
Illustration. Image: MadRobot
As of November 19, 2026, the AI bubble has not popped, but it is wobbling: the Nasdaq fell into a correction on November 5, Nvidia's shares fell even after record sales, Alphabet's, Amazon's and Meta's free cash flow has turned negative as they spend on data centres, and the International Monetary Fund now names an AI market correction as one of the biggest risks to the global financial system in its October Global Financial Stability Report. What hasn't happened yet is the thing that ends a bubble: the money pouring in has not stopped, and Anthropic's shares opened 34% above their IPO price.
Below: the signs to watch, where each stands now, and what stage the boom is at. We'll keep it updated.
Last checked: November 19, 2026, against Nvidia's November 17 third-quarter results, Anthropic's first day of trading on November 19, the November 5 stock market sell-off, the IMF's October 2026 Global Financial Stability Report, the Bank of England's September 2026 Financial Policy Committee record, the Federal Reserve's May 2026 Financial Stability Report and October 28 rate decision, Nvidia's latest results, and Alphabet's, Amazon's, Meta's and Microsoft's third-quarter results.
Is the AI bubble popping right now?
Not yet, on the evidence so far. A bubble pops when prices fall and keep falling because the buyers have run out, and that is not what 2026 looks like. Prices have dropped hard several times, most sharply on November 5, but spending has kept going up. The newest worry is cash: the build-out now costs three of the richest companies more than their businesses bring in.
- July: AI company shares "fell sharply", and the fall was made worse by leveraged investors being forced to sell, according to the Bank of England's Financial Policy Committee record. Markets kept working and there was no wider crisis.
- October 14 and 15: AI stocks fell again after US inflation came in hotter than expected and traders bet on more interest rate rises. The pressure centred on heavy borrowers building data centres: CoreWeave fell 15% in a day and Oracle nearly 9%, and both companies' bond yields hit record highs. Even so, the Nasdaq was only about 3% below the record it set the week before.
- October 28 and 29: Alphabet's shares fell about 10% in after-hours and pre-market trading after its third-quarter results showed free cash flow of minus $9.1 billion, deeper than the minus $5.9 billion of the quarter before, as it spent $51.4 billion on property and equipment in three months. A close more than 9.77% lower would be its worst reaction to results since it listed in 2004. Meta, which reported on October 21, also had negative free cash flow, of $2.1 billion. Microsoft, whose cash flow stayed positive, rose nearly 4%.
- October 29: Amazon's shares fell about 7% in after-hours trading, from a close of $243.79, after its third-quarter results showed free cash flow of minus $42.4 billion over the past 12 months, against plus $14.8 billion a year earlier. It spent $77.3 billion on property and equipment in the quarter, more than double a year earlier, a rise it says "primarily reflects investments in artificial intelligence". AWS, its cloud business, grew 39%, its fastest in nearly five years.
- November 5: the worst day for US shares since April 2025, after Democrats swept the midterms. Details below.
- November 17 to 19: Nvidia reported record sales but its shares fell, then Anthropic's shares jumped on their debut. Details below.
The cracks are in the most indebted corners of the boom, in Big Tech's cash flow and, since the midterms, in investors' faith that data centre building will keep its pace. Actual demand for AI chips hasn't fallen yet.
Update: Nvidia's record quarter, then Anthropic's 34% debut
Nvidia's third-quarter results on November 17 cleared the bar this guide had set: revenue rose 103% on a year earlier to a record $115.8 billion, ahead of its own $108 billion forecast, and it expects about $130 billion in the current quarter, which has an extra week. They were its first results since Jensen Huang resigned as chief executive. The shares still fell 2.9% the next day as investors weighed life without him, AP reported, and the Nasdaq had closed on November 17 at 24,328, close to where it ended on November 5. Two details matter for the bubble: Nvidia has stopped shipping AI chips to China, and interim chief executive Colette Kress said it would be "rigorously evaluating" new investments and guarantees, having given $118 billion of guarantees for customers' land, power and hardware financing.
Then the biggest test of investors' appetite went the boom's way. Anthropic priced its IPO at $83, above its $70 to $78 range, raising $81.3 billion, and its shares opened 34% higher at $111.50 on November 19, about $2 trillion counting every share and stock award, CNN calculated. Ryan Detrick of Carson Group pointed out that big IPOs since 2004 have had a median first-day gain of 31%, but also a median worst fall of 53% over the following 12 months.
