 27, 2026: ChangXin Memory Technologies, or CXMT, China’s largest DRAM manufacturer, listed on Shanghai’s STAR Market. The stock surged 466% on its debut, reaching a market capitalization of approximately ¥3.3 trillion. The dramatic listing reignited concerns that China’s expanding DRAM capacity could eventually create additional global supply and pressure memory pricing.

July 28, 2026: SK hynix reported record Q2 2026 results, but operating profit came in below analyst consensus. Combined with renewed CXMT concerns and reports of a Chinese DUV lithography breakthrough, SKHY ADRs fell as much as 8.69%, reaching a new post listing low and trading below the $166 IPO reference price.

July 29 to 30, 2026: Selling pressure related to CXMT continued across the memory sector. SK hynix and Micron remained weak even though SK hynix had just delivered the strongest quarterly results in its history on an absolute basis.

Key Risk Factors

1. Memory industry cyclicality

This is a demonstrated risk, not simply a generic industry disclaimer. During the 2023 memory downturn, SK hynix recorded an operating loss of ₩7.73 trillion and a net loss of ₩9.14 trillion.

That provides a clear historical example of how quickly profitability can reverse when memory pricing turns. A business generating a 76% operating margin today can move into losses when supply and pricing conditions deteriorate.

2. CXMT and Chinese DRAM competition

China’s CXMT has already demonstrated that it can materially affect SK hynix’s share price through industry sentiment alone. SKHY fell 8.69% on July 28 following the CXMT developments.

The larger concern is a sustained expansion of Chinese DRAM production. If CXMT successfully increases competitive capacity, it could eventually weaken the pricing power supporting SK hynix’s record margins.

3. China manufacturing and export controls

Approximately 30% to 40% of SK hynix’s DRAM and NAND output is manufactured in China.

The US government has reportedly been considering whether to revoke waivers that currently allow SK hynix, Samsung, and TSMC to import certain equipment and upgrade technology at their Chinese fabs without obtaining individual licenses.

This represents a specific regulatory risk that could affect SK hynix’s ability to maintain or upgrade its Chinese manufacturing operations.

4. SKHY ADR premium risk

The Nasdaq ADR carries an additional structural risk that is separate from the underlying business.

SKHY has traded at a significant premium to KRX parity, reaching as high as approximately 33%, compared with roughly 25% to 30% implied at current prices. The premium developed partly because SK hynix limited conversion of common shares into ADRs to 2.5% of shares outstanding, with that quota fully exhausted at the listing.

As two way conversion between the ADR and Korean shares normalizes, this premium could compress toward zero for purely mechanical reasons.

This means SKHY ADR investors can lose value even if the underlying SK hynix business performs well, simply because the ADR premium disappears.

5. Late cycle capex risk

FY2026 capex guidance has been raised to ₩35 trillion+, compared with ₩29 trillion in FY2025.

The concern is not the absolute level of investment. SK hynix clearly needs additional capacity to meet AI and HBM demand. The concern is timing.

The company is committing to significantly higher capacity at what may prove to be a cyclical peak in memory pricing. Historically, aggressive capacity expansion near the top of the memory cycle has contributed to subsequent periods of oversupply and falling prices.

Bottom line: SK hynix enters the next phase of the cycle with an exceptionally strong balance sheet, record profitability and unmatched HBM positioning. The key risk is that today’s extraordinary economics encourage the entire industry, particularly China, to add capacity too aggressively. The investment case therefore depends increasingly on whether AI and HBM demand can grow fast enough to absorb new supply without destroying the pricing power that currently drives SK hynix’s exceptional margins.

DECISION TREE

Level 1 — Macro & Cost of Capital Filter

Read more

META PLATFORMS, INC. (NYSE: META)

Current Price: $596.10 | Market Cap: $1.52 trillion | Shares Outstanding: 2.55 billion

Table of Contents

1. Follow Up Company Overview

2. Financial Performance — Most Recent Results

3. Forward Guidance

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

17. Entry/Risk framework

Follow Up Company Overview

Meta operates through two reported segments:

Family of Apps includes Facebook, Instagram, WhatsApp, Messenger, and Threads. This is the core advertising business, supported by 3.60 billion daily active people. Increasingly, Meta is also using this distribution network to build an AI assistant business.

Reality Labs covers AR and VR headsets and AI enabled smart glasses. The segment remains deeply loss making.

The core business model remains straightforward: sell targeted advertising to a massive and highly engaged user base.

