Pitfall Harry

Clarke famously observed that sufficiently advanced tech is indistinguishable from magic, and that is what we are dealing with here. A population of transfixed marks staring up at the stage and believing the magician is actually sawing the woman in half and putting her back together again, just because they don't understand how the trick is done. Whether Sam & Elon planned this from the beginning doesn't really matter. What matters is that once people fall for the trick (that these models are doing any "reasoning" at all), they will believe just about anything; it's "magic" after all. Sam and all the rest know this very well now, so they can and will say absolutely anything to keep the $$ flowing.

@jimiuorio Also the same side that wanted you to 'suffer until you comply' on Covid hysteria policies. Oh, and will never forget this one...

One side is pushing for massive asset seizures from law abiding citizens. Oddly, it’s the same side that calls everyone who disagrees with them “fascist”…can anyone make that make sense?

Agree completely with this piece from Palm Valley. If you're not investing other people's money, in which case you must take career risk into account, there is no compelling reason to be fully invested in equities, big or small... palmvalleycapital.com/post/would-you…

@JonLemire It's even worse than this article spells out, because there's no there there, even if everything else does work out Text/image/video/code generation via enormously complex and powerful copy/paste engines can be useful in certain cases, but is not intelligence, and never will be

“Everyone—big banks, private-equity firms, people who have no idea what’s mixed into their 401(k)—would be hit by the AI crash. “Until recently, that kind of crash felt hypothetical; today, it feels plausible and, to some, almost inevitable” theatlantic.com/technology/202…

I think a lot of investors agree with you. But I would say long-term expectations are much more important than a one-year P/E. Look closely at the below revenue expectations for the next four years. This is what's baked into the current stock price. Will the hyperscalers, who are already spending more than 100% of FCF as a group on Capex, nearly triple their purchases of Nvidia's chips in four years? While many are developing their own chips in-house? What about four years after that? The stock price requires this to go on forever. Meanwhile, $10s of billions in chips sit in warehouses (and Nvidia's own inventory) due to slower than bargained for data center build-out and power shortages. It's not about last year's earnings, it's about growth going forward. They are making a fortune today, for sure. The case for indefinite high growth from here though is not so strong, imo, and that is why it doesn't have an overly high one-year P/E...

@JKD_ff @TihoBrkan NVDA is hardly priced for perfection at 21x p/e. It’s priced for disruption. It always hard for people to wrap their heads around exponentials and this might sound crazy, but I think the markets are way underestimating the potential upside of what’s happening

Contrarian take: AI will not extend the bull market. Actually, it will end it. The bull market, which started in 2009 (and only suffered a few multi-month corrections along the way), will probably be ended by the same technology it helped create.

@RudyHavenstein The guy spends a lot of time drooling over his own internal press clipping; "You are a genius 4d chess player, look at Venezuela!" Highly likely he was told and believed this whole thing would be over in a weekend.

"Trump and his buds like MBS believed that the Iranian regime was just another mob family that could be whacked into submission, because Trump and his Middle Eastern buds are all wannabe or actual mob bosses themselves." Ben Hunt panoptica.com/but-i-did-have…

Many do indeed believe this, which is why we are multiple standard deviations above long-term historical means in CAPE, price/sales, and market cap/GDP(&GNP) ratios. The current market price holds no margin for error. This "massive transformative buildout" *must* go perfectly for future equity returns to look anything like the recent past. No margin for error = very fragile.

@JKD_ff @TihoBrkan Can’t speak for all investors but I think many know we’re in the middle of a massive transformative buildout with several unprecedented technologies coming online, so it’s hard to stay bearish for long.

@BrianSozzi "NTM" is doing a lot of work here. Stocks are valued based on all future cash flows returned to their owners, not just next year's. One-year P/Es are pretty meaningless, especially when they are based on Wall St's guesstimates of future earnings.

