Trump’s threats of invading Iran did not materialise, and don’t get me wrong, I am glad restraint held. But no strike also means no resolution.

The US is back to economic pressure, with the military option taking a back seat for now. Treasury Secretary Scott Bessent has been throwing around phrases like “economic D-Day” and “economic onslaught”. Washington even threatened severe secondary sanctions on any country or entity that keeps doing business with Iran. China fired back and warned it will retaliate. It has been buying Iranian oil at a discount and has little reason to stop.

My view? This leads nowhere, again. The US has sanctioned Iran for years. If sanctions worked, military action would never have been on the table in the first place.

Threats against China have not worked either. Trade wars, sanctions and bombs did not stop China from buying Iranian oil. So why would China’s support stop this time?

The Black Sea Adds Another Problem

And things are getting worse elsewhere. Russia and Ukraine are escalating.

Russia has been hammering Ukrainian Black Sea ports, particularly around Odesa, hitting grain elevators, port infrastructure, export terminals and even merchant vessels. A strike on 20 July killed 10 people on a maize carrier off the coast of Odesa.

Ukraine has responded with an unprecedented campaign against Russian vessels and transport networks in the Sea of Azov and the Black Sea. A major drone attack on 12 August heavily damaged grain terminals at Novorossiysk, Russia’s main Black Sea export port, potentially putting facilities out of operation for months.

The chain reaction is simple. Shipments get disrupted, commodities can’t get to where they are supposed to go. Supply tightens, prices go up. Shipping and insurance rates go up too, because risk has gone up. And higher fuel prices push both commodity and shipping costs up again.

You can see it in the prices. Look at the oil, wheat, corn, sugar and soybean ETFs over the past month in the chart below. All of them have risen, with sugar gaining the most at 17.63%.

The Breakwave Dry Bulk Shipping ETF (BDRY), a proxy for shipping rates on major dry bulk commodities like coal, iron ore and grain, is up 16% over the same period.

The Bond Market Is Already Telling You

Read more

Some companies, however, continue to operate their businesses as usual.

The chances of AI replacing such businesses are close to zero. It’s boring, but it does give investors a good night’s sleep.

Imagine this. Every week, without fail, households and businesses across America produce trash. Someone must collect it, haul it, and put it into a landfill.

Getting rid of trash has become so convenient that most of us take the last mile collection and management of trash for granted. Yet nobody wants a dump next to their house, and the permitting process takes years and tens of millions of dollars. Most attempts fail.

So whoever already owns the landfills owns something close to a license to print money. And that also means other haulers, including rivals, must pay to dump in those landfills.

Talk about first-mover advantage.

Republic Services (NYSE: RSG) is the second-largest owner of that infrastructure in North America, behind only Waste Management (NYSE: WM).

Its operations span the United States and Canada through 389 collection operations, 260 transfer stations, 84 recycling centres, and 208 active landfills. It also runs 88 landfill gas-to-energy and other renewable-energy projects.

Think Toll booths, but for trash trucks

Read more

It’s a chemical reaction as old as iron, and no algorithm is going to dissolve it. As long as metal exists, it will corrode, and something will need to displace the moisture that causes it. WD-40 has occupied that need for 73 years, on a formula finalised in the 1950s and barely touched since. A product that solves a problem this permanent is not an obvious candidate for disruption, least of all by AI.

So much so that there was a stretch when the internet joked WD-40 could fix anything, including heartbreak. The brand name is so strong that the company that invented it renamed itself after the product. It’s a listed company, founded 73 years ago.

That ubiquity is the entire investment thesis for WD-40 Company (NASDAQ:WDFC). It might be exactly why the stock is starting to look interesting again, even at what still looks like a rich valuation.

A Near-Perfect Little Machine

WD-40 stands for “Water Displacement, 40^th formula,” the fortieth attempt at a rust-preventing solvent, and apparently the one that stuck.

It’s a tool that is rarely bought, used occasionally, but almost always solves the problem it’s asked to. Its track record speaks for itself. No one stands in a hardware aisle agonising over whether to save eight cents on the generic penetrating oil next to it.

That itself sounds moat-y, and it is a great business model. WD-40 has genuine pricing power for the most durable reason a brand can have it. The product is cheap in absolute terms, essential when needed, and trusted across generations. There are real competitors (CRC, Liquid Wrench, PB Blaster) and functionally their products are fine. But brand recognition and habit mean WD-40 wins the shelf anyway.

Read more

Netflix first featured in FAANG in 2017, alongside Facebook, Amazon, Apple and Google. But it subsequently faded from the stock acronyms that followed.

We had MAMAA in 2021. Most of the original FAANG members kept their places, with some name changes, while Microsoft replaced Netflix.

Then came the Magnificent Seven. NVDA and TSLA joined the previous MAMAA cohort.

And just 2 months ago, in Jun’26, MANGOS emerged. AAPL got replaced by Anthropic, while OpenAI and SPCX replaced AMZN.

So, what happened to Netflix? Did it really fall out of favor or simply off the radar?

Netflix’s performance L5Y

Objectively, Netflix has done relatively well over the last 5 years. Its share price increased by around +50% during that period.

That’s more than AMZN’s +44% across the same period, and only slightly behind TSLA and META, both at +60%.

Not too shabby for a stock that has largely fallen off investors’ radar.

Read more

MELI is the largest online commerce ecosystem in the region, operating in 18 countries including Brazil, Argentina, and Mexico. The company offers a fully integrated suite of e-commerce and digital financial services designed to facilitate both digital and offline commercial transactions.

These are the products the company segregates itself into:

Mercado Libre Marketplace: An online platform for listing and selling various merchandise categories. Includes both 3rd party as well as its own direct sales.

Mercado Pago: A fintech platform providing secure digital payment solutions.

Mercado Envios: A logistics service offering fulfillment and shipping solutions.

Mercado Ads: An advertising solution for promoting products and services.

Mercado Shops: A platform for users to create and manage their own online storefronts. Standalone storefront builder tools for merchants integrated into the stack.

Mercado Credito: A financial solution offering credit lines to consumers and merchants.

Read more

In fact, the index itself barely budged, just 5.5% peak to trough over June and July. The selldown was concentrated in AI trades. Top names like SanDisk fell more than 50% from peak to trough. But sectors like software, cybersecurity, oil and gas, and banking held up, and the broader market supported the index while AI stocks fell. So if you were just holding the index, you might not have felt a thing.