However, Meta is now layering a much more capital intensive strategy on top of that business.

The company is investing heavily in its own AI infrastructure, including:

• Data centers

• Custom silicon

• Third party GPUs

The goal is to improve its advertising and recommendation systems while also supporting its broader AI ambitions.

The advertising benefits are already visible in the numbers. In Q2 2026, ad impressions increased 14% year over year, while price per ad increased 12%.

Management has also talked about pursuing general purpose and “superintelligence” AI models. That is management’s strategic ambition rather than a filed financial metric, so this analysis treats it as a strategic narrative rather than a proven financial outcome.

Financial Performance — Most Recent Results

The Important Earnings Quality Point

The headline GAAP EPS figures from the last two quarters are materially distorted by unusual items.

This analysis therefore adjusts both quarters before using them as the earnings base.

Q1 2026

Reported diluted EPS was $10.44.

However, Q1 included an $8.03 billion one time income tax benefit related to Corporate Alternative Minimum Tax transitional relief.

After removing that benefit, adjusted diluted EPS was approximately $7.31, rather than $10.44.

Q2 2026

Reported diluted EPS was $6.18.

The quarter included:

• $2.40 billion of legal proceedings charges

• $1.18 billion of severance expense related to an 8,000 person headcount reduction

After adjusting for these items on an after tax basis, EPS was approximately $7.35.

That figure was actually close to the $7.14 to $7.22 Street consensus.

What This Means

The underlying operating business is executing well.

The problem is that the headline GAAP EPS series has been unusually noisy in both directions during 2026.

For valuation purposes, this analysis therefore uses:

Q1 adjusted EPS: ~$7.31

Q2 adjusted EPS: ~$7.35

rather than the reported $10.44 and $6.18 figures.

That provides a more useful picture of Meta’s underlying earnings power.

Forward Guidance

Q3 2026 revenue: $61 billion to $64 billion

FX impact: approximately 1% headwind

FY2026 total expenses: $165 billion to $169 billion

FY2026 capex: $130 billion to $145 billion, including finance lease principal payments

FY2026 tax rate: 15% to 17%

The most important point is the continued increase in the capex outlook.

Meta has raised its 2026 capex range three times:

The tax rate guidance was also raised from 13% to 16% to 15% to 17%.

The key issue is therefore not simply that Meta is spending heavily on AI. The bigger concern is the direction and pace of the spending revisions.

Capex continues to move higher during the year rather than converging toward a stable plan. That is the opposite of the disciplined and predictable investment pattern that would give investors greater confidence that the AI spending will ultimately translate into higher revenue and free cash flow.

The core advertising business remains exceptionally strong. The question for investors is increasingly whether AI infrastructure investment can generate returns quickly enough to justify the rapidly rising capital requirements.

Balance Sheet & Capital Allocation

Meta ended Q2 2026 with $90.26 billion of cash and marketable securities against $83.66 billion of long term debt, leaving approximately $6.6 billion of net cash.

That cushion has narrowed significantly. Meta raised $25 billion to $30 billion of bonds in each of two separate offerings, in October 2025 and May 2026, specifically to help fund its AI data center buildout.

The bigger concern is cash generation. Q2 free cash flow fell to just $784 million, dramatically below historical levels, as $31.08 billion of quarterly capex absorbed most of the cash generated by operations.

Meta also repurchased zero shares in Q2 2026. This is a filed change in capital allocation and is consistent with management prioritizing cash preservation for AI infrastructure investment.

The dividend continues at approximately $2.10 per share annualized, equivalent to a yield of around 0.35%, but it remains immaterial compared with the scale of Meta’s AI capex and debt financing.

Recent Developments Since the Prior Quarter

July 29, 2026: Meta reported Q2 2026 results after the close. Revenue beat expectations, but GAAP EPS was affected by legal and severance charges. Meta also raised its capex range again. The stock fell as much as 9% the following morning, to around $530.

July 30, 2026: Several major sell side firms reduced their price targets while maintaining positive ratings:

July 31 to August 3, 2026: Meta rebounded approximately 14% from its post earnings low. Sell side commentary largely attributed the recovery to improving sentiment toward large scale AI infrastructure spending after a peer cloud company’s better than expected results.

Importantly, this appears to have been more sentiment driven than Meta specific. There was no new Meta disclosure during this period that resolved the underlying concerns around rising capex and collapsing free cash flow.