Update: the Nasdaq falls into a correction
The Nasdaq Composite fell 5.52% to 24,315 on Thursday, November 5, its worst day since April 2025, and closed more than 10% below the record it set in October, the usual definition of a correction. The S&P 500 lost 4.36% and the Dow 1,833 points, or 3.44%, according to CNBC. Chip stocks took the worst of it: Nvidia fell about 9%, wiping out more than $440 billion of its market value, Broadcom 10% and AMD and Micron 15% or more, while the Philadelphia Semiconductor index had its worst day since March 2020, NBC News reported. The seven biggest US tech companies lost more than $1 trillion between them.
The trigger was politics as much as profits. Investors fear the Democrats' sweep of both houses of Congress, and the candidates and ballot measures that promised to curb data centres, will slow the building that keeps AI chipmakers busy. Weak forecasts from Qualcomm and Arm added to the selling, and the heavily borrowed data centre firms fell again: CoreWeave almost 10% and Oracle nearly 7%. Not everyone reads it as the end. Ed Yardeni of Yardeni Research told NBC News: "I don't think this is an AI bubble bursting".
Signs the AI bubble is bursting: the checklist
These are the signs economists, central banks and investors watch most closely, with where each one stood on November 19, 2026.
| Sign | What to watch | Where it stands |
|---|---|---|
| Big Tech cuts AI spending | Alphabet, Amazon, Meta and Microsoft lowering their capital spending plans | Not happening Meta narrowed its 2026 range upwards again, to $135 billion to $145 billion; Alphabet kept $195 billion to $205 billion and expects a significant rise in 2027; Amazon still expects about $220 billion |
| Nvidia's growth stalls | Chip sales slowing, or a weak forecast | Not happening Revenue up 103% to a record $115.8 billion in the quarter to October 25, beating its $108 billion forecast, with about $130 billion forecast for the next; it has stopped shipping AI chips to China |
| AI stocks fall and stay down | Sell-offs that don't recover | Wobbling The Nasdaq fell into a correction on November 5, more than 10% below its October record, and was still close to that level on November 17; Nvidia's shares fell 2.9% the day after its record results |
| Debt stress at weaker players | Rising bond yields and default insurance costs at data centre builders | Showing CoreWeave's and Oracle's bond yields hit records in October, both shares fell again on November 5, and CoreWeave has lost nearly a third of its value this year |
| Circular deals unwind | AI firms that fund each other cancelling or cutting deals | Building up The IMF counts $1.4 trillion of revenue promises tied to these deals; none has unwound yet, but Nvidia says it will be "rigorously evaluating" new investments and guarantees |
| AI revenue doesn't show up | Paying customers failing to keep pace with spending | Mixed Google Cloud revenue doubled to $30.7 billion, Microsoft's Azure grew 48% and Amazon's AWS 39%, but Alphabet's, Amazon's and Meta's free cash flow was negative as spending outran it |
| IPOs and funding rounds get pulled | Big listings delayed or priced lower | Not happening Anthropic priced its IPO at $83, above its range, raising $81.3 billion, and its shares opened 34% higher on November 19 |
| Interest rates rise | Higher rates make borrowing to build data centres dearer | Showing The Fed held rates at 3.75% to 4% on October 28, but two policymakers voted for a rise; the 10-year US Treasury yield was above 5% on November 18 |
| Regulators sound the alarm | Central banks and the IMF warning about valuations | Showing The IMF, the Bank of England and the Federal Reserve have all flagged AI |
Green means no sign of trouble, amber means mixed, red means the warning sign is present.
The money is still flowing in
The biggest sign a bubble is ending is the spending stopping, and it hasn't. Meta's third-quarter results say it now expects 2026 capital spending of $135 billion to $145 billion, "narrowed from our prior outlook of $130-145 billion", and it is selling up to $95 billion of new shares to help pay for its data centres. Alphabet, which lifted its 2026 range to $195 billion to $205 billion in July, kept it there in its October 28 results, and finance chief Anat Ashkenazi told analysts: "We continue to expect our capex to increase significantly in 2027." Microsoft's cash spending on property and equipment roughly doubled to $39.4 billion in the three months to September 30.
Nvidia's results for the quarter to October 25 showed data centre revenue of $107.6 billion, up 110% on a year earlier. "Our demand is accelerating. Our supply is growing," interim chief executive Colette Kress told analysts.
The debt is where the worry is
What has changed in 2026 is how the boom is paid for. Big Tech is increasingly borrowing to build. The Bank of England's committee cited Morgan Stanley estimates of about $450 billion of AI-related debt issued worldwide by early September, more than double the whole of 2025, and warned that "increasing leverage, opacity and, at times, 'circular arrangements'" could "amplify losses if expectations disappointed". We covered that warning in full.
