For three years, almost any company that could tie its story to artificial intelligence saw its shares rise. Chipmakers, cloud providers and software firms became the largest companies in the world, and a handful of them now carry much of the US stock market on their own.
In October 2026 that run started to wobble. Worries about how AI companies pay for their data centers, mixed earnings and a jump in oil prices hit tech stocks hard. The Nasdaq had its worst day since June on October 22, days before the next Fed rate decision (source: CNBC, October 22, 2026). Bitcoin fell with it. Well-known investors said out loud what many traders had been asking privately: whether the whole AI trade has become a bubble.
That question matters even if you never buy a tech stock. AI shares now move the indexes that set the mood for every risky asset, including crypto. The money flowing into AI data centers is also borrowed and recycled in ways that can spread losses quickly if expectations change.
What is the AI bubble?
The AI bubble is the idea that the prices of companies tied to artificial intelligence have risen far faster than the profits AI can realistically produce. The claim covers chipmakers such as Nvidia, the cloud giants spending hundreds of billions on data centers, and private AI labs valued in the hundreds of billions of dollars.
In economist Robert Shiller's description, a bubble starts when rising prices spread excitement from one investor to the next. More buyers join who doubt the real value but buy anyway, partly through envy and partly through "a gambler's excitement" (source: Robert Shiller, Nobel lecture). The Federal Reserve uses a drier term, "valuation pressures", for asset prices that are high compared with economic fundamentals or past norms. It warns that they raise the chance of unusually large price falls (source: Federal Reserve, Financial Stability Report, May 2026).
So the AI bubble debate is really two questions. The first is whether AI companies will earn enough to justify today's prices. The second is what happens to everything connected to them if they do not.
Does a bubble have to burst?
A bubble does not have to end in one sudden crash. Shiller notes that there is no widely accepted definition of the word, and that a sudden, final crash is not an essential part of it. Even the abrupt ends of the booms of 1929, 2000 and 2007 were followed before long by new booms (source: Robert Shiller, Nobel lecture).
That matters when you read headlines, because a sharp drop in AI stocks, like the one in October 2026, is a sign that expectations are being tested, not proof that a bubble has burst. It can turn out to be a pause in a long rise, or the start of a much longer decline, and nobody can tell which while it is happening.
Why people say AI is a bubble
People who call AI a bubble point to four things. Valuations are near dot-com levels, and the market leans heavily on a few AI companies. Some companies now spend more on data centers than they earn in cash. And a web of deals has AI companies funding each other's purchases.
Start with valuations, the measure critics cite first. The Shiller CAPE ratio, which compares stock prices with ten years of inflation-adjusted earnings, stood at 40.36 for the S&P 500 at the close on October 23, 2026. Its record is 44.19, set in December 1999 at the height of the dot-com bubble, so today's reading is about 91% of that peak (source: multpl, Shiller PE ratio).
Concentration adds to the risk, because a few companies now move the whole index. The Bank of England estimates that AI companies make up around half of the S&P 500, up from around a quarter in 2022 (source: Bank of England, Financial Stability Report, July 2026). Earnings growth is just as narrow. Before Meta reported, FactSet expected the seven big tech stocks known as the Magnificent 7 to grow earnings 33.2% in the third quarter of 2026. Without Nvidia and Meta, that figure would have been 7.5% (source: FactSet, October 16, 2026).
Then there is spending, which is where cash flow comes in. As Meta's AI capex and cash flow explains, Meta spends heavily ahead of revenue. It spent $34.9 billion on capital expenditures in the third quarter of 2026, against $32.8 billion of cash from operations. That left free cash flow of negative $2.1 billion, and it expects $135 billion to $145 billion of capital spending for the year (source: Meta, third-quarter 2026 results). Amazon has said about $220 billion for 2026, from its report for the quarter to June (source: CNBC). Alphabet raised its 2026 plan to $195 billion to $205 billion in July (source: Bloomberg), and Alphabet's AI capex and cash flow shows what that does to its free cash flow. Morningstar puts Microsoft's calendar-2026 spending at about $175 billion after an accounting change (source: CFO Dive), so the four biggest spenders are on course for roughly $725 billion to $745 billion this year.