Now that the S&P 500 is back at new highs, investors are split. One camp breathes a sigh of relief, calling the drawdown over and the market ready for another bull leg. The other sees stretched valuations and stays bearish.

I belong to the first camp. The signs point to strength, not weakness.

Excess Cleared, Breadth Restored

We’ve already touched on the first sign. AI trades corrected heavily over the last two months, and that flushed out a lot of leverage and excess from the market. Stocks that are no longer over-stretched have less reason to correct further.

At the same time, the index didn’t tank, because other sectors were rallying. That broadening is a positive, not a negative. It removes the unhealthy concentration we had before.

Look at new highs versus new lows on the NYSE (not the S&P 500). We’re seeing more highs than lows, and that’s strength, not weakness. In a bear market, new lows outnumber new highs. And if valuations were truly overstretched, we’d see new highs spike far above new lows, the way they did in February, just before the Iran War (see the spikes in the chart below). We’re nowhere near those levels today. That’s a healthy bull, by my diagnosis.

New Highs = Bullish

How many new highs has the S&P 500 made in 2026, as of 5 Aug 2026?

25

That might surprise you. Many investors assume a new high means the market is expensive and due for a fall. It’s the opposite. Markets hit new highs because they’re bullish, and a new high can be broken many times in a single year. A new high isn’t a ceiling. It’s proof the market has the energy to keep breaking through. You can see the track record of S&P 500 making numerous new highs in the past years.

So the S&P 500 making new highs again is a bullish sign, not a bearish one.

Supported By Earnings Growth

The third reason is that this rally is backed by earnings growth. S&P 500 earnings grew 50.4% year on year in Q2 2026. Fifty percent!

Some of that growth came from investment gains rather than operations. Alphabet alone pushed Communication Services earnings growth to 117%. Without Alphabet, that sector’s growth would have been only 6%. Q2 GAAP EPS included $53.4bn of gains, largely related to its Anthropic investment. Consumer Discretionary earnings growth came out at 91.6%. Without Amazon, it would have been only 7.5%.

Strip out both, and earnings growth still comes in around 32%. That’s still very strong.

In the chart below, you can see the forward 12 month EPS line accelerating since 2025, with the S&P 500 price only slightly above it. That tells you the fundamentals behind these companies are strong and genuinely profitable. Valuations aren’t as stretched as people think. Sure, you’ll find pockets of overvaluation, but on the whole, the index is backed by earnings growth. Of course, forward EPS is an expectation and things can fall apart once that positive expectation tapers.

Fed Unlikely To Hike Interest Rates

The prolonged Iran War sparked fears that inflation would come roaring back. New Fed Chair Kevin Warsh didn’t make things easier, keeping his cards close and refusing to signal the Fed’s intentions. Investors had to adjust to this new style, and that created fresh uncertainty.

The good news is that US inflation fell to 3.4% in July, giving the Fed one less reason to hike.

Based on the latest futures market, the probability of a rate hike in September 2026 has dropped to 34.7%, down from more than 50% just a week ago.

No rate hike is good news for AI capex. Companies like Alphabet are issuing new bonds to fund their expansion, and a hike would have raised their cost of funds, squeezed profits, and possibly delayed plans to borrow more and expand faster to meet demand. That worry is off the table for now, at least for the next two months, so they can carry on.

Commodity prices are one thing worth watching. Not just oil, but copper, corn, and other raw materials that go into whatever we produce. They tend to lead the inflation numbers rather than follow them.

More Bull Than Bear

So I see more reasons to be bullish than bearish, barring any sudden bad news. But let’s be clear about the timeframe. This is my read for the next one or two months, not a forecast for the rest of the year. And I could be wrong too. How things unfold from here will decide whether that call still holds.

That has really been the case. Ferrari deliberately limits how many cars it builds each year (13,640 in 2025, down slightly from 13,752 in 2024) to keep them scarce. Because demand always outstrips supply, Ferrari doesn’t need to discount, and buyers often pay extra for personalization (custom colors, materials, and options). That’s why 2025 revenue still grew 7% even though the number of cars sold fell. Customers paid more per car, not that more cars were sold. This “sell fewer, charge more” approach is the core of how Ferrari protects its brand and its profit margin.

Got the money to buy one. Sorry, you’ve got to wait.

There’s a downside to this though. Ferrari’s growth isn’t driven by volume, it’s driven by price and mix. Unit sales are flat by design, so all future growth must come from charging more per car and selling more limited editions. And price and mix has a ceiling. So basically, don’t expect Ferrari to grow fast.

Historically, Ferrari’s revenue has compounded about 10.4% a year over the past 10 years. Earnings have grown faster still, at 16.7% a year, which tells you prices are moving up faster than the cost of production. That’s Ferrari’s pricing power at work.

Margins are impressive. Net profit margin sits at 22%, ROIC at 30.7%, and ROE at 40.9%. These are numbers most companies can only dream of, but Ferrari prints them year in, year out. While automakers fight for survival in the bottom and middle tiers, scraping together profits where they can, Ferrari sits above them and avoids the bloodbath. Yes, it still has competitors, but there are far fewer of them at the top.

It is All About the Brand

The second source of money is sponsorship, commercial deals, and the Ferrari brand itself, about €820 million in 2025 (11% of revenue, and the fastest growing piece, up 22% year over year). This includes money from Formula 1 racing, both sponsorship deals and Ferrari’s cut of what Formula 1 earns from broadcasting and other commercial deals. It also includes brand licensing, Ferrari branded clothing, merchandise, theme parks, and stores, most of which are run by outside partners under license rather than by Ferrari directly.

That means Ferrari monetizes the brand very well. A highly desired brand, after all, that people are willing to pay for the logo, not just the car.

We’d argue Ferrari is the category name for luxury supercars. Yes, there are other options like Lamborghini, but we think the mindshare for Ferrari remains stronger, and reaches more people. We can’t prove it, but one way to test it is to ask around. “What brand do you think of first when we talk about supercars?” We believe Ferrari’s name would come up pretty much most of the time.

The brand is what keeps Ferrari desired. If the branding weakens, so does the revenue and profitability of the company. If Ferrari starts to sell more cars than the market wants, or future generations lose their affinity with the brand, then Ferrari goes downhill. But we believe both are low probability. So this isn’t about a competitor outspending Ferrari. As long as the Ferrari brand is intact, business should stay as good as usual.

Is an Electric Ferrari Still a Ferrari?