The IMF went further. Its blog on the report says a review of 14 deals since 2023 between the cloud giants and AI firms "found about $300 billion in equity investments and $650 billion in debt" in exchange for "$1.4 trillion in future revenue commitments". If one of the key companies hit trouble, it warns, revenue forecasts and valuations would fall, "weakening the value of the accumulated debt in those deals". Anthropic's own IPO paperwork shows $518 billion of compute bills, around 80% of them non-cancellable, which is the kind of commitment the IMF means.
What stage is the AI bubble at?
Economists often describe bubbles in five stages, a model built on the work of Hyman Minsky and Charles Kindleberger:
- Displacement: a new technology changes what seems possible. For AI, that was ChatGPT in late 2022.
- Boom: money rushes in and prices climb steadily. Roughly 2023 to 2025.
- Euphoria: caution goes out of the window, borrowing soars.
- Profit-taking: early investors start selling, and prices get jumpy.
- Panic: prices fall, lenders pull back, and forced selling feeds on itself.
Measured against that, 2026 has features of stages three and four. Borrowing is surging and prices are high, the Federal Reserve's May Financial Stability Report says the ratio of S&P 500 share prices to earnings "stayed in the upper range of its historical distribution", and a run of sharp sell-offs, the worst on November 5, shows investors getting nervous. There is no sign of stage five: no lender has pulled out, the biggest buyers of AI chips are still raising budgets, and investors piled into Anthropic's record IPO. But a boom can sit in these stages for years, and it's only clear afterwards which wobble was the last.
AI bubble vs the dot-com bubble
The comparison everyone reaches for is 2000, when internet stocks collapsed. The two booms differ in one big way: profits. In 2000 many of the hottest companies had no earnings at all. Today the companies spending the most, Alphabet, Amazon, Meta and Microsoft, are among the most profitable businesses in history, and Nvidia keeps about 75 cents of every dollar of sales as gross profit.
The similarities are uncomfortable, though. Before 2000, telecoms equipment makers lent to their own customers so they could buy more kit, and were hit twice when those customers failed. The IMF's "circular financing" warning describes the same pattern: chip makers and cloud giants investing in AI labs that spend the money on their chips and cloud. And as in 2000, a handful of companies make up a large share of the US market.
Bank of England governor Andrew Bailey made the point in an interview with the BBC on October 1: the technology can be real and still produce losers, because "everybody is currently priced to be a winner". Looking back at the early internet, he said:
Nobody can remember Netscape today. It doesn’t exist. So not everybody always wins.
Andrew Bailey, Governor of the Bank of England, to the BBC
What happens if the AI bubble bursts?
The IMF's report says weaker AI investment or profits "could lead to a correction in AI-related asset prices", and that "corrections could be amplified by rising leverage and concentrated exposures". In practice, a burst would most likely mean:
- Pensions and savings fall: index funds, 401(k)s and workplace pensions hold large amounts of AI-linked shares.
- Lenders take losses: private credit funds and banks that lent to data centre builders would be hit first. The IMF's stress test found banks holding 22% of global banking assets would fall below their capital limits in its adverse scenario, up from 18% a year earlier.
- Government borrowing gets harder: the Bank of England warns that because growth forecasts lean on AI productivity gains, a rethink "could therefore affect not only AI-related asset valuations but also sovereign debt markets".
- Slower growth: data centre building has been propping up US growth, so jobs well beyond tech would feel it.
- Dearer or scarcer AI tools: if cheap money dries up, AI companies would be under pressure to cut free tiers, raise prices or shut weaker products.
AI itself wouldn't disappear, any more than the internet did after 2000.
Is the AI bubble bursting in the UK, Europe and Asia?
The boom is centred on US shares, but it isn't a US-only risk:
- South Korea: Samsung and SK Hynix make the high-bandwidth memory AI chips need, and they dominate the Kospi index, which is why Korean shares have swung hardest in each AI sell-off this year.
- Taiwan: TSMC makes almost every advanced AI chip, so any slowdown in orders lands there quickly.
- Japan: SoftBank is one of OpenAI's largest backers.
- UK: AI hyperscalers accounted for 47% of sterling corporate bond issuance this year, the Bank of England says, and British data centre firms such as Nscale are heading for the stock market.
- Europe, Canada and Australia: chip toolmakers such as the Netherlands' ASML depend on AI chip demand, and big pension funds hold large stakes in US tech.