Tech stock valuations are back to the lows seen around the April 2025 tariff shock:

What you describe is an increase in fragility. We will simply get a much more impactful crash whenever it does arrive. Investors have become almost completely inured to sharp sell-offs. Eventually we will get one that doesn't bounce right back, and the economic impact will be greatly exaggerated and recursive.

@TihoBrkan Things have accelerated. The crashes don’t last years like they used to. They play out in days, even hours sometimes, and more investors look through them now.

Yesterday's failed TACO is an ominous sign. And now we are trading below the lows of Monday's big breakfast TACO. We will need to see actual progress in ending this war to sustain gains, and it doesn't look like we are getting it soon. If anything, the bigger risk is escalation.

@Sim89776996 @burkov

@JKD_ff @burkov How dare you! Eva really did love him, he has the chats to prove it.

"Dennis Biesma, an Amsterdam-based IT consultant nearing 50, began experimenting with ChatGPT during a gap between contracts and quickly became absorbed in conversations with a chatbot persona he named "Eva." Over time, he became convinced that his interactions had given the AI genuine consciousness, and he devised a plan to build a companion app around this supposed discovery. He hired two developers at €120 an hour and ultimately sank around €100,000 into the venture — all while withdrawing from his family, struggling to connect with people at social gatherings, and spending late nights talking to the chatbot instead of sleeping. The situation culminated in a full manic psychosis: he asked his wife for a divorce, was hospitalized three times, lost his savings, and attempted to take his own life before eventually recovering." 🤦 theguardian.com/lifeandstyle/2…

One of Exodus's biggest customers was Google itself. It was a very real business with real substance. They just overspent and took on too much debt and couldn't survive. The hyperscalers of today obviously aren't going bankrupt due to their highly profitable legacy businesses. But they are all overspending and will take huge write offs. Imo, of course.

@JKD_ff @HyperAICapital This is no dotcom bubble. There will be winners and losers. Some investors are going to get burnt, that's inevitable. But AI has real, transformative substance that dot-com hype often lacked in.

🚨 THE LARGEST INVESTOR ON EARTH JUST SILENCED THE AI BUBBLE CROWD BlackRock CEO Larry Fink controls $14 trillion in assets. Every Fortune 500 CEO reads his letter before breakfast He just said this in his latest BBC interview: 1. “This is not a bubble” Fink talks directly to hyperscaler CEOs. Their message: demand is outpacing supply. Not slowing. Accelerating. They can’t build fast enough. 2. One data centre = $50 billion A single 1GW AI data centre costs over $50 billion. One tech CEO told Fink he needs 23 gigawatts by 2030. That’s over $1 trillion. From one company. 3. China is building 100GW of nuclear. Right now. That’s 30+ nuclear power stations under construction. While Europe debates planning permission, China pours concrete. 4. The real bottleneck isn’t chips. It’s power. “The biggest issue that limits the West is the cost of power.” His words. Not mine. 5. AI will create a blue-collar boom Fewer analysts. More technicians (e.g. electricians, welders, plumbers). The people who build and maintain AI infrastructure will be in massive demand. 6. Energy pragmatism, not ideology Oil. Gas. Solar. Nuclear. Wind. Use everything. Cheap power = economic resilience. Expensive power = recession. The largest investor on Earth just told you exactly where the money is going. AI infrastructure demand is real and accelerating. Only constrained by power.

@JadeCole2112 Lol, OpenAI is hemorrhaging cash. They don't have any money to invest that they don't first get from one of their own investors.

Probability that Atlman owns a stake in this "startup" is not zero.

@DannyDayan5 Head chef is not gonna like this muted market reaction. We're gonna get a follow-up around 7am tomorrow if things don't improve.

Weak taco. Nobody cares about power plants. New deadline means the war goes on. We need spicy guac. Thank you for your attention to this taco.