Ferrari’s premium rests on engine engineering, on sound, feel, mechanical mystique. So with the world moving to EV, and Ferrari announcing its first pure EV, the Luce, is it still the same Ferrari? Electric motors make raw speed cheap and common. If performance becomes a commodity, the thing that justifies a €300k+ price weakens. Investors were doubtful, and it was one of the reasons the stock got sold down.

Read more

The BUG ETF, which holds a basket of cybersecurity stocks, fell more than 30% from peak to trough over the past year. Then it rallied hard, finishing up almost 20% from where it started.

What’s the deal here?

It started when Claude Mythos was reported to be a super hacker, powerful enough to pose a real risk if it fell into the wrong hands. Anthropic held back its full release. Even Claude Fable, the version made available to the public, was a toned down variant. The takeaway for the market was blunt. AI powered hacking had arrived, and established cybersecurity business models suddenly looked outdated, maybe even incompetent. That triggered a massive sell off, sending heavyweights like CrowdStrike, Palo Alto Networks, Cloudflare, and Okta plunging. The logic was simple. If an AI could hack anything effortlessly, paying for legacy cybersecurity looked like a waste of money.

The narrative flipped entirely within a few months. Mythos didn’t make cybersecurity companies obsolete. It became their best sales pitch. Corporate boards, CEOs, and even the US government, including the Treasury and the Fed, realized that with hackers now wielding AI, baseline defenses were no longer enough. That set off a wave of urgent spending as companies rushed to upgrade to AI powered defense systems.

So here we are. Cybersecurity stocks are hot again. But that also means a lot of the value from this narrative shift has already been captured. CRWD, FTNT and PANW, the top cybersecurity stocks, have rallied between 77% and 102% year to date.

The reality is there are plenty of other cybersecurity stocks out there, and one of them doesn’t compete head on with these three market leaders at all.

Instead, it focuses on vulnerability detection. Think of vulnerability detection as hiring a highly trained security inspector to walk around your house every single day with a clipboard, aggressively checking for weaknesses before a burglar finds them.

In an AI era where hacking grows more powerful by the month, and AI use keeps proliferating, vulnerability detection only becomes more important. Companies and organizations need to find their own weak points before malicious actors using AI find them first.

Read more

Most recently it closed at $77.29. That’s down 53% from our original look, and about 70% off its 2021 all-time high near $250. So what happened? What did we get wrong?

3 things.

1. Librela stopped being a sure thing.

Librela, an injectable antibody for arthritis pain in dogs, was supposed to be the next billion-dollar franchise. Roughly 27 million dogs in the US have osteoarthritis, and Zoetis was only reaching a fraction of them, so the market looked enormous and mostly untapped.

Then adverse-event reports piled up, the FDA issued safety warnings tied to neurological complications, and vets got cautious about prescribing it. Global Librela sales fell 13% in Q1 2026, and management’s own language shifted from talking about accelerating adoption to talking about rebuilding prescribing confidence.

Several shareholder law firms have filed a securities class action alleging Zoetis kept describing its portfolio as resilient even as Librela adoption, Simparica Trio share, and Apoquel share were all weakening faster than it let on publicly. It’s a private class action, not a government investigation, and it’s still in the early stages, so nothing has been proven and no liability has been established. But it adds a risk. If discovery surfaces internal data showing executives knew the trends were worse than they said, that’s a different problem than a normal earnings miss.

Management once hoped Librela’s stabilization plus new treatments would support a return to growth in 2026. That case weakened, partly because rebuilding vet trust is taking longer than modeled, and partly because the real fix, a longer-acting successor drug called Lenivia, is already selling in Europe, the UK, and Canada but isn’t expected to win US approval until 2027. The rest of the pipeline that’s supposed to reignite growth doesn’t meaningfully arrive until around the same time. So 2027 isn’t a vague promise pushed out for optics. It’s tied to when the actual next-generation product and pipeline show up in Zoetis’s biggest market.

2. For the first time in years, Zoetis has real competitors.

For most of the last decade, Zoetis simply had no equal. It holds something like a quarter of the global companion animal drug market, roughly double its nearest rival, built on deep vet relationships and a trio of blockbuster franchises (Simparica for parasites, Apoquel and Cytopoint for dermatology, Librela and Solensia for pain) layered on top of a diagnostics business that kept clinics locked into the Zoetis ecosystem. Nobody had managed to attack that position at scale.

Elanco is the one that finally did, and it wasn’t an accident of timing so much as a deliberate pivot. Elanco used to be a legacy livestock company. After absorbing Bayer’s animal health business, it sold off other divisions, paid down debt, and redirected its pipeline specifically at Zoetis’s two biggest cash cows.

In dermatology, Apoquel’s core patent was set to lapse around 2026, opening the door to generic versions, and Elanco launched Zenrelia, a once-daily pill marketed as comparable or better, right ahead of that cliff at a lower price. Zoetis ended up facing branded and soon-to-be generic competition on the same drug at the same time. In parasiticides, there was no patent expiring (Simparica Trio’s exclusivity runs to 2028), so Elanco won on product design instead. Credelio Quattro added tapeworm coverage that Trio’s own label admits it doesn’t have, worked faster against ticks, and undercut Trio on price.

The overall pet-drug market didn’t grow to absorb the new competitor. Vet visits were already softening, so every dollar Elanco won was a dollar coming directly out of Zoetis’s franchises rather than out of a bigger pie. Two flagship franchises getting squeezed on price and features in the same stretch of quarters, with no market growth to cushion the blow.

Zoetis’s share price has fallen about 49% over the past year while Elanco’s has climbed roughly 87%.

3. Pet owners pulled back.

Household budgets generally have been stretched by a few years of inflation, and vet care inflation has run even hotter than that. Vet service prices have been climbing roughly 5 to 8 percent a year recently, more than double the overall inflation rate. Part of that is ordinary cost inflation; part of it is structural, since a wave of private-equity consolidation has put a large share of US vet clinics, and most specialty practices, under corporate ownership, which several analysts point to as a driver of the faster price increases. The result has been four straight years of falling clinic visits nationally.

What moves is which pet products and services people choose. They’re trading down to cheaper food, doing grooming at home, stretching the time between checkups, and picking the less expensive drug when a cheaper one exists.

Millennial and Gen Z owners, who make up a growing share of the pet-owning population coming out of the pandemic adoption boom, are less attached to the traditional vet-clinic relationship than older generations and more willing to shop around for value. Tighter budgets plus a demographic less loyal to any one brand is exactly the environment in which a cheaper competitor gains share fastest.