The IMF says the risks could compound: an AI correction, rising government bond yields and a major cyberattack might each be manageable alone, but not all at once.
When will the AI bubble burst?
Nobody knows. Bubbles usually end when something forces the money to stop: rates rise, a big borrower fails, or the biggest spenders blink. What to watch next:
- 2027 spending plans: Alphabet, Amazon, Meta and Microsoft have all reported third-quarter results and none cut its spending plans; Amazon kept its 2026 forecast at about $220 billion. A cut by any of them would be the clearest sign yet.
- Big Tech cash flow: whether Alphabet's, Amazon's and Meta's free cash flow turns positive again in their fourth-quarter results, due in early 2027.
- Nvidia's next results: due in early 2027, measured against its $130 billion forecast, and whether its retreat from China and more careful approach to customer guarantees slow its sales.
- US interest rate decisions: more rate rises would make the borrowing behind the boom dearer.
- Anthropic's shares: they opened at $111.50 on November 19, about the $2 trillion valuation it had hoped for, up from about $1 trillion at its November 10 price range. Whether ANT holds well above its $83 IPO price once the first-day excitement fades is the next test of appetite.
- Politics after the midterms: whether the new Congress and the states move to slow or halt data centre building, the fear behind the November 5 sell-off.
- Data centre debt: bond yields and default insurance costs at CoreWeave, Oracle and other heavy borrowers.
AI bubble FAQ
Frequently asked questions
Is the AI bubble popping?
Not yet. As of November 19, 2026, AI stocks have had sharp sell-offs in July and mid-October, the Nasdaq fell into a correction after its worst day since April 2025 on November 5, Alphabet's and Amazon's shares fell after their results showed AI spending outrunning their cash, and the most indebted data centre companies are under strain, but Big Tech is still raising its AI spending, Nvidia's sales hit a record $115.8 billion last quarter, and Anthropic's shares opened 34% above their IPO price.
Is the AI bubble real?
Central banks take the risk seriously. The IMF, the Bank of England and the US Federal Reserve have all warned that AI share prices rest on optimistic profit expectations and growing debt. Whether it is a bubble will only be clear afterwards: the technology is real, but valuations could still be too high.
What are the signs the AI bubble is about to burst?
The big ones are Big Tech cutting its data centre spending, Nvidia's sales growth stalling, sell-offs that don't recover, debt trouble at heavily borrowed data centre builders, and AI firms that fund each other cancelling deals. In October 2026 only the debt stress and rising interest rates were clearly showing.
When will the AI bubble burst?
No one can say. Watch Big Tech's 2027 spending plans, Nvidia's next results in early 2027, Big Tech's cash flow, US interest rate decisions and how Anthropic's shares hold up after its IPO. Bubbles usually end when the money stops flowing in, and so far it hasn't.
What stage of the AI bubble are we in?
Using the classic five-stage bubble model (displacement, boom, euphoria, profit-taking, panic), 2026 shows signs of euphoria and profit-taking: heavy borrowing, high valuations and nervous sell-offs. There is no sign yet of the panic stage.
Is the AI bubble like the dot-com bubble?
Partly. Unlike most dot-com stars, today's biggest AI spenders are hugely profitable. But the IMF's warning about circular financing, where chip and cloud firms invest in AI labs that then buy their products, echoes how telecoms suppliers lent to their own customers before 2000.
What happens if the AI bubble bursts?
Shares in AI-linked companies would fall, hitting pensions and index funds; lenders to data centre builders would take losses; and growth would slow. The Bank of England warns it could even unsettle government bond markets. AI itself would not disappear.
Should I sell my AI stocks?
MadRobot can't give financial advice. If you are worried, check how much of your pension or index fund is in a few big tech names, and speak to a regulated financial adviser before making changes.
Correction: an earlier version of this guide said Alphabet's free cash flow turned negative in the third quarter, when it was already negative in the second, and gave a Microsoft spending figure not found in its results release, now replaced.
Sources: IMF Global Financial Stability Report, October 2026, IMF blog (Oct 13), Bank of England Financial Policy Committee record (September 2026), Federal Reserve Financial Stability Report (May 2026), Nvidia (Aug 26), Nvidia (Nov 17), Meta (Oct 21), Alphabet (Oct 28), Microsoft (Oct 28), Amazon (Oct 29), Federal Reserve (Oct 28), CNBC (Oct 29), BBC (Andrew Bailey interview), Reuters (Oct 15 market report); CNBC, CNBC (Nov 5); NBC News; AP (Nov 17, Nov 18); CNN (Nov 19).





