November 5 was an ordinary Thursday, but Wall Street experienced a textbook case of a major shift in market sentiment. U.S. stocks lost $3.2 trillion in market value overnight—more than the entire crypto market capitalization of $2.5 trillion. In other words, the market wiped out the equivalent of 1.2 entire crypto markets in a single night.
The sell-off was triggered by a sharp collapse in AI-related stocks, setting off a chain reaction across the broader market. Nvidia plunged 9.3% in a single session, wiping out $446.6 billion in market capitalization—approximately RMB 2.98 trillion, roughly equivalent to the market value of Agricultural Bank of China disappearing in one day. AMD fell 17.3%, SanDisk dropped 22%, while Micron and Western Digital each lost more than 17%. Combined, the so-called “Magnificent Seven” lost approximately $1.04 trillion in market value.
The Philadelphia Semiconductor Index crashed 12.25%, dragging down the tech-heavy Nasdaq Composite by 5.52% and the Nasdaq 100 by 6.36%. Both recorded their worst single-day performances since the consecutive market circuit breakers of March 2020.
The U.S. stock market rout quickly spread across global markets, and Bitcoin was no exception. Within a short period, BTC plunged from $81,000 to $75,000. The entire crypto market lost approximately $140 billion in market capitalization, while 24-hour liquidations reached $1.26 billion. More than 240,000 traders were liquidated, with long positions accounting for as much as 88% of the total.
Overnight, everyone was asking the same question: Is the AI bubble finally starting to burst, just like the dot-com bubble around 2000？
Democrats Win Midterm Elections, Triggering a Major AI Valuation Repricing
Looking only at the price declines, it would be easy to dismiss this “Black Thursday” as another tech stock correction. But what really deserves closer examination is why it happened on this particular day, why the AI supply chain was hit so hard, and why the losses were so severe.
The immediate trigger was neither corporate earnings nor the Federal Reserve. It was the ballot box, following the November 4 elections.
The U.S. midterm election results came in, with Democrats winning control of both chambers of Congress and securing victory in the Texas gubernatorial race. Gina Iñiguez, the newly elected governor of Texas, explicitly stated that approvals for data center projects would be suspended until legislation protecting local residents was introduced.
Texas is one of the largest data center construction hubs in the United States, with approximately 125 GW of planned project capacity, accounting for roughly 27% of the national pipeline. Meanwhile, Democratic Representative-elect Yassamin Ansari publicly stated that the party's top priority in the next Congress would be to pause the development of advanced artificial intelligence and impose a ban on “superintelligence.”
Taken together, these developments struck at the most vulnerable point in the AI investment narrative: the pace of AI infrastructure expansion is not determined solely by companies' willingness to spend. It is heavily dependent on regulatory approvals and the broader policy environment.
For months, markets had been pricing semiconductor stocks, data center operators, and power equipment suppliers on the assumption that the AI computing arms race was unstoppable. Suddenly, policy risks moved from background noise to center stage.
UBS strategists offered a straightforward estimate following the sell-off: approximately 30% of the additional data center capacity expected to come online in the United States by 2028 could be affected by potential construction suspensions.
For an AI industry whose valuations depend heavily on the chain of “new data centers → new chip orders → new revenue,” this is no minor disruption. It represents a direct downward revision of future revenue expectations.
The AI Bubble May Be Approaching a Tipping Point, but This Won't Be a Repeat of the 2000 Dot-Com Crash
Ray Dalio's assessment, made just days before the sell-off, has been widely cited: AI has become a “classic bubble,” and the market is approaching a critical point at which that bubble could burst.
His central argument is that AI infrastructure requires astronomical levels of capital investment, with an increasing proportion of funding coming from debt financing.
Data cited by Michael Burry offers an even more striking comparison: approximately 38% of current high-yield bond issuance is related to AI, compared with a 40%–50% share for TMT bonds during the dot-com bubble around 2000.
These warning signs are real and should not be dismissed. But jumping straight to the conclusion that “AI is simply repeating the year 2000 and the bubble is about to burst” would be an equally simplistic way of looking at the market.