It all came to a head on May 7, 2026, when Zoetis cut 2026 guidance (organic growth to just 2% to 5%), and the stock dropped 21.5% in a single session.

Markets Repriced Zoetis And Burry Picks It Up

Read more

“Aschenbrenner may be one. Or he could just be another Cathie Wood who rode the trend at the right time. Either way, time will tell.

Till the next 13F.”

Time told us a lot sooner than expected. Situational Awareness did not even last until the next 13F.

Yesterday the news was plastered all over the internet. Aschenbrenner was hit with margin calls and had to liquidate his public stock positions.

The WSJ then reported that Citadel, Ken Griffin’s hedge fund juggernaut, bought over those positions.

It is easy to point fingers at a hedge fund manager who blew up. But over in South Korea, the exact same thing is happening to mom and pop investors, who took on leverage to bet on the hot AI trend right up until prices became unsustainable. Same mistake, different account size.

What is uncanny is the timing. It coincided with a huge rebound in AI stocks after weeks of heavy selling.

That gives us further validation that the past month of selldown was largely a deleveraging event, not a verdict on the businesses. Investors got complacent, bought with margin, and some were force sold when their collateral fell short.

There are two lessons here.

Lesson 1. Do not take leverage in the first place

First, investors generally should not take leverage when investing. The reason is that as long as you take leverage, there is a chance that you can lose it all. You become fragile.

Here is the math.

Stocks can crash 50% or more. That is not a freak scenario, it is just what markets do.

Now say you take 2x leverage. You have $1m and you borrow another $1m, giving you $2m of exposure. The market drops 50%, and your $2m is worth $1m. Except that $1m belongs to your broker. You walk away with nothing.

And here is the cruel part. All this selling may well have been unwarranted. The AI trend is structural, chip companies are still making big money, hyperscalers are still spending, and there is no sign of a slowdown. You may have every intention of holding and waiting out the rebound.

But leverage does not give you the luxury of time. The market can rebound eventually. Leverage shortens your runway, the margin call arrives first, and your game is over before you get to be right.

Keynes put it best. Markets can stay irrational longer than you can stay solvent.

Without leverage, you stay in the game for as long as the market takes. To make money, survive first.

Lesson 2. If you have taken leverage, honour your cut loss

Set the level, and when the trigger comes, act on it. Do not delay.

If you set a 10% cut loss, honour it when it happens.

If you set a price to sell, honour it when it happens.

We have already established that leverage can take you to zero. So once you are on leverage, your only job is to limit the damage. The moment you do not honour the cut loss, the problem begins.

I know this well because it happened to me more than ten years ago. I was overleveraged on options. I was selling them, not buying them, so my losses were theoretically unlimited.

I had predetermined cut loss levels. I just did not want to take the loss. So I kept rolling the options and sank deeper into the red, until the margin call came and my broker closed out the entire portfolio. The account closed negative. I still had to repay the difference.

You are probably thinking you would honour yours. Everyone thinks that. But you are thinking it now, calm, with nothing at stake and the market closed. That is not the version of you who has to press the button.

The version who presses the button is watching the position gap down, doing the math on what the loss means, and telling himself it will come back by Friday. He is not calm and he is not the one who wrote the plan.

So leverage plus a stop loss is still a bet. You are not just betting on the market. You are betting on how you will behave under stress, and that is the one variable you have never tested.

It was an expensive lesson, and it has served me well ever since. Best not to take leverage at all, because it makes you fragile. If you must, honour your stop loss and act on it. Stop digging the hole deeper.

So is the AI trade back?

Not so fast. A single day rebound, however strong, proves nothing. The encouraging sign is that the deleveraging may be largely done, which reduces the odds of huge downsides. But asking AI stocks to reclaim their previous highs in a short span of time is asking for too much.

Leopold may yet be proven right on AI. He just will not be around to collect. That is the whole point.

But what lifts your wealth fast can bring it down just as quickly. That is the price of volatility.

SpaceX closed at $201.80 on 16 June 2026. It has since tumbled 43% to $115.07, which also puts it 15% below its IPO offer price of $135. A round trip and then some, in a matter of weeks.

It wasn’t just SpaceX. Tesla fell hard after earnings and is now down 29% year to date. The culprit was negative free cash flow of $1.09 billion in Q2 2026, as Tesla spends big on AI, on top of robotaxi and Optimus. Investors have turned sensitive towards heavy capex spenders, and Tesla has just joined that group. The selling is consistent with how the market is treating everyone else in that camp.

With both stocks sliding, Musk has lost his trillionaire status. Over $300 billion gone in about a month. That is more than the entire net worth of second placed Larry Page. Wow. Those numbers.

He is still the world’s richest man. Comfortably.

SpaceX’s fall has dragged the rest of the space sector down with it. A handful of these names are down more than 20%. No surprise there. When the leader falls, the followers follow.

SpaceX listed with enormous fanfare. I stayed out, because I avoid IPOs generally.

The reason is simple. IPOs are difficult to price correctly. There is limited information and no trading history to anchor against, so the market has to discover the price in real time. For a stock as controversial as SpaceX, that discovery process takes even longer before it settles. Better to let the dust settle than to be part of the experiment.

The last month is that argument playing out in public.

To me, SpaceX is a lottery ticket kind of stock. It defies almost every principle of sound fundamental investing.

First, the valuation is always a stretch. Same story with Tesla, priced wildly above its competitors. I accept that a market leader deserves a premium, but the gap suggests something more than that. There is an Elon Musk premium embedded in the share price, and it is substantial.

Second, much of what you are buying sits in the future. Colonising Mars or the Moon, data centres in space, humanoid robots. None of it is commercialised. Yet it is priced today as though it already is. That is not valuation. That is buying into a vision.

So if you want to invest in Musk’s companies, you cannot use the conventional lens. Which is exactly why these stocks are so divisive. On one side, the naysayers listing every fundamental principle being violated. On the other, loyal fans who saw the vision early and believe in it.

Most investors sit somewhere on that spectrum. I sit in between, probably a little closer to the fundamentalist end.

But I also think a small position in SpaceX can make sense. If it fails, it costs the portfolio very little. If it delivers 10x or 100x, even a small position grows into something meaningful and actually moves the needle. That is the lottery ticket logic, and it only works if the ticket stays small.