How circular financing works in AI
Circular financing is when an AI supplier invests in, or lends to, a customer that then uses the money to buy the supplier's products. The deal can lift both companies' reported revenue and valuations at once, so critics say part of the AI boom is money moving in a loop.
The IMF counts 14 such deals since 2023, worth about $300 billion in equity and $650 billion in debt, in return for $1.4 trillion of promised future revenue (source: IMF blog, October 2026). The best-known examples involve OpenAI. Nvidia said in September 2025 that it intended to invest up to $100 billion in OpenAI as OpenAI deployed Nvidia systems (source: NVIDIA). AMD gave OpenAI a warrant for up to 160 million AMD shares alongside a deal to supply its chips (source: AMD).
The pattern is still live, and Meta is the latest example. On October 26, 2026, Meta launched a sale of $22.5 billion of common stock plus convertible preferred shares. Its chip suppliers Nvidia, AMD and Broadcom indicated interest in buying up to $5.0 billion of the common stock (source: Meta, SEC filing).
How much AI spending is funded by debt
A growing share of the AI build-out is paid for with borrowed money. The Bank of England reports a Morgan Stanley estimate that AI-related debt issuance reached around $450 billion by early September 2026. That is more than double the total for all of 2025. It also cites a JPMorgan estimate that about $4.1 trillion of AI spending will be paid for with debt between 2026 and 2030 (source: Bank of England, Financial Policy Committee, September 2026).
Debt changes how a downturn plays out, because shareholders can absorb a fall in stock prices, but borrowers still owe interest when revenue disappoints. The five AI hyperscalers made up just 3% of outstanding US investment-grade debt at the end of 2025 but more than 15% of new issuance by early May 2026 (source: Bank of England, Financial Stability Report, July 2026). The Bank of England warns that leverage, hidden risks and circular deals could make losses bigger if expectations disappoint (source: Bank of England, Financial Policy Committee, September 2026).
The case against an AI bubble
The case against a bubble rests on profits. Unlike many dot-com companies, today's AI leaders earn large and growing amounts of money, and much of their spending is paid for from that cash. On one common measure, the stock market is not especially expensive by its own recent history.
FactSet puts the S&P 500's forward price-to-earnings ratio, which uses analysts' estimates of the next 12 months of profits, at 18.7 on October 23, 2026, below its 10-year average of 19.1 (source: FactSet, October 23, 2026). The two gauges disagree because CAPE looks back at ten years of earnings, while forward P/E looks ahead at a year of fast-growing estimates. If those estimates come true, prices are less stretched than CAPE suggests.
The earnings themselves are strong: FactSet expects S&P 500 earnings to grow 29.5% in the third quarter of 2026, which would be a third straight quarter above 25% (source: FactSet, October 23, 2026). Nvidia reported revenue of $96.2 billion for its quarter to late July, up 106% from a year earlier (source: NVIDIA). Microsoft's capital spending in its fiscal year to June 2026 was about 63% of its operating cash flow, so it paid for its build-out from the cash its business brings in (source: Microsoft). Nvidia sells the chips, and its chief executive, Jensen Huang, said in August 2026 that AI "has reached its inflection point" and is "doing useful work" (source: NVIDIA). Even the IMF, while warning about a correction, says the companies behind the AI boom have shown "exceptional earnings growth" (source: IMF blog, October 2026).
Even some of the people who use the word "bubble" expect the technology to last. Citadel founder Ken Griffin said in October 2026 that the companies left standing after a bubble often lead their markets for decades (source: Decrypt), the way Amazon outlived the dot-com crash.
AI bubble vs the dot-com bubble
The dot-com bubble is the comparison most people reach for. Stock valuations today are close to their 1999 peak. The difference is that the biggest AI spenders pay for much of their build-out from the cash their businesses bring in. The telecom carriers that built the internet borrowed heavily, and many went bankrupt when demand fell short.