A more accurate description of today's AI trade is this: supply is expanding at an extraordinary pace, while demand is being absorbed at a more conventional rate. The gap between the two is no longer just a theoretical concern. It has become a real risk that can be repriced by policy developments. However, the industry itself is still far from reaching the point where its underlying demand has been disproven.
Consider this analogy: the internet bubble of 2000 was like a glass of water made almost entirely of foam. Fiber-optic networks had been built, but nobody was using them. Websites attracted traffic but generated little revenue. “Eyeballs” were treated as valuable assets in their own right.
Today's AI bubble is more like a glass of beer with a thick layer of foam on top.
The foam represents questionable AI projects, overvalued speculative stocks, and data center developers raising capital primarily on the strength of the AI narrative. But beneath that foam is real beer: businesses with actual revenue, practical applications, and genuine industrial value.
Microsoft's Azure grew 48% in a single quarter, AWS expanded 39%, and Google Cloud surged 103%. Cloud revenue among hyperscalers has accelerated for six consecutive quarters. Demand is real, and so are the bills customers are paying.
The problem is that even if the beer itself is excellent, once a needle bursts the foam on top, the liquid inside can still spill out and splash everyone around it.
2000 vs. 2026: The Similarities Are Striking, but So Are the Critical Differences
The most honest conclusion when comparing today's market with 2000 is this: the similarities are alarming enough to warrant concern, but the differences are significant enough to undermine the narrative that history is simply repeating itself.
The similarities are primarily concentrated in debt structures. Nearly 40% of high-yield bond issuance is now AI-related, approaching levels seen during the dot-com bubble. The AI ecosystem could generate more than $2 trillion in debt financing demand, while the investment-grade corporate bond market may only be able to absorb less than $1 trillion of that demand by 2030. This could leave a funding gap exceeding $1 trillion, with private credit widely viewed as a leading candidate to fill it.
Even more concerning is the growing use of off-balance-sheet special purpose vehicles (SPVs). Meta and Blue Owl, for example, established a joint venture that issued $27.294 billion in bonds to finance data center construction. Once completed, the facilities are leased back to Meta. Under the applicable accounting treatment, this debt does not appear on Meta's balance sheet.
This structure—moving risk off corporate balance sheets while relying on the underlying investment narrative to sustain creditworthiness—does bring to mind the layers of securitization that accumulated before the 2007 subprime mortgage crisis, rather than a straightforward stock market bubble like the one in 2000.
One analyst put it bluntly: this time, the situation looks more like 2007, except that the collateral has changed from houses to GPUs—and GPUs can lose more than half their value within just three years.
However, the differences are equally decisive.
First, the direction of the interest rate cycle is reversed. The 2000 bubble burst after the Federal Reserve raised interest rates six consecutive times, adding a total of 1.75 percentage points and pushing the federal funds rate to 6.5%. The resulting liquidity squeeze directly undermined technology stock valuations. Today, although the Fed maintains a hawkish stance and has not ruled out further rate hikes this year, the broader rate cycle is still moving down from elevated levels. Real interest rates are therefore much less restrictive toward long-term returns than they were in 2000.
Second, the earnings foundations are different. The forward P/E ratio of today's five largest technology companies is approximately 23, compared with around 65 near the peak of the dot-com bubble in 2000—nearly three times higher.
Third, the bubbles are at different stages of their life cycles. After fitting various bubble indicators to current AI market conditions, several institutions have concluded that the AI trade is roughly in the “early stages of euphoria,” making it more comparable to 1998 than to the first quarter of 2000.
The real danger lies in a potential path to a bubble bursting that is different from the dot-com scenario.
In 2000, the sequence was: disappointing earnings → Fed rate hikes → valuation collapse.
Today's more likely risk scenario is: policy shock (such as the recent election) → downward revisions to capital expenditure expectations → tighter financing conditions → widening credit spreads → earnings downgrades → valuation repricing.
The starting point of this chain reaction may not be the stock market at all, but the credit market.
Rather than Nvidia's next earnings report, the earliest warning sign could be a sustained widening in the credit spreads of an AI-related bond. In other words, lenders may begin demanding significantly higher interest rates before they are willing to lend money to AI-related companies.