I am still not buying though.

Not yet, anyway. The share price is still sliding, and there is no reward for catching it mid fall.

What I want to see first is the selling stop, then the price form a horizontal base and hold it for a while. After that, I want to see genuine upward movement off that base. That is the point where betting on Musk reclaiming his trillionaire status starts to look like a calculated bet rather than a hopeful one.

Until then, I can wait.

Here is the strange part. Alphabet’s results were not bad. They were stellar.

Revenue grew 24%. The standout was Google Cloud, where growth accelerated to 82% from just over 30% in the previous quarter. That is a huge jump off a much bigger base. Operating margin also expanded from 20.7% to 35.6%. Alphabet is monetising its AI tools and compute, and it has reached scale.

So why was the stock beaten up?

Capital expenditure.

Last quarter, Alphabet guided 2026 capex to $180B to $190B. This quarter, management raised it by another $15B to $195B to $205B. AI spending is not peaking. It is not even slowing. It is accelerating.

And that spending produced the chart investors have been dreading. Alphabet’s free cash flow has turned negative for the first time since it listed in 2004.

It gets worse for shareholders. Alphabet has stopped its share buyback because it needs the cash for AI, while it continues to issue stock to raise funds and to reward employees with options. Share count is up 1.2% in six months. That is dilution, and it is new. Previously the buybacks more than offset the issuance. On top of that, Alphabet took on $19.1B of mandatory convertible preferred shares and $56.2B of debt. And the fundraising is not ending soon. Roughly $106B of securities are expected to be issued over the next six months.

So the question investors are asking is simple. Will Alphabet earn this money back, and will shareholders see any of it?

The encouraging sign is the backlog. Google Cloud’s backlog grew $51.9B to $513.9B, roughly 5.2 years of cloud revenue at the current run rate. Customers are exceeding their contracted commitments by more than 50%. Demand is real. Supply is not enough. That is exactly why Alphabet is spending so hard. Fall behind, and the revenue does not wait. It goes to whoever can serve it.

But real demand and full capacity do not guarantee a good return on that capital. Alphabet may fill every server and still earn a poor return on what it paid to build them. That worry spread across the group, which is why Amazon and Microsoft slipped 5% and 2% respectively.

Alphabet’s Loss Is AI’s Gain

For the past month, the market has been whispering that the AI trade is done. Memory stocks took the brunt of it, on the belief that chip prices had peaked. Their share prices fell hard.

Alphabet’s guidance says otherwise. Higher capex and a shortage of AI cloud compute mean the spending cycle is still running. And that is music to the ears of everyone selling into it.

So while the Magnificent Seven fell 5% and QQQ fell 2%, memory stocks rose.

One day does not make a trend. But look closer and they have been climbing for three straight days. A relief rally is taking shape.

It helps that some AI names are surprising to the upside. Super Micro reported gross margins double its earlier target and a record $60B order book. The stock jumped more than 20%.

Next week is the real test. Microsoft, Meta, Amazon and Apple all report. The first three are the big capex spenders, and any further increase confirms the trade is intact. Watch the suppliers too. SK Hynix, Seagate, Corning, KLA and Arm are all reporting. Strong numbers there would tell you the AI build out is very much alive.

That shattered the illusion that outspending everyone else was the only path to AI supremacy.

Nvidia took the hardest hit, plunging nearly 17% intraday and erasing roughly $589 billion in market value. It remains the largest single-day loss of market capitalisation for any company in history. Other chipmakers bled too. Broadcom dropped 17% and Taiwan Semiconductor Manufacturing Company (TSMC) fell 13%.

The good news is that AI stocks recovered, and then some. They surged to record highs as Western models kept improving and held their lead, quashing fears that cheaper Chinese models could catch up. For a while, that justified the massive capital expenditure the West had committed to.

Kimi K3 Reruns The Playbook

Last week, the DeepSeek moment was relived. Moonshot AI, another Chinese startup, released its Kimi K3 model, and it can rival the frontier models from Claude and ChatGPT. On artificialanalysis.ai’s Intelligence Index, K3 scored 57, placing third behind GPT-5.6 Sol at 59 and Claude Fable 5 at 60. That is not far off.

The alarming part is the price. K3 runs at about $0.95 per task, cheaper than GPT-5.6 Sol at $1.04 and far below Claude Fable 5 at $2.75.

Moonshot didn’t stop there. Over a single 48-hour run, Kimi K3 autonomously executed a full semiconductor chip design flow, from RTL to tape-out simulation, for a 45-nanometer chip. The kicker is that it did this using only open-source tools, completely bypassing the expensive proprietary software licences that Cadence and Synopsys sell to chipmakers.

Cadence and Synopsys stocks tanked 9.47% and 7.85% respectively last Friday.

So why this particular demo? I don’t think Moonshot is really aiming at Cadence and Synopsys. Most AI models are judged on how well they answer a question, write a Python script, or pass an exam. Chip design is a different beast. It is a gruelling, multi-step engineering process. By completing all of it autonomously over 48 hours with no human intervention, Moonshot proved K3 isn’t just a chatbot. It is a long-horizon agent that can run a complex, multi-day engineering project from start to finish.

Impressive as it is, this is still far from the end for the two EDA giants. The demo leaned on the Nangate 45-nanometer Open Cell Library. In the semiconductor world, 45nm is ancient history. It is the node that built consumer chips back in 2007 and 2008.

Modern AI accelerators are fabricated on cutting-edge 3nm or 2nm nodes. At 45nm, the physics of chip design are relatively straightforward and forgiving. At 3nm, engineers wrestle with extreme ultraviolet (EUV) lithography, quantum tunnelling, and severe thermal density. Designing a 45nm chip autonomously does not prove an AI can handle the staggering multi-physics complexity of modern frontier silicon. At 3nm and 2nm, Cadence and Synopsys software is a necessity, not an option.

K3 Still Needs A Mountain of Compute

There is another catch. Cheap to use does not mean cheap to run. Kimi K3 still devours computing power, and this is not something you fire up on your home PC, however beefy. It needs at least 1.4 TB of video RAM (VRAM) just to load. For perspective, a single high-end Nvidia H100 costs tens of thousands of dollars and holds only 80GB.

So you are not shopping for a graphics card. You are shopping for a rack. NVIDIA’s flagship GB200 NVL72 packs 72 Blackwell GPUs into one liquid-cooled rack that costs around $3 million, and serving K3 at any real scale means one or two of those.