Measure | Dot-com bubble | AI boom |
|---|---|---|
S&P 500 Shiller CAPE | 44.19 at its December 1999 peak | 40.36 on October 23, 2026 |
What happened next | CAPE fell to 21.96 by October 2002 (source: multpl, Shiller PE by month) | Still being tested |
Nasdaq Composite | Record close 5,048.62 on March 10, 2000; low close 1,114.11 in 2002 | Fell 2.6% on October 22, 2026 |
How the build-out was paid for | Heavy borrowing by telecom firms, plus financing from their equipment makers | Operating cash flow, plus fast-growing debt, share sales and circular deals |
Internet stocks peaked in March 2000, and from that record close to the 2002 low the Nasdaq Composite fell about 78% (source: Reuters via Business Standard). The deeper damage came from the telecom companies laying fiber-optic cable. They borrowed heavily on forecasts of soaring demand, while new technology multiplied what each cable could carry. When demand fell short, prices collapsed and many could not repay their debts (source: François Bar, Stanford). Telecom carriers had raised more than $1 trillion in equity and debt since 1998, not counting financing from the equipment makers whose gear they bought (source: House Financial Services Committee).
Griffin draws the same line for today. He compared AI chips to airline seats, warning that a glut could send their price into freefall, as fiber prices did after 2000 (source: Decrypt). Amazon's own path through both booms is traced in Amazon's AI capex and cash flow.
Who is warning about an AI bubble
Michael Burry, Ken Griffin, the IMF and the Bank of England have all warned about AI valuations in 2026. They disagree on what would trigger a fall and when: Burry has bet on a market top within months, while the IMF describes a steep drop only as a scenario it tested.
Michael Burry is the investor known for betting against the US housing market before 2008. On October 11, 2026, he wrote that he had moved most of his bearish bets into put options on the Nasdaq 100 index and on the VanEck Semiconductor ETF (SMH), which expire in June 2027. A put option gains value when the price of what it tracks falls. Burry said he sees a "significant top" in the market within the coming nine months (source: Business Insider via Yahoo Finance).
Ken Griffin said on October 20, 2026 that there are "almost certainly" bubbles in AI right now, while adding that big changes to the economy usually come with a bubble (source: 24/7 Wall St.). Official bodies are more measured. IMF Managing Director Kristalina Georgieva said in October 2026 that the world economy is "standing on one leg, the AI leg". An IMF scenario found that a steep drop in AI investment could cut global output by 1.1% by 2028 (source: Reuters via BusinessDay). That is a test of what could happen, not the IMF's forecast.
Signs an AI bubble is bursting
No single number tells you a bubble is bursting, but traders watch a handful of signals, and each one has a place you can check it yourself. The common thread is money: whether AI revenue keeps up with AI spending, and whether lenders and new investors keep supplying the cash.
The first sign is a cut to spending plans. If the largest companies cut their capital spending guidance, chipmakers and data-center suppliers lose their biggest customers. You find the guidance in each company's quarterly earnings release and call, and how earnings season works explains how to read them. Microsoft (source: Microsoft) and Alphabet (source: Alphabet) report on October 28, 2026, with Amazon on October 29 (source: Amazon).
The second is free cash flow, which is the cash a company has left after its capital spending. When it turns negative, as it did at Meta in the third quarter, a company has to borrow, issue shares or spend its savings to keep building. It appears in the cash flow statement of the same earnings release.
The third is credit, the price lenders charge. Rising bond yields or credit default swap prices on AI borrowers mean lenders want more to take the risk. On October 22, CNBC reported that the bond yields and credit default swap prices of CoreWeave, the leading neocloud, stayed high as its shares fell 10% (source: CNBC, October 22, 2026). A neocloud is a cloud company built mainly to rent out AI chips. Meta's compute deals and rented AI capacity explains how they sell that capacity.