Bitcoin Is Not a Safe Haven but a “Leveraged Nasdaq”: Where Does the Market Go From Here？
Bitcoin is increasingly demonstrating through its actual market performance that it currently behaves more like a high-beta technology asset than digital gold.
The November 5 crash further exposed the close relationship between U.S. equities and the crypto market. Bitcoin's 30-day correlation with the Nasdaq 100 has climbed to 0.68, indicating a clear tendency to move in tandem.
This means that when the AI investment narrative weakens and technology stocks come under selling pressure, Bitcoin is unlikely to escape unscathed.
When institutions reduce risk exposure, they often prioritize selling highly liquid assets with concentrated leverage. BTC happens to possess both characteristics.
A sharp decline in U.S. equities triggers a broader contraction in risk appetite, which in turn forces Bitcoin long positions to close. This creates a negative feedback loop:
U.S. stocks fall → risk exposure shrinks → Bitcoin liquidations accelerate → BTC prices fall further.
So, where does the market go from here？
The direction of U.S. equities over the coming weeks and months will depend largely on policy signals and corporate capital expenditure.
If U.S. regulatory policy becomes more pragmatic and concerns surrounding data center approvals begin to ease, technology stocks could stage a technical rebound.
However, if major technology companies cut their AI capital expenditure forecasts for 2026–2027, or if credit spreads on AI-related bonds continue to widen, the market correction could deepen.
Goldman Sachs has pointed out that approximately one-sixth of S&P 500 constituents have already fallen more than 50% from their respective peaks. The index's heavy dependence on a handful of technology giants means concentration risk cannot be ignored.
On the other hand, Morgan Stanley has found that hedge fund exposure to AI-related stocks has fallen to its first net-underweight position since early 2023. This suggests that crowded trades are being unwound. Once market sentiment stabilizes, short-covering could potentially fuel a rebound.
Bitcoin's medium-term outlook is also heavily influenced by the AI narrative and Federal Reserve policy.
As expectations of further rate hikes remain elevated, BTC is likely to continue behaving like a high-beta risk asset. ETF flows, regulatory developments, and the post-halving supply structure can all influence its long-term trajectory. However, when systemic risks emerge, these factors may not be enough to offset the downward pressure coming from a declining U.S. stock market.
So, what should investors do under these circumstances？
The core principle can be summed up in one sentence: Rather than trying to predict exactly when the bubble will burst, manage your risks before it happens.
First, reduce asset concentration. Never put all your eggs in one basket.
Second, increase allocations to cash and short-term bonds. This preserves the flexibility to re-enter the market at lower prices. Gold can also be considered as a portfolio hedge.
Investors experienced in derivatives may even consider using put options to hedge their exposure. The objective is not to bet on a market crash, but to protect existing positions at a manageable cost.
Finally, and most importantly, control leverage. With Bitcoin's correlation to the Nasdaq reaching 0.68, investors should not casually assume that crypto can rally independently of the broader technology market.
Long-term holders should ensure their portfolios can withstand a further 30%–40% decline in BTC. Short-term dip buyers, meanwhile, need to understand the difference between an asset being cheap and a position being safe.
The most important thing in the current market is not predicting when the AI bubble will burst, but preparing for the possibility that the correction could continue.
For investors in both U.S. equities and crypto, managing concentration, reducing leverage, and preserving liquidity may prove far more important than accurately predicting the next rebound.
Returning to the question in the headline: Is the AI bubble starting to burst？
The more honest answer is that the outermost layer of the bubble was already punctured by policy uncertainty on November 5, 2026, but the “beer” underneath has not been discredited.
The lesson from 2000 is clear: When a market narrative becomes unquestionable, even the smallest crack can be amplified by the pricing mechanism.
The real message behind yesterday's AI stock collapse is that the market has begun demanding a risk premium for two factors that were previously underestimated in the AI trade: policy dependence and financing structures.
For investors, the most valuable exercise right now is not simply predicting the date when the bubble might burst. More importantly, it is taking an honest look at their own portfolios.
If another policy shock of similar magnitude hits tomorrow, which holdings would you be forced to sell at the bottom？ And which positions have genuine resilience？
The answers will differ from investor to investor, but everyone needs to ask themselves the same question.