Which brings us back to the question in the title. The market treated K3 the way it treated DeepSeek, selling first and asking questions later. But both big fears, that EDA software gets disrupted and that AI stops needing so much hardware, look overdone once you read the fine print.

So no. Kimi K3 will not reduce the need for computing power. If anything, the opposite. This is where Satya Nadella’s point about Jevons paradox rings true. Because K3 is cheaper, more people can afford to run it. And because it still demands enormous resources, cheaper access means more servers get spun up, not fewer.

IBM is often seen as a dinosaur stock, irrelevant in the new era. It used to be a foremost tech name, but not anymore. The Magnificent 7 has taken over, and IBM is a “mere” $200 billion market cap company today.

So why does it matter what happened to IBM? Because it’s still tied to tech, specifically what’s going on with software and AI.

IBM CEO Arvind Krishna noted that the global rush to build out AI infrastructure has created severe supply shortages and driven up the cost of memory chips, servers, and storage across the industry.

In the final weeks of June 2026, IBM saw corporate clients aggressively shift their capex budgets. Customers rushed to buy up physical, supply-constrained hardware to get ahead of anticipated price increases.

Krishna bluntly admitted in his letter to investors that IBM “did not adapt and move quickly enough” to this sudden shift, stating “this quarter we faltered.”

This rapid pivot toward AI hardware by enterprise clients drained budgets away from IBM’s core, higher-margin businesses.

Infrastructure fell 7%. Clients pulled spending away from IBM’s flagship mainframes (the IBM Z systems) and related transaction processing software. IBM had originally projected low-single-digit declines for this segment, but the capex shift made the drop far more severe.

Software rose 5%. Still IBM’s strongest segment, but growth slowed sharply from the 11% growth seen the previous quarter, as budgets were diverted elsewhere.

Consulting was flat, with no meaningful growth.

For those who don’t know, IBM still sells hardware, mainly to customers who want to control their own servers, especially in closed private systems. On the software side, IBM acquired Red Hat in 2019, which helps enterprises manage both private and public cloud. Both businesses remain necessary, but both are still AI vulnerable.

So the rapid pivot to AI infrastructure dented confidence in software stocks and, on the flip side, boosted AI stocks for the day.

Yes, a day.

Because the following day, we saw heavy selldowns in AI stocks once more, and sector rotation continued.

Look at the 10 biggest losers in the S&P 500 on 16 Jul 2026 below. They are all AI related, except for Robinhood and Moderna.

Now look at the top 10 gainers on 16 Jul 2026. They are boring, old school companies such as FedEx, Cintas, and Domino’s Pizza.

So sector rotation is still ongoing. AI stocks continue to slide as excess leverage in the sector still needs time to clear. The IBM one day drop looked significant, but it was really just a relief rally for AI stocks, and only for a day.

Now let’s look at software stocks in the S&P 500 on the same day. They were up too. So no, software isn’t dead, it’s finally showing signs of recovery.

Some may ask, then, is it a good time to buy IBM after the big drop, given capital is rotating to old school companies? Let’s examine further.

Read more

Take the Iran war. Oil prices rose and the energy sector did well as a result. For years the AI trade has lifted technology, and just this past month we saw financials and healthcare take their turn. So there is no single sector that wins forever. But there is always at least one sector that is working.

That is the basis for a rotation strategy. You move capital into the sectors that are doing well, or that you expect to do well. The method varies. It can be driven by macro, by fundamentals, or by momentum.

Sector rotation is the established idea. Factor rotation is the newer cousin, and many investors are not even familiar with factors to begin with.

Factors are characteristics of stocks with evidence of outperformance. Value, quality, momentum, size and low volatility are the foremost ones. Factors rotate too, and none of them wins all the time. Value investors know this better than anyone. Value languished for years before roaring back over the last two.

So if the two strategies sound alike, do they behave alike? And can either of them beat the S&P 500?

Helpfully, there are sector and factor rotation ETFs, which makes the comparison straightforward, both against each other and against the index.

Here is the punchline. The sector rotation ETF did not just lose to the factor rotation ETF. It could not even beat the S&P 500 itself. The factor rotation ETF beat both.

But there are nuances worth knowing. Let’s break it down.

Read more

A quick check told me both brands are owned by Authentic Brands Group. In fact ABG owns a long list of once-iconic names that have simply gone less popular, Reebok and Guess among them.

ABG buys fading brands and keeps them running, probably at a steep discount because nobody else wants them. It is cigar butt investing, the term Buffett used for picking up a business cheap enough that there is still one good puff left in it. If venture capital chases young, dynamic startups, Authentic sits at the other end of the spectrum, hunting mature companies with a fading outlook.

We are not here to talk about Authentic Brands as an investment idea though. It is a private company.

But a company applying a similar playbook to software has just gone public. Bending Spoons, ticker BSP, priced its IPO on 30 June 2026 at $29 a share, above the initial $26 to $28 range, valuing the business at roughly $19 billion. It popped 40% on debut day and closed at $40.50. As I write this it trades closer to $36, putting the market cap near $27 billion.

The name itself is a nod to The Matrix. Founder Luca Ferrari, one of four Italian co-founders, was inspired by the “there is no spoon” scene. Not essential point, just in case you wonder about the name.

Bending Spoons operates like a private equity firm with a software focus. It buys software companies that already have a large, loyal user base but have stopped growing, which means they can be picked up at far lower price multiples than hot startups chasing distribution and customers for the first time. Same logic as Authentic buying fashion labels that fell out of vogue, just applied to code instead of clothing.

The portfolio includes some genuinely iconic names. Evernote. Vimeo. WeTransfer. Eventbrite. AOL. And more.

Here is where the comparison with Authentic breaks down. Once Bending Spoons acquires a company, it typically follows with a brutal round of layoffs, sometimes cutting more than half the workforce. When it acquired WeTransfer in 2024, roughly 75% of the staff were let go within months.

The acquired software then gets migrated onto Bending Spoons’ own centralised operating platform, run by a tight, concentrated engineering team in Italy that the company calls “Spooners.” Buy cheap, strip costs, centralise operations. That is a textbook leveraged buyout or private equity approach, just wearing a software jacket.

Bending Spoons does not stop at cutting costs though. It pushes hard on the revenue side too. One lever is limiting what free users can do. Shortly after acquiring Evernote, it capped free accounts at 50 notes, effectively forcing power users to pay. It has also raised prices significantly across its apps.