The fourth is how new listings are received by public investors. A weak reception for big AI offerings would show that public investors are less willing to pay private-market prices. Anthropic filed publicly for its initial public offering (IPO) on October 26, 2026. Its filing reported an operating profit for the quarter to June 2026 but an operating loss for the first half of the year (source: Anthropic, Form S-1). What the Anthropic IPO means covers how it is valued.
The fifth is breadth, meaning how many companies share in the gains. If fewer companies carry the market's earnings growth, as FactSet's Magnificent 7 figures already show, the rally rests on fewer shoulders, and a miss from one of them hits the whole index.
What happened to AI stocks in October 2026
AI stocks fell on several days in October 2026. On October 8, the Nasdaq dropped 1.25% after reports that OpenAI's annualized revenue was below what it had signaled, as Oracle fell more than 5%, Nvidia nearly 3% and AMD almost 4% (source: CNBC, October 8, 2026).
The sharpest fall came on October 22, when the Nasdaq slid 2.6%, its worst session since June, on renewed concerns about AI spending. Meta fell about 7% despite beating revenue expectations, AMD more than 8% and Nvidia 4%. Nvidia chief executive Jensen Huang's remark that capital was scarce for neoclouds also unsettled traders (source: CNBC, October 22, 2026). The 10-year Treasury yield fell below 5% that day, for the first time in a month, as investors grew cautious about the borrowing behind AI data centers, and the next day the Dow fell 1.23% (source: CNBC, October 23, 2026).
How bitcoin trades when AI stocks fall
Bitcoin has moved closely with AI stocks in October 2026. On BloFin, the daily changes of the BTCUSDT Perpetual and the QQQUSDT Perpetual, which tracks the Nasdaq 100, had a correlation of 0.72 across the 17 trading days from October 1 to October 23.
Bitcoin's correlation with the NVDAUSDT Perpetual was 0.77 over the same days. A correlation of 1 would mean the two always moved together, and 0 would mean no link at all. On the three days the QQQUSDT Perpetual fell most, bitcoin fell too, and it rose with Nvidia in the October 21 rebound.
Day | BTCUSDT Perpetual | NVDAUSDT Perpetual | QQQUSDT Perpetual |
|---|---|---|---|
October 14 | −1.75% | −3.12% | −1.92% |
October 16 | −0.96% | −1.39% | −2.50% |
October 21 | +2.31% | +2.51% | +1.11% |
October 22 | −3.57% | −4.25% | −2.84% |
Each change runs from one US stock close (4PM ET, 8PM UTC) to the next. From September 30 to October 23, bitcoin fell 2.78% on BloFin, while the Nasdaq 100 perpetual fell 4.75% and the Nvidia perpetual 7.67%. Seventeen days is a short sample, and moving together does not mean one caused the other, but it shows that an AI sell-off has recently reached crypto on the same day. How Nvidia stock and bitcoin move together looks at the longer history. To watch whether an AI sell-off reaches crypto, put the BTCUSDT Perpetual chart beside the NVDAUSDT Perpetual on the same timeframe.
How to trade AI-stock swings on BloFin
AI-stock swings cluster around known dates, especially big tech earnings reports. So plan each trade around when its market is open. BloFin lists the NVDAUSDT Perpetual and the QQQUSDT Perpetual for direct exposure to Nvidia and the Nasdaq 100, while the BTCUSDT Perpetual trades 24 hours a day.
BloFin also lists perpetuals on other AI names, such as AMD, Alphabet, Amazon and Meta, and TradFi assets you can trade on BloFin covers its other stock, fund and commodity markets.
The hours matter most on earnings nights. Microsoft, Alphabet and Amazon all report after the US close. BloFin's stock perpetuals follow the US regular session, 9.30AM to 4PM ET. That is 1.30PM to 8PM UTC until US daylight saving time ends on November 1, and 2.30PM to 9PM UTC after that. Once the session closes, BloFin may accept only reduce-only orders on them, meaning you can cut or close a position but not open a new one (source: BloFin, stock futures trading hours). So when results land in the evening, you may only be able to cut an NVDAUSDT or QQQUSDT position until the next open. Bitcoin keeps trading and can react first.
