AI Bubble or Correction? The $6 Trillion Selloff (Kobeissi Letter)
The Kobeissi Letter argues that the $6 trillion US stock selloff since the S&P 500's October 6th record is a normal correction, not a popped AI bubble. Bob and Alice unpack its case and test it against the skeptics. Not investment advice. In this episode: - Big Tech AI capex versus free cash flow - The midterms, the Fed and Nvidia - Dot-com lessons and the skeptics' case Sources: The Kobeissi Letter (@KobeissiLetter); Jason Fernando; James Chen; Mary Hall; David Cahn; Dave Smith; Peter Dizikes Big AI ideas, made clear for builders. luongnv.com/podcast
Listen on Build Wiz AI Show Download MP3 Transcript
Based on -$6 Trillion In 45 Days: Is This A "Bubble" Or An Opportunity? by The Kobeissi Letter (@KobeissiLetter)
Transcript
Bob: 6 trillion dollars of US stock market value, wiped out in about six weeks.
Alice: So is this the AI bubble popping? Or is the panic itself the story?
Bob: Welcome to Build Wiz AI Show. Big AI ideas, made clear for builders. I'm Bob, and today I'm the one asking the questions.
Alice: And I'm Alice. Today we're unpacking a brand-new X article by The Kobeissi Letter, a market newsletter. It breaks down why US stocks lost about 6 trillion dollars in 45 days, and asks: is this a "Bubble" or an opportunity?
Bob: Before we go any further, one important note. This episode isn't investment advice. We explain the debate; we don't tell anyone what to buy, sell or hold.
Alice: Exactly. And one piece of context about the source: the newsletter says it has been highly bullish on AI since November 2022, calling it the biggest technological revolution in history. So every forecast we share today, we'll label as theirs.
Bob: Fair enough. So what's their headline answer? Bubble, or not?
Alice: Their answer is no: they argue this is a normal correction, not a popped bubble. We'll test that with outside voices: a skeptical essay from Sequoia Capital, the dot-com story in Fortune, and an MIT economist's numbers.
Bob: Set the scene for me. What actually happened over these six weeks?
Alice: The S and P 500 last hit a record high on October 6th. Since then, as reported by The Kobeissi Letter, US stocks have lost an average of 225 billion dollars per day, for 31 straight trading days.
Bob: Wait. Per DAY?
Alice: Per day. It was a crowded stretch: Iran War headlines, earnings season, and worries about AI capex and free cash flow. Then the October Fed meeting, the US midterm elections, and a controversy at Nvidia.
Bob: Okay, but when people say 6 trillion dollars was lost, did that money just vanish into thin air?
Alice: Think of it like the estimate on your house dropping when the whole street goes quiet: you haven't lost cash unless you sell, but you certainly feel poorer. That's my reading of a market cap loss: it's real for anyone who sells now, and paper value for anyone who holds.
Bob: Real pain for sellers, paper pain for holders. Got it. And on top of all that, there was a giant IPO?
Alice: Yes. The article says Anthropic's IPO just raised over 100 billion dollars, the biggest IPO in history, and calls it a clear strain on market liquidity over the last few sessions. We'll come back to that one.
Bob: So with all of that going on, what's the one question they want listeners to ask?
Alice: They put it simply: "What has changed and what is priced in?" Priced in just means the bad news is already reflected in today's prices. Keep that question in mind, because the whole article hangs on it.
Bob: So where did the selloff actually start?
Alice: With spending. The article says it began over concerns about Big Tech's capital expenditures on AI, and what they mean for free cash flow and profits looking into 2027 and 2028. And they add, plainly: "These are real concerns."
Bob: Capex. Free cash flow. Translate, please.
Alice: Jason Fernando's Investopedia explainer puts it well: free cash flow is the cash a company generates after covering operating expenses and capital expenditures. Capex is the spending on long-term assets, the buildings and the machines; for AI, think data centers and chips.
Bob: So when capex shoots up, free cash flow goes down?
Alice: Picture a family renovating their whole house at once: their salaries are great, but the builders' bills empty the bank account before the new rooms earn a cent of rent. That's the worry with Big Tech: profits look strong, while record building spend squeezes the cash left over.