The bet is that its acquisition targets sit on a highly inelastic user base, people who are annoyed enough to complain but not annoyed enough to leave. The numbers back this up. Between December 2023 and March 2026, total paying subscribers across all apps grew from 3 million to 9 million.

Financials

Bending Spoons is growing revenue fast, which is not surprising given that acquiring and folding in companies is one of the quickest ways to boost the top line. The company has reported 95% year on year revenue growth in 2025.

Interestingly, the company stayed profitable through most of this run, aside from a loss in 2025. That loss was driven by accounting and acquisition financing effects rather than any deterioration in the underlying businesses. An unusual $166 million tax expense arose from transferring certain acquired businesses to Italy, and interest expense jumped from $32.6 million to $142.6 million, up 337%. Bending Spoons spent about $1.92 billion on acquisitions in 2025, including Brightcove, Komoot, Harvest, MileIQ and Vimeo, funded with significant borrowing.

That last point is the real risk in the model. As Bending Spoons scales, debt interest, restructuring costs and acquisition accounting can eat a surprisingly large share of operating profit. Free cash flow has stayed positive through this though, and without those one off accounting losses, the company was back to profitability in Q1 2026, reporting net income of roughly $28 million.

I have not really seen an operating model quite like this before. Other private equity shops buy software companies too, but I have not seen the same depth of integration and centralisation that Bending Spoons runs.

I originally assumed legacy software was a tough business to be in, especially with the risk of obsolescence rising as AI reshapes how people work. I also assumed Authentic Brands had an easier road, since fashion can always find a retro revival and does not really face obsolescence the same way. Bending Spoons is growing revenue, cash flow and profit despite the AI wave, which does not fit that assumption.

But there is a wrinkle worth separating out. Bending Spoons’ own IPO filing shows that AI authored code jumped from under 10% of its engineering output in Q1 2025 to over 90% by the end of Q1 2026, with around 70% written by AI alone. That tells you AI is making Bending Spoons more efficient at running its playbook, cutting headcount further and squeezing more margin from each acquisition. It does not tell you the acquired products themselves are immune to AI driven disruption. Those are two different claims, and it is worth watching whether Evernote or WeTransfer eventually lose users to AI native alternatives even as Bending Spoons gets leaner at extracting profit from them in the meantime.

I also assumed B2C software would not be very sticky, since switching to a competitor costs little compared to enterprise software, where switching is often too painful a project to even start. But raising prices did not push users away here, it grew revenue further. The real trick is buying software with a genuinely loyal user base, not just a large one.

I will not buy an IPO stock. That is a personal principle of mine, since market pricing has not stabilised yet. And even setting that aside, as interesting and promising as the business model is, it does not scream must own stock to me yet. But I will be watching it.

AI stocks will crash. It’s just a matter of time, though no one knows exactly when, or how much higher they’ll climb before gravity catches up. That makes them poor long-term holdings.

On the other hand, there are stocks that transcend time, what I call Lindy stocks. They’re often boring and slow growing, but they stand the test of time. No example beats McDonald’s. It’s a place where parents bring their kids, and one day those kids bring their own. It’s intergenerational and never goes out of style. Compare that to a fashion label tagged to an older generation and shunned by the next. This is what makes McDonald’s the kind of stock that lasts, because it stays relevant no matter who’s ordering.

McDonald’s is Like a REIT

Read more

Start with the memory group, the biggest rally among AI related names. All were down more than 10% in just two sessions.

Semiconductor equipment makers fell just as hard.

Photonics too, got hammered.

There was nowhere to hide. These stocks go up together. They come down together.

So what’s driving the fear this time?

The apparent trigger is AI overcapacity worries. In a surprising move, Meta wants to play hyperscaler like Amazon, Microsoft, and Google. It was reportedly looking to sell off excess AI compute. The market read that one way, too much AI infrastructure is being built.

But here’s the twist. Just days before Meta’s announcement, the Financial Times reported that Google had capped Meta’s access to its Gemini AI models, because Google couldn’t meet the demand coming from Meta. Both companies declined to comment.

So which is it?

Excess capacity, or a shortage?

Did Google catch wind of Meta’s plan to lease out its own compute and move to box out a new competitor?

Or are Meta’s models falling short, pushing it to lean on Google’s and others, while offloading its own capacity to defray costs?

We don’t actually know. And honestly, investors don’t care. These stocks have rallied so hard that they are now hypersensitive to any whiff of weaker demand. The moment demand looks shaky, the selloff follows.

Making things worse, investors have been piling on leverage.

U.S. margin debt, what investors borrow from brokerages to buy securities, jumped 54% year on year to a record $1.4 trillion in May, according to Finra data.

Leveraged ETFs tell the same story. Assets have been swelling, including single stock leveraged ETFs tracking individual AI names.

My take is that this looks more like a deleveraging exercise than a crash. Investors are trimming AI exposure, not abandoning the thesis. I don’t think this is 2000 all over again, yet.

Once the deleveraging plays out, these stocks should stabilize. And frankly, after nine straight weeks of gains and multi bagger moves packed into a short stretch, a breather was overdue. It’s just that the breather can look like a sharp, ugly pullback, the kind most investors aren’t prepared for.

The key reason I’m not sounding the alarm is that fundamentals still look solid and supply is still tight.

There’s no confirmation on why Meta is suddenly offering out cloud capacity, or whether “excess capacity” is even real. No other hyperscaler has reported anything similar.

But this is a spot worth watching. Hyperscalers have poured billions into building capacity. If we’re at the point where excess capacity is genuinely emerging, that signals overbuilding, which means demand for AI hardware could shrink, unwinding the entire AI stack narrative. That’s the point where the bubble actually bursts.

For now, though, there’s no evidence of that.

My instinct? If most people are already calling it a bubble, it probably isn’t one yet. When fear is widespread, buyers step back and the market stabilizes rather than spirals. That said, a pullback or a period of consolidation wouldn’t surprise me. The volatility needs to work itself out.

So if not AI, where do investors look?

Sector Rotation Has Quietly Begun

About two weeks ago, something shifted. Healthcare and utilities started gaining ground while AI names sold off. I can’t point to a fundamental catalyst as sector rotation just happens. But if you’re asking me to pick one sector worth paying attention to, it’s healthcare.