Start the size from the loss you can accept. The NVDAUSDT Perpetual fell 4.25% on October 22, so allow for at least 5%. If 25 USDT is the most you want to lose in a day, a position of about 500 USDT fits, since 5% of 500 is 25. Opened at 5x leverage, it ties up 100 USDT of margin, and a 5% move against you takes a quarter of that. How liquidation works explains what happens as losses approach your maintenance margin.
Short positions carry their own risk in a bubble debate, because prices can keep rising long after valuations look stretched, as they did through 1999. A stop-loss caps a steady move against you. But a price that opens sharply higher or lower after an evening report can skip past it and fill at the next available price. How to set a stop-loss on BloFin shows how to place one. You can also copy a lead trader on the NVDAUSDT, QQQUSDT and BTCUSDT Perpetuals, with a limit you set on how much of your balance each one uses.
Looking to trade the AI bubble debate? To get started, you'll need to first create a BloFin account, fund your account with cryptocurrency, and navigate to the NVDAUSDT Perpetual, QQQUSDT Perpetual or BTCUSDT Perpetual page, or browse the lead traders you can copy on BloFin.
Frequently asked questions
What is the Shiller CAPE ratio?
The Shiller CAPE ratio divides the S&P 500's price by its average earnings over the past ten years, adjusted for inflation. Using ten years smooths out booms and recessions. Since 1871 it has averaged 17.44, with a low of 4.78 in December 1920 and a high of 44.19 in December 1999 (source: multpl, Shiller PE ratio).
Has the AI bubble burst?
A burst can only be confirmed well after the fact. AI stocks fell sharply in July 2026. In September, the Bank of England noted that the risk of "a sharper correction with spillovers to core markets persisted" (source: Bank of England, Financial Policy Committee, September 2026). AI stocks fell again in October, before the Nasdaq jumped about 1% in trading on October 26 on enthusiasm for Anthropic's IPO (source: CNBC, October 26, 2026). Watch spending plans, free cash flow, credit and new listings for direction.
Why do neoclouds matter in the AI bubble debate?
Neoclouds sit where AI demand meets borrowed money. CoreWeave is the leading one, and its contracts with OpenAI totaled up to about $22.4 billion as of September 2025 (source: CoreWeave). Many neoclouds borrow heavily to buy chips, which makes their bond prices a closely watched sign of AI credit risk.
What happens if the AI bubble bursts?
Official bodies have tested that as a scenario, not a forecast. In July 2026 the Bank of England modeled a 45% fall in the US stock market over six quarters, with corporate credit spreads 3.5 percentage points wider. It found this would cut UK output by 2.2 percentage points, mostly through financial markets, with wider credit spreads accounting for about half of the hit (source: Bank of England, Financial Stability Report, July 2026).
How expensive is Nvidia stock?
Nvidia traded at about 44 times its trailing earnings at the close on October 20, 2026, with a market value of about $5.2 trillion. It made up about 7.6% of the Invesco QQQ fund at the end of June (source: 24/7 Wall St.). Its earnings are also growing fast, which is why investors disagree about whether that multiple is high, as whether Nvidia is a good investment sets out.
Researched and written by the BloFin Academy editorial team with AI-assisted drafting. All facts independently verified. Primary sources include the Federal Reserve, the Bank of England, the IMF, FactSet, company filings and releases from Meta, Microsoft, Nvidia, AMD and Anthropic, Robert Shiller, CNBC, Reuters and BloFin, current as of October 27, 2026.
Nothing in this article constitutes financial advice, and nothing in it is a recommendation to buy, sell or hold any stock, fund or crypto asset. Views quoted from investors and official bodies are theirs, and scenarios are not forecasts. AI stocks and bitcoin can fall sharply on the same day, stock perpetuals may accept only reduce-only orders outside US regular trading hours, and a leveraged position can be liquidated and lose all the margin behind it. BloFin services are not available in restricted locations, including the United States. Past performance does not indicate future results. Do your own research and consider your risk tolerance before you trade on BloFin.