Bob: So Big Tech is living in a house full of scaffolding.
Alice: Very expensive scaffolding. As reported by The Kobeissi Letter, the combined capex of Amazon, Alphabet, Meta and Microsoft rose 36.9 billion dollars in the third quarter of 2026, to a record 201.9 billion. That's up 356 percent since the first quarter of 2024.
Bob: And over a few years, how much in total?
Alice: Big Tech has spent 1.6 trillion dollars on capex over the last three years, according to the article. And the quarterly figure more than doubled in just twelve months.
Bob: Okay. That does sound scary.
Alice: Their twist: they call record infrastructure spending a necessary condition for the biggest technological revolution in history. A brief period of negative free cash flow is, to them, the short-term cost of a long-term buildout. They expect the fear to last until returns get clearer, likely into the first half of 2027.
Bob: Is that a fair reading, though? Can a cash dip actually be a good sign?
Alice: Fernando's explainer says it can go either way. If a company is investing strongly to grow, and free cash flow plus capex is still trending up, that could be good for the stock. But falling free cash flow that looks very different from earnings and sales trends signals a higher likelihood of negative price performance.
Bob: So free cash flow is a kind of early warning light?
Alice: That's how Fernando frames it: declining or volatile free cash flow may reveal operational strain before it appears in reported earnings. He also says it should be analyzed over time and alongside other metrics, not in isolation.
Bob: So the same numbers can tell two stories. What about profits right now?
Alice: This is the article's favourite statistic. S and P 500 companies have reported 31.9 percent year-over-year growth in earnings per share so far in the third quarter, the strongest rate in at least four years. In their words: "There has never been a market with so much fear yet so much growth."
Bob: And have companies been beating forecasts too?
Alice: Yes. The article says actual earnings have beaten estimates for thirteen consecutive quarters, and that the AI buildout is stronger than most people realize. That last part is their reading.
Bob: Theo, you brought a skeptic's take on all this spending. Where does it land?
Theo: In an essay for Sequoia Capital, David Cahn asked one blunt question: "Where is all the revenue?" He saw a big gap between the revenue expectations implied by the AI infrastructure build-out and the actual revenue growth in the AI ecosystem. The essay was published in 2024, so it's an older warning, not a comment on this selloff.
Bob: And how did he do the math?
Theo: He took Nvidia's revenue forecast, doubled it for the total cost of AI data centers, then doubled it again for the margin the end customer needs. That's how, he wrote, the 200 billion dollar question became a 600 billion dollar question.
Bob: But isn't that the railroad argument? Build the tracks, and the trains will come?
Theo: Cahn partly agrees, but he says GPU computing is turning into a commodity, metered per hour, with much less pricing power. Then there's depreciation: think of a GPU like this year's flagship phone, which loses value the moment a better model ships. Railroad tracks don't age like that.
Bob: The next cause on their list is political. Alice, give it to us straight, just as the article tells it.
Alice: As The Kobeissi Letter tells it, a major part of the tech drawdown is tied to the outcome of the US midterm elections. In their view, that outcome ended what they call the "Trump Put". The idea: markets took comfort believing that if things got ugly, President Trump would intervene, because stable markets have been a top priority for him.
Bob: Why call it a put, though?
Alice: It builds on an older idea. James Chen's Investopedia piece on the Greenspan put defines a Fed put: the market belief that the Fed would step in to limit a stock market decline. And he stresses it was a notion that was never officially confirmed.
Bob: Where did that belief come from in the first place?
Alice: Chen traces it to Alan Greenspan, Fed Chair from 1987 to 2006. After the 1987 stock market crash, Greenspan immediately lowered rates and set a precedent that the Fed would intervene in times of crisis.
Bob: So it's a safety promise that nobody actually signed.
Alice: Imagine a trapeze artist with a safety net below: she tries bolder tricks, because a fall won't hurt much. Take the net away, and the very same jump suddenly feels terrifying. Chen notes the downside of the net, too: critics said the Greenspan put encouraged risk-taking.
Bob: And in their telling, the election took the net away?
Alice: That's the article's argument. It reports the Democrats' "sweep" of the midterms. It says the President's ability to intervene in markets has been partially eroded by the change in control of the US Senate and House.