January 20, 2026: The FTC filed a notice of appeal against the November 2025 district court ruling that cleared Meta of illegal monopoly allegations related to its Instagram and WhatsApp acquisitions. Oral arguments before the D.C. Circuit are tentatively expected in late summer or autumn 2026.

Key Risk Factors

AI capex continues to rise. Meta has increased its 2026 capex guidance three times, from $115 billion to $135 billion, then $125 billion to $145 billion, and now $130 billion to $145 billion, with some reporting putting the current lower end at $135 billion. There is still no filed evidence showing when this spending will begin translating into incremental revenue or when the AI investment will generate sufficient returns to justify its scale.

1. Free cash flow has collapsed.

   Q2 2026 free cash flow was only $784 million, down dramatically from historical levels, as $31.08 billion of quarterly capex absorbed most operating cash generation.

2. The balance sheet cushion is shrinking.

   Meta’s approximately $6.6 billion net cash position is now only a fraction of its historical cushion, following two large bond offerings over the past year. Another significant debt raise could push Meta into a genuine net debt position, something that would represent a meaningful change in its financial profile as a mature company.

3. Reality Labs remains a major financial drag.
   The segment has accumulated approximately $88 billion of losses since it was first separately reported in late 2020. The losses are still increasing, reaching $4.62 billion in Q2 2026, compared with $4.03 billion in Q1, with no filed timeline for profitability.

4. Antitrust risk remains unresolved.
   The FTC appeal of the November 2025 ruling is active at the D.C. Circuit, with oral arguments potentially taking place before the end of 2026. Meta won at the district court level, but an adverse appellate decision could reopen questions around the ownership of Instagram and WhatsApp. This is a separate structural risk from the AI capex debate.

5. The recent share price recovery may not be fully durable.
   Much of the rebound has been attributed to improved sentiment following a peer company’s cloud results rather than to any Meta specific improvement in capex efficiency or free cash flow. If that broader AI infrastructure read through fades, part of Meta’s recent recovery could reverse.

Decision Tree

Level 1 — Macro & Cost of Capital Filter

Read more

The AI Manufacturing Bottleneck

The biggest bottleneck in AI is no longer GPU de…

Read more

What Actually Happened

This was the week the AI infrastructure trade split into …

Read more

Here’s the updated full portfolio (as of 31st July 2026):

Read more

SENTINELONE, INC. (NYSE: S)

Current Price: ~$18.22 | Market Cap: ~$6.1B | Shares Outstanding: ~335M

Table of Contents

1. Company Overview

2. Financial Performance — Most Recent Results (Fiscal year ends June 30)

3. Forward Guidance & the May 28 Restructuring

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

17. Entry/Risk framework

Company Overview

SentinelOne provides an AI native cybersecurity platform called Singularity, centered on autonomous endpoint detection and response. Its core differentiation is the ability of its AI engine to detect, investigate, and remediate threats without waiting for human analyst approval.

The platform has expanded well beyond endpoint security into:

1. Cloud security

2. Identity security

3. AI driven data and SIEM capabilities

4. Purple AI, its agentic AI security assistant

Purple AI has recently added Agentic Investigation and a Prompt Security integration with Amazon Bedrock AgentCore, extending SentinelOne’s AI capabilities further into automated security investigation and AI application security.

The company operates in a highly competitive endpoint security market. Its largest competitors include CrowdStrike and Microsoft Defender for Endpoint. Third party market share estimates generally place SentinelOne around third, with approximately 9% to 10% market share, compared with roughly 20% for CrowdStrike and 11% to 14% for Microsoft Defender. These estimates vary by methodology and should therefore be treated as directional rather than precise.

Financial Performance — Most Recent Results (Fiscal year ends January 31)

FY2026 was an important milestone, with SentinelOne crossing $1B of annual revenue for the first time.

Gross margins remained strong:

1. GAAP gross margin: 74%

2. Non GAAP gross margin: 79%

The more important issue is the widening gap between GAAP and non GAAP profitability.

GAAP net loss margin deteriorated from 35% to 45%, while non GAAP net margin improved from 2% to 7%.

This divergence is largely explained by stock based compensation, which represented roughly 30% of FY2026 revenue, together with other non cash and non recurring items. It does not necessarily indicate deterioration in the underlying operating economics, but the level of stock based compensation is high and remains an important shareholder cost.

The customer base also continued to expand. Customers generating more than $100,000 of ARR increased 18% to 1,667.

Q1 FY2027: The more important growth signal

The May 28, 2026 results showed a potentially meaningful acceleration in future growth.