Here’s some context. Over the last month, the S&P 500 was down 2.9%. The Magnificent 7 dragged it lower, punished for their massive capital expenditure on AI data centers. But beneath the surface, more sectors delivered positive returns than negative ones. Market breadth like that is healthy as it means the selling is concentrated, not broad.

And topping the sector returns was healthcare. That’s a new development. Over the past five years, healthcare was a middling performer, weighed down by stocks like UnitedHealthcare, which got absolutely battered.

Zooming in on individual names, the big healthcare stocks are doing the heavy lifting. Eli Lilly (LLY) and Johnson & Johnson (JNJ) are both up more than 10% over the past month and have hit new all-time highs.

Even the previously beaten-down names are bouncing back convincingly. UnitedHealth (UNH) has nearly doubled from its August 2025 low and is now in a Stage 2 bull phase.

Read more

After reporting earnings on Wednesday evening, Micron (MU) surged nearly 16% on Thursday. Not 16% from some beaten-down base. This was a stock that had already hit an all-time high of $1,134 just days earlier. And it still popped 16%.

Let’s break down what happened.

Shocking Numbers

Micron posted fiscal Q3 2026 revenue of $41.46 billion. That’s not a typo. Just last year in the same quarter, they did $9.3 billion. In other words, revenue more than quadrupled year-over-year.

To put it another way, Micron grew by nearly $18 billion in a single quarter compared to the prior quarter’s $23.86 billion. The beat was across the board. Revenue came in 16% above analyst estimates, and adjusted EPS of $25.11 beat forecasts by over 22%.

Data center revenue hit $25 billion in one quarter, an annualized run rate of over $100 billion. Enterprise SSD revenue alone exceeded $5 billion, more than doubling sequentially.

And if you thought Q3 was good, management’s Q4 guidance was something else entirely.

Analysts expected Q4 revenue of around $43 billion. Micron guided to $50 billion.That’s nearly a $7 billion beat on guidance.

Free cash flow in Q4 is expected to exceed $30 billion. HBM3E and HBM4 are fully booked through calendar 2027, with demand already extending into 2028. They also locked in $22 billion in strategic customer agreements, including $18 billion in upfront cash deposits. Customers are essentially pre-paying to secure memory supply.

When your customers are handing you cash in advance just to get in line, you’re not in a commodity business anymore. You’re controlling supply and dictating prices.

But Wait, Didn’t Micron Just Drop 12%?

Here’s the part that makes this story even more interesting.

Just 48 hours before this blowout results, Micron was down sharply. It was part of a brutal semiconductor selloff on June 23 tied to AI valuation fears and a crisis in South Korean leveraged ETFs tracking memory chip peers. Sandisk fell 12.5% that same day. The mood was ugly.

And then earnings hit.

That’s the brutal reality of momentum trading in AI stocks. One day you’re down 12%, the next you’re up 16%. The swings are violent in both directions, and only those with the stomach and the discipline can ride them.

Micron’s results didn’t just move Micron. They sent a wave through the entire semiconductor ecosystem.

SanDisk (SNDK) surged over 20%, building on its position as the top-performing S&P 500 stock of 2026 with a staggering 748% year-to-date return. The logic is straightforward. If Micron’s NAND and HBM supply is fully booked, the same shortage is lifting SanDisk’s pricing power.

The rest of the AI value chain were up too. Applied Materials (AMAT) rose around 13% after launching six new chipmaking systems designed for DRAM production and advanced AI chip packaging. More Micron orders means more machines needed to make the chips. Teradyne (TER) and Corning (GLW) also had strong days, up around 10 to 11% respectively. You might wonder what a fiber optics company like Corning has to do with memory chips. The answer is that as AI data centers scale, everything around them scales with it, from testing equipment to optical connectivity. These are the picks-and-shovels plays of the AI build-out.

Micron’s Gain Is Apple’s Pain

The same memory shortage that’s minting money for Micron is creating a serious headache for Apple (AAPL).

On Thursday, Apple announced price hikes across its MacBook and iPad lineups. The MacBook Air jumped from $1,099 to $1,299. The MacBook Pro 14-inch went from $1,699 to $1,999. The iPad Air rose from $599 to $749. CEO Tim Cook described the memory supply squeeze as a “hundred-year flood,” with DRAM and NAND prices surging as the entire industry prioritises AI data centre chips over consumer electronics.

Apple’s stock fell over 6% on the day, its worst single-day drop since April 2025.

This illustrates a zero-sum dynamic playing out in real time. The AI infrastructure boom is diverting semiconductor supply away from consumer devices. What fills Micron’s order books reduces Apple’s margin.

And it’s not just Apple.

The MAGS ETF, tracking the Magnificent Seven, fell about 6% over the past week.

The hyperscalers and AI capex spenders, Microsoft, Google, Amazon, Meta, have all seen their share prices drift lower about 6% this week. The market is effectively penalising the buyers of AI infrastructure while rewarding the sellers.

Think about what that means. The companies spending hundreds of billions on AI data centres are seeing their valuations compress, while the companies supplying the memory, chips, and equipment are seeing their valuations soar.

It’s a bifurcated market. AI capex spenders on one side, AI capex beneficiaries on the other.

The Investor’s Dilemma

If you’re sitting on a heavy allocation to the Magnificent Seven right now, you’re facing a genuine dilemma.

Selling them to buy into Micron, SanDisk, and semiconductor stocks feels like chasing a runaway train. These stocks are already up hundreds of percent. The risk of being the last buyer is real.

But staying put feels punishing too, watching your portfolio drift sideways or lower while AI infrastructure stocks surge higher week after week.

Here’s how I think about it.

If you’re a short-term momentum trader, go with the leaders. The AI beneficiaries, memory makers, chip equipment companies, data centre infrastructure plays, have the momentum. Follow it, but have your sell rules ready. These things can reverse as fast as they run.

If you’re a long-term investor, don’t abandon your long-term winners for short-term excitement. Microsoft, Alphabet and Amazon are still building the AI era’s critical infrastructure. Their capex today is an investment in their competitive moat tomorrow. Yes, the market is punishing them now for spending. But those who stay patient may be rewarded when that spending translates into earnings.

The worst thing you can do is panic out of quality compounders at a cyclical low point, only to chase the hot momentum trade at the peak.

Just as we thought AI stocks had run too far, Micron delivered a quarter that made new highs look cheap. That’s the thing about momentum. It doesn’t care about your valuations, your gut feeling, or when you think it should stop.

The AI party is far from over. But make sure you know which kind of investor you are before you decide what to do.