Bob: The article also mentions calls to ban AI?
Alice: It does. It shows a post from a US Congresswoman calling for an AI ban, dated November 4th, 2026. It says calls for a ban have gained traction, and that markets have priced in aggressive regulation and potential bans when Democrats take control in January 2027.
Bob: Do they think a ban will actually happen?
Alice: Their view is no. They argue regulation concerns are overdone, and that a US ban would mean surrendering the global AI race to China. They do expect more regulation and likely new taxes; that's their forecast, not a fact.
Bob: I think for builders that's the one to watch. Regulation lands on products directly.
Alice: Agreed, and the article itself says the end of the put will spur sharper drawdowns, even as it calls the ban fears overdone. So the risk they describe is volatility, more than a ban.
Bob: Then there's the Fed. What's going on with interest rates?
Alice: According to the article, the Fed is hiking interest rates aggressively. US inflation has been above the Fed's 2 percent target for 67 straight months, and the Iran War nears its 300th day. Fed Chair Warsh has said his top priority is getting inflation back to 2 percent.
Bob: Why do higher rates hit tech stocks so hard?
Alice: Mary Hall's Investopedia explainer gives the mechanism: with higher rates, the discount rate used for future cash flow is higher, so future discounted valuations are lower. And the effect is often more pronounced for growth stocks, which get more of their value from expected future earnings.
Bob: Okay. Now say that in normal words.
Alice: Imagine a friend promising to pay you back next year. If your savings account suddenly pays more, that future promise looks less attractive today. AI companies are mostly future promises, which is why Hall says growth stocks feel rate moves the most.
Bob: So my friend's promise loses to my savings account. Harsh.
Alice: The newsletter adds that building AI is getting even more expensive, and Hall explains why debt matters: borrowing costs rise with rates. US investment-grade corporate bond sales are expected to rise 27 percent from a year earlier this November. That's 175 billion dollars, the highest for any November on record.
Bob: Is there any relief in sight?
Alice: Oil, possibly. The article reports a roughly 22 percent drop in oil prices this month. Their view: "higher for longer" stays for now, but the chance of less restrictive Fed policy into mid-2027 is growing, especially if the Iran War gets closer to its end.
Bob: Lena, history has a rates story too, doesn't it?
Lena: It does. Dave Smith's piece in Fortune on why the dot-com boom burst says the first critical blow came from the Federal Reserve, which raised rates through 1999 and 2000. The federal funds rate climbed from around 4.7 percent to 6.5 percent by May 2000, making speculative bets less attractive than safer bonds.
Bob: Rate hikes, then trouble. That sounds uncomfortably familiar.
Alice: It's one parallel, not proof. Hall's explainer also notes rate hikes don't hit every sector the same way, and real rates after inflation matter too. In the article's telling, rates are the third pressure, after spending fears and the end of the put.
Bob: Now the company stories. What happened at Nvidia?
Alice: This one is serious, so we'll stick to the article's words. It says Nvidia has lost Jensen Huang, whom it calls the leader of the AI Revolution, and that the stock fell 40 percent in a matter of weeks. The article doesn't lay out the details of the controversy, so neither will we.
Bob: And what's the newsletter's position on it?
Alice: They took an early stance that Jensen Huang is innocent until proven guilty, and called for the board to leave the door open until an independent investigation establishes the facts. Their reading is that the stock now prices in a near-certainty that he won't return.
Bob: And AMD?
Alice: A short seller report on November 19th accused AMD of smuggling chips into China. The newsletter stresses that this claim has been far from substantiated. So, an accusation, not a finding.
Bob: So what did all of that do to valuations?
Alice: As reported by the newsletter, the tech sector's forward price-to-earnings ratio collapsed to its lowest since March 2020. Nvidia's is down to nine times, which they call one of the cheapest in the S and P 500.
Bob: Forward P/E. Quick translation for the rest of us?
Alice: Picture a bakery with next year's order book completely full. What you'd pay for that bakery depends on whether you believe the orders will really come. A low forward P/E means buyers won't pay much for expected profits; the article says the market simply doesn't believe the E.
Bob: Theo, does Cahn's skeptical essay say anything about Nvidia itself?