Net new ARR increased 55% year over year, a record, adding $44M during the quarter. Total ARR growth remained at 23%, meaning the forward growth rate is improving faster than the headline ARR growth rate currently shows.

AI security ARR was particularly strong, nearly doubling during the quarter.

Cash generation also improved:

1. Adjusted free cash flow: $61.4M, up from $45.4M.

2. Adjusted FCF margin: 22%, up from 20%.

Cash, cash equivalents, and investments stood at $812M as of April 30, 2026, compared with approximately $1.2B around nine months earlier. The change could not be fully reconciled from public sources in this analysis, so it should be treated as a monitoring item rather than attributed to a specific cause.

The company carries no significant debt, providing additional financial flexibility.

Forward Guidance & the May 28 Restructuring

FY2027 revenue guidance is $1.195B to $1.205B, representing approximately 20% year over year growth at the midpoint.

Q2 FY2027 guidance is:

1. Revenue: $289M to $291M

2. Year over year growth: approximately 20%

This represents modest growth deceleration from 22% revenue growth in FY2026.

At the same time, management announced a restructuring affecting approximately 300 employees, or roughly 8% to 10% of the workforce depending on the source.

The restructuring carries a one time charge of approximately $25M and is expected to generate approximately $45M of annualized savings.

Importantly, management plans to redirect those savings toward:

1. AI

2. Data

3. Cloud

4. Endpoint R&D

The restructuring therefore appears designed to improve operating efficiency while continuing to invest in the areas generating the strongest growth.

The financial targets also improved:

1. FY2027 non GAAP operating margin: approximately 10% at the midpoint.

2. FY2027 non GAAP operating income: $115M to $125M.

This is an important distinction: the restructuring is being used to improve profitability without materially reducing investment in the product areas central to the company’s growth strategy.

Balance Sheet & Capital Allocation

SentinelOne has a relatively conservative SaaS balance sheet, with no significant debt and approximately $812M of cash and investments as of the latest quarter.

Capital allocation is focused almost entirely on reinvestment, particularly:

1. Research and development

2. Go to market investment

3. AI product development

4. Cloud and endpoint security capabilities

There is no dividend and no meaningful disclosed share buyback program.

The largest capital structure concern for shareholders is therefore stock based compensation and the resulting dilution.

Stock based compensation represented approximately 30% of FY2026 revenue, which is unusually high and should be treated as a genuine economic cost to shareholders even though it is classified as a non cash expense.

The improving non GAAP profitability and free cash flow are encouraging, but they should be evaluated alongside this dilution rather than viewed in isolation.

Recent Developments Since the Prior Quarter

The May 28, 2026 Q1 results and restructuring announcement initially hit the stock hard. Shares fell approximately 18% to around $14.69 as investors focused on the layoffs and interpreted the guidance profile as cautious despite the underlying operational beat.

The stock has since recovered meaningfully.

Shares are now around $18.22, approximately 24% above the post earnings low, although still roughly 15% below the 52 week high of $21.40 set earlier in the year.

Scotiabank upgraded SentinelOne to Outperform from Sector Perform in early July 2026 and raised its price target to $23.50 from $16. Shares gained approximately 6% to 7% following the upgrade.

The thesis focused on:

• SentinelOne’s AI native platform positioning
• Purple AI Agentic Investigation
• Prompt Security integration with Amazon Bedrock AgentCore
• The possibility that AI enabled attacks will push CISOs to increase cybersecurity budgets

This is important because the argument is structural rather than simply cyclical: more sophisticated AI driven attacks could require greater spending on AI enabled defense.

Jefferies also raised its target to $25 from $23 and maintained a Buy rating. Wells Fargo remained at Hold as of July 13, 2026.

Sell side sentiment is therefore constructive but not unanimous.

Consensus remains bullish but widely dispersed:

• Median target: approximately $20
• High target: $30 from Cantor Fitzgerald
• Low target: $16 from DA Davidson

The wide range reflects genuine disagreement over how much value investors should assign to the improving ARR and margin trends versus competitive pressure and shareholder dilution.

Key Risk Factors

1. Persistent GAAP losses and heavy dilution

FY2026 GAAP net margin was (45%), while stock based compensation was approximately 30% of revenue. The improving non GAAP numbers are encouraging, but they do not eliminate the economic cost of dilution.