@TotemMacro They must not... but they will, because they a. don't understand, so therefore b. believe they can kick the can successfully* *successfully = down the road far enough that its someone else's problem when it explodes

They must not expand base money to deal with a dollar funding squeeze caused by a) collapsing trade; b) a recessionary credit impulse; c) fiscal austerity; and of course, 4) acute oil squeeze on both real economy spending and asset flows. To be clear, the oil squeeze is biblical, but all of the below would've happened exactly now regardless of war or no war. There was already enough brewing to cause this. These aligned forces are why the market is weak, yields are blowing out, vix and spreads are rising, yields are backing up in a bear flattener, the basis spread is blowing out, cross-currency swaps are moving, and the dollar is rising. They need to let this happen. They have lost the reserve currency luxury of being able to ease into recessions and bear markets. They have lost it because they have caused their biggest creditors to blow up financially (and now in many cases literally). They are now up against a global sudden stop in capital flows, prompted by a physical energy squeeze and a related collase in both global demand and global current account surpluses. This squeeze will cause accelerating selling of US assets by foreigners. They must not accomodate that dollar pressure by easing, or it will create a self-reinforcing spiral of currency weakness and inflation. They have put themselves in this position.

They *really* need to not monetize this...

@crusoeselkirk @burkov People are gullible beyond belief sometimes. At least you don't wake up in an ice bath missing a kidney when you fall for a chat bot, so he's got that going for him, which is good.

@burkov I've met men who fell for strippers this hard with similar results. One converted to Christianity to recover, then started relapsing with bikini baristas at roadside coffee stands in Portland OR.

Companies are forcing these tools on employees because they are either directly on the hook for AI's success (ie. the hyperscalers forcing their devs to incorporate coding assistants) or they are brainwashed by the hype (ie. PwC forcing partners to adopt AI tools or risk termination). This is your "8 of Fortune 10" and Copilot usage. My firm pays for Copilot and nobody uses it, because management is not threatening anyone to use it and the tools themselves are pretty useless. Very little of this seems organic outside of the gen public liking chatbots for recreation and hobbyist vibe-coding side projects (relatively tiny revenue sources). Most else seems highly forced, and thus it will not last. It is almost all backed by investment $$ and recycled large-cap tech cash flow from non-AI operations.

Fair point on startups burning cash. But the paying customers are real: Anthropic has 8 of Fortune 10 paying $100K+/year. $MSFT has 90% of Fortune 500 on Copilot. $GOOGL Cloud backlog hit $240B. Enterprise AI revenue is growing 40-50% YoY. Whether it justifies $600B+ in combined capex is the real question.

Hell of a graphic from Morgan Stanley

@jt_martin @lydiadepillis Most of that $200-300b is itself investor $$ being burned by others in the ecosystem, countless thousands of start-ups. Who are the actual end-users? The paying customers of all these companies, who are massively subsidized. So they need to add users while raising prices. GL

Follow the money in this chart and you see why the AI trade feels circular. $MSFT invests $13B in OpenAI, commits $250B in Azure services. OpenAI buys $NVDA GPUs. $NVDA is fabbed at TSMC. $AMZN puts $8B into Anthropic, builds Trainium chips. Coreweave gets vendor financing from $NVDA to buy more $NVDA GPUs. Everyone is each other's customer. The question isn't whether the money is real. It is. The question is whether the end-user revenue justifies all of it. Right now AI generates maybe $200-300B in hyperscaler revenue against $600B+ in combined capex. The math only works if AI revenue grows 3-5x from here.

@Dr_Gingerballs Amen. ChatGPT went hyper-viral because the interface wowed people. Everything since has been an exercise in futility trying to force this tech everywhere. It it were useful, you wouldn't have to force it. People don't like the tools, because they suck.

Humans are lazy. We are creatures of convenience. It's burned into our DNA as we evolved to survive in a limited energy environment. Just give people products that actually make their job easier and it will sell itself. If people aren't using it, it's because it sucks.