Theo: Back in 2024 he noted that about half of Nvidia's data center revenue came from the large cloud providers. And even as a skeptic, he wrote that companies like Nvidia deserve enormous credit for enabling this transition.
Bob: I'd still buy the bakery, though. For the croissants.
Alice: Noted. The article says Nvidia guides about 750 billion dollars in revenue for its 2028 fiscal year, even without Jensen Huang, and the market refuses to believe it. Their takeaway, and it's their view: there's a ton of room for surprise.
Bob: The article says sentiment is at crisis levels. When did the mood actually turn?
Alice: Earlier than the headlines, they say. On October 15th they published an analysis called "Stealth Bear Market", pointing to a bear market under the surface of the S and P 500, especially in tech. In their telling, the midterms and Nvidia only accelerated the shift.
Bob: Quick one: what counts as a bear market?
Alice: Chen's Investopedia piece gives the usual yardstick: a stock market decline of over 20 percent typically denotes a bear market. A stealth one, as I read their term, is that kind of fall hiding inside some sectors while the headline index looks calmer.
Bob: And the so-called fear gauge?
Alice: The Volatility Index, the VIX, is up over 100 percent from its October low. On November 5th it posted its biggest daily gain since February 2018.
Bob: And then the Anthropic IPO lands on top of that fear.
Alice: Right. As reported, institutional demand for the IPO hit 400 billion dollars, with retail orders at 100 billion, nearly 8 times the initial 64.4 billion dollar raise that was expected. The article argues that the IPO significantly reduced liquidity in the market.
Bob: Liquidity. Explain it like I'm five.
Alice: Imagine a food court where one giant new stall opens, and everyone's lunch money goes there. The other stalls aren't worse, they just lose their customers that day. The article's version: a shift in sentiment plus a drawdown in liquidity sent everything lower, from stocks to gold to crypto.
Bob: Everything at once. Who was selling?
Alice: Hedge funds posted their largest trailing six-week net outflow in over a decade this month; the article calls it a dash for the exit. It's like a crowd hearing a fire alarm: people run because everyone else is running. Their view is that once liquidity returns, the market will stabilize.
Bob: Maya, does research say anything about hype and how fast money moves?
Maya: Peter Dizikes in MIT News reports the view of economist Daron Acemoglu: hype is a tangible aspect of the economics of AI, because it drives investment. Acemoglu's warning is that the faster you go, the harder a course correction becomes: "It's very difficult, if you're driving 200 miles an hour, to make a 180-degree turn."
Bob: Real fire or burnt toast, nobody waits around to check. So, the big question: is it a bubble, and did it pop?
Alice: Their answer is no. They argue the market has begun pricing in a far more pessimistic outcome than the fundamentals suggest. Even a further 10 percent drop in the S and P 500, they say, would still be within a healthy correction.
Bob: And the doom calls flying around out there?
Alice: They say they'd be cautious about calls for a drop of 30 percent or more, which are quickly spreading. To be fair, they're careful too: they don't claim the bottom is in, and they say further downside is entirely possible.
Bob: And the opportunity half of the title?
Alice: Think of it like a clearance sign in a shop window: it can mean a bargain, or a shop that's closing, and the sign alone doesn't tell you which. The newsletter reads it as a bargain: they believe polarized sentiment will create an opportunity, as it always does. That's their forecast, and nothing we say here is a reason to buy or sell anything.
Bob: Theo, your skeptic wrote about frenzies like this one, didn't he?
Theo: He did. Cahn says railroad history teaches that a lot of people lose a lot of money during speculative technology waves, and that picking losers is easier than picking winners. Yet his own line is calm: "Speculative frenzies are part of technology, and so they are not something to be afraid of."
Bob: Lena, back to your dot-com history. What else does it say about a moment like this?
Lena: Smith says the overbuild is the most instructive parallel. Telecom companies laid more than 80 million miles of fiber optic cable, and even four years after the bust, 85 to 95 percent of it sat unused, nicknamed dark fiber. It's like building a giant highway for traffic that takes years to show up.
Bob: And today's spending versus today's AI revenue?