2. Structurally weaker competitive position

CrowdStrike has roughly twice SentinelOne’s estimated market share in core endpoint security. Microsoft Defender also benefits from Microsoft’s enormous enterprise distribution ecosystem, which SentinelOne cannot replicate.

3. Recent guidance and restructuring shock

The May 28 announcement triggered an approximately 18% single day decline. This demonstrates that the market can punish SentinelOne severely when expectations change.

4. Unreconciled cash decline

Cash and investments declined from a reported $1.2B to $812M over roughly nine months. The reason could not be fully confirmed from available public information and therefore remains an open monitoring item.

5. Execution risk around the AI security strategy

Management is redirecting restructuring savings into AI, data and cloud R&D. The strategy needs to produce sustained ARR acceleration, rather than simply one strong quarter, to prove that the investment is working.

6. Enterprise IT spending sensitivity

Cybersecurity tends to be more resilient than discretionary IT spending, but a broader enterprise technology spending slowdown would still affect SentinelOne.

PHASE 1: DECISION TREE

LEVEL 1 — Macro & Cost of Capital Filter

Read more

ALPHABET INC. (NASDAQ: GOOGL)

Date: July 27, 2026 | Current Price: ~$326.56 (roughly 19-20% below the all-time high near $402-408 set mid-May 2026 | Market Cap: ~$3.9–4.0T

Table of Contents

1. Company Overview

2. Financial Performance — Most Recent Results

3. Forward Guidance & the Capex Surge

4. Balance Sheet & Capital Allocation

5. Recent Developments Since the Prior Quarter

6. Key Risk Factors

7. Decision Tree

8. Level 1 — Macro & Cost of Capital Filter

9. Level 2 — Industry Attractiveness

10. Level 3 — Company Moat & Management Quality

11. Level 4 — Financial Health & Unit Economics

12. Level 5 — Valuation, Scenario Mapping & Expected Value

13. Level 6 — Technical & Execution Confluence

14. Final Recommendation

15. Expected Value Summary

16. Next Actions / Monitoring Items

17. Entry/Risk framework

Company Overview

Alphabet is the parent company of Google and operates through three main business segments:

• Google Services: Search, YouTube, Android, Chrome, Google Play, and advertising.

• Google Cloud: Google Cloud Platform (GCP) infrastructure and enterprise AI solutions.

• Other Bets: Businesses including Waymo.

Alphabet occupies a unique position in AI. It develops its own AI models through Gemini, designs custom TPU chips, and owns one of the world’s largest consumer platforms through Search, Chrome, Android, and YouTube.

This gives Alphabet exposure from both sides of the AI market. It benefits from AI adoption through Cloud and Gemini, while also facing the risk that AI changes how people search and consume information.

A key point on earnings quality

The reported GAAP EPS figures require careful interpretation.

Both Q1 and Q2 2026 were significantly boosted by non operating gains from Alphabet’s equity investments, widely reported to include stakes in companies such as Anthropic.

• Q1 2026: $37.7 billion of other income added approximately $2.35 to GAAP EPS of $5.11.

• Q2 2026: A roughly $98 to $99 billion gain on equity securities increased net income to $112.1 billion and lifted GAAP diluted EPS to $9.11.

However, the comparable adjusted EPS used by analysts was only $2.85, versus expectations of $2.89, indicating the operating business performed largely in line with forecasts.

The underlying business remains healthy:

• Operating income increased 30% year over year to $40.8 billion in Q2.

• Operating margin remained strong at 34%.

• Google Services revenue grew 15% to $94.5 billion.

Because of these unusually large investment gains, the reported GAAP EPS should not be used directly for valuation. This analysis adjusts for those gains in Level 5.

Google Cloud continues to accelerate

Google Cloud remains Alphabet’s strongest growth engine.

This marks three consecutive quarters of accelerating growth.

As of Q1 2026, Google Cloud also reported a $460 billion backlog, providing a substantial, filed indicator of future demand rather than a promotional estimate.

Forward Guidance & the Capex Surge

The biggest takeaway from the Q2 2026 results was not revenue or earnings. It was the sharp increase in capital spending.

Management raised full year 2026 capital expenditure guidance to $195 billion to $205 billion, up from the previous $180 billion to $190 billion range, and indicated that investment will continue to increase in 2027.

Quarterly capex reached $44.9 billion in Q2 2026.

As a result, free cash flow turned negative, reaching negative $5.9 billion for the quarter.

This is a significant change for Alphabet, which has historically generated substantial positive free cash flow.