Lena: Smith cites tech writer Ed Zitron's estimate for Microsoft, Meta, Tesla, Amazon and Google: about 560 billion dollars invested in AI infrastructure over two years. Against that, just 35 billion dollars in AI-related revenue combined. That's one writer's estimate, so treat it as a rough gauge.
Bob: So is AI just the same story again?
Lena: Smith points to a crucial difference: unlike many dot-com companies with no revenue, major AI players are generating substantial income. But the crash came down to companies that couldn't justify their valuations with actual business results. And the internet did change the world, just not as quickly as its early champions promised.
Bob: Counterpoints. The strongest objection first: what if the AI payoff is simply smaller than everyone expects? Maya?
Maya: That's where Acemoglu's numbers come in, as Dizikes reports them. Acemoglu estimates that AI will produce a modest increase in GDP of 1.1 to 1.6 percent over the next ten years, with a roughly 0.05 percent annual gain in productivity.
Bob: That's a lot less exciting than a revolution.
Maya: In his words, it's "just disappointing relative to the promises that people in the industry and in tech journalism are making." He calls himself a realist rather than a skeptic, and he says hype is making us invest badly in the technology.
Bob: Alice, how would the newsletter answer that?
Alice: With earnings: 31.9 percent growth this quarter, and thirteen straight quarters of beats. But that's corporate profit today, while Acemoglu's estimate is about the whole economy over a decade. In my view, both can be true at once.
Bob: Second objection. The newsletter has a point of view here. They've been bullish on AI since 2022.
Alice: Fair, and they say so openly. Their view is "own assets or be left behind". So weigh their forecasts as the bull case, and Cahn's and Acemoglu's as the skeptics' case.
Bob: Theo, does the depreciation point cut against their capex story?
Theo: It can. Cahn argues the market overestimates how well today's top chips will hold their value in three to four years, because each new generation improves so fast. Physical infrastructure, he says, doesn't follow that kind of curve.
Alice: And Fernando's explainer adds a fair caveat on the other side: purchases are subtracted from free cash flow in the year they're made, not spread over many years. So one bad cash year, on its own, doesn't settle the question.
Bob: Time for Builder's Corner. Alice, one thing builders can do this week. And not a stock tip!
Alice: Not a stock tip. Run your own free cash flow check: cash coming in, minus running costs, minus what you spend on compute and hardware. Fernando's explainer warns that a company with consistently low or negative free cash flow might be forced into costly rounds of fundraising.
Bob: And then stress-test it?
Alice: Exactly. Redo it as if borrowing gets pricier and your compute bill rises, because, as Hall notes, growth companies are heavily reliant on capital. My suggestion: if your runway still holds, good; if not, you've learned it before the market teaches you.
Bob: Takeaways. Mine: 6 trillion dollars in losses is alarming, but the article's own question is the useful one. What has changed, and what is priced in?
Alice: Mine: the spending worry is real. Even the newsletter calls the capex and free cash flow concerns real, and the bills are record-sized.
Bob: And the politics and the Fed?
Alice: Two pressures, in their telling: the end of what they call the Trump Put, and rate hikes that make AI more expensive to build. Whether regulation fears are overdone is their call, and only time will settle it.
Bob: And the skeptics?
Alice: My take: keep them in the room. Cahn's revenue gap, Acemoglu's modest estimate and the dark fiber story are reasons to watch returns, not just spending.
Bob: And once more, for everyone listening: none of this is investment advice.
Alice: Which leaves the open questions. The biggest one: will the returns on all this capex show up? The newsletter itself expects clarity only around the first half of 2027.
Bob: And on Nvidia?
Alice: Whether an independent investigation establishes the facts, and whether the market starts believing the company's guidance again. Nobody knows yet, the article included.
Bob: And for builders: whether regulation lands softer or harder than markets fear. I'd bet that's the one we'll be talking about next year.
Alice: Thanks to The Kobeissi Letter for their article on the selloff. And to Jason Fernando, James Chen and Mary Hall at Investopedia, Dave Smith in Fortune, David Cahn at Sequoia Capital, and Peter Dizikes in MIT News.
Bob: One last reminder: we explained a debate, and nothing here is investment advice. That's all for today on Build Wiz AI Show. Thanks for listening, and see you next time.