Despite beating revenue expectations and delivering another strong Cloud quarter, the stock fell roughly 6% to 7% after earnings as investors focused on rising capex and declining free cash flow.

Balance Sheet & Capital Allocation

Alphabet continues to have one of the strongest balance sheets in global technology.

Historically, the company has funded growth almost entirely through operating cash flow while maintaining minimal debt.

However, the current AI investment cycle is changing that profile.

The rapid increase in capital spending is consuming cash at a much faster pace than in previous years, with the negative $5.9 billion quarterly free cash flow serving as the clearest evidence.

Alphabet continues to return capital to shareholders through:

• A quarterly dividend of $0.22 per share (next payment scheduled for September 14, 2026).

• Ongoing share repurchases, although a precise Q2 buyback figure could not be confirmed from available sources.

Overall, Alphabet’s capital allocation strategy is shifting from returning excess cash to shareholders toward funding one of the largest AI infrastructure buildouts in the industry. This increase in capital intensity is a key trend investors should continue to monitor.

Recent Developments Since the Prior Quarter

Antitrust risk remains a major overhang

Regulatory pressure continues to be one of Alphabet’s biggest long term risks.

• The U.S. Department of Justice (DOJ) has appealed the September 2025 search remedies ruling and is continuing to seek a forced divestiture of Chrome.

• The D.C. Circuit Court is expected to hear arguments in late 2026 or early 2027, making this a clear upcoming catalyst rather than an indefinite legal overhang.

• Alphabet also continues to face parallel regulatory proceedings in Europe.

A forced sale of Chrome would be far more significant than a financial penalty because it would directly weaken Alphabet’s distribution advantage across Search, Chrome, and Android.

Gemini adoption continues to accelerate

Gemini usage continues to show strong momentum.

• Gemini is now processing 16 billion tokens per minute, up 60% quarter over quarter.

• Gemini Enterprise paid monthly active users increased 40% quarter over quarter.

These are meaningful, filed indicators that AI adoption is accelerating beyond product announcements.

Waymo continues to scale

Waymo has now surpassed 500,000 autonomous rides per week.

CEO Sundar Pichai has also told employees that he expects Waymo to make a meaningful financial contribution as early as 2027.

While Waymo remains a relatively small contributor today, it represents a growing source of long term optionality.

Analysts remain positive despite the selloff

Wall Street sentiment remains broadly constructive.

• Median analyst price targets are around $440, implying roughly 27% upside from a reference price of $346.

• Some analysts have raised targets as high as $493.30.

This optimism contrasts with the stock’s roughly 20% decline from its May high. Unlike many AI stocks, this pullback has been driven primarily by concerns over rising capex and negative free cash flow rather than a deterioration in the underlying business.

Key Risk Factors

1. Antitrust risk

The DOJ’s effort to force a Chrome divestiture remains Alphabet’s most significant structural risk.

The appeal is expected to be heard in late 2026 or early 2027, providing a clear timeline for an outcome that could materially weaken Alphabet’s distribution ecosystem.

2. Negative free cash flow

Alphabet already reported a negative $5.9 billion free cash flow quarter after spending $44.9 billion on capex in Q2 2026.

Management expects heavy investment to continue through the rest of 2026 and into 2027, meaning this is an ongoing trend rather than a one time event.

3. Earnings quality

Reported GAAP earnings are currently distorted.

Two consecutive quarters included very large gains from equity investments, significantly inflating net income and EPS. These headline figures should be adjusted before being used for valuation.

4. AI disruption to Search

Alphabet is benefiting from AI through Gemini and AI Overviews, but AI also creates long term risk for its core Search business.

Whether AI driven answer experiences, either from Alphabet or competitors, ultimately change search advertising economics remains an important unanswered question.

5. Capex execution

Alphabet plans to spend $195 billion to $205 billion on capital expenditures during 2026.

This investment assumes AI infrastructure spending will generate durable growth in Cloud and AI revenue. The $460 billion Cloud backlog supports that thesis, but execution and long term returns at this scale remain unproven.

6. Intensifying AI competition

Competition across AI continues to increase.

Alphabet competes directly with OpenAI, Microsoft, Amazon, and Anthropic for AI models, cloud infrastructure, and enterprise AI workloads. Anthropic is also an Alphabet investment, creating the unusual dynamic of being both a strategic investment and a direct competitor.

PHASE 1: DECISION TREE

LEVEL 1 — Macro & Cost of Capital Filter

Read more