In the last two weeks of October 2026, the conversation around artificial intelligence shifted from pure exuberance to open worry. Anthropic filed for what could become the largest IPO in history. Meta sought tens of billions more for its AI buildout. SpaceX locked in $40 billion to lease chips. On a single day, roughly $235 billion in AI-related deals crossed the wires—almost 1 percent of U.S. GDP. Yet the same week, Chinese AI and semiconductor stocks continued a brutal decline, and U.S. markets registered fresh losses in the companies most tied to the boom.
The question is no longer whether AI is transformative. It is whether the scale of capital now chasing that transformation has outrun the cash it can generate in the near term.
What has already broken in China
China’s AI equity bubble has deflated hard. The STAR50 index of AI and semiconductor leaders fell about 40 percent in less than four months, erasing roughly $480 billion in market value. Shares of robot-maker Unitree, chip firms MetaX and Moore Threads, and other retail favorites have halved or more from their peaks. China’s broader blue-chip index is down about 20 percent from its June high.
The reversal followed a rally that pushed AI-related stocks to dominate both market capitalization and trading volume. When demand doubts and export restrictions hit, the speculative layer collapsed first. Western markets have so far seen milder drawdowns—Nvidia and Oracle down roughly 10 percent from recent highs—but the Chinese experience is a live demonstration of how quickly sentiment can reverse when returns lag expectations.
The U.S. picture: enormous spending, lagging cash flow
Hyperscalers and AI infrastructure firms continue to spend at historic rates. Estimates for AI-related capital expenditure in 2026 sit in the $700–800 billion range, with projections for the next two years pushing past $1 trillion. Amazon, Microsoft, Google, Meta and others have raised hundreds of billions in new debt to fund data centers and chips. Some specialized “neocloud” providers that lease GPU capacity have seen borrowing costs spike and share prices crater as investors question whether the interest bills can be covered.
On the revenue side, the gap is visible. Combined 2026 revenue for OpenAI and Anthropic is estimated in the low-to-mid tens of billions to around $100 billion range depending on the source—material growth, yet still a fraction of the infrastructure capital being deployed behind them. OpenAI has already trimmed the valuation it is seeking in its latest funding talks from $1.4 trillion to $1.25 trillion while chasing roughly $30 billion. Anthropic’s IPO filing arrives against this more cautious backdrop.
Michael Burry has argued that a single season of revenue disappointment could be enough to force boards to cut spending. Ken Griffin of Citadel has said the market is “almost certainly” in a bubble. Former SEC chair Gary Gensler has publicly warned of a potential crash within 18 to 30 months. The Bank for International Settlements and the IMF have both flagged the risk that a sharp pullback in AI investment could transmit into the broader economy.
Why this is not a simple replay of 2000
The comparison to the dot-com crash is inevitable and imperfect. Many of today’s largest AI beneficiaries—Microsoft, Alphabet, Amazon, Meta—are highly profitable cash generators with diversified businesses. Valuations, while elevated, sit well below the peak multiples of the late 1990s for the sector as a whole. Real enterprise adoption is underway; the technology is not vaporware.
What looks more bubble-like is the infrastructure layer: the circular financing (chipmakers investing in the companies that buy their chips), the rapid rise in leverage among pure-play data-center and GPU lessors, and the assumption that revenue will scale linearly with capital expenditure for years. When that assumption is tested—whether by slower enterprise spending, open-source competition, power constraints, or simply slower model progress—the companies with the weakest balance sheets feel it first.
What to watch next
Three signals matter more than any single headline:
- Capex guidance from the hyperscalers. A material cut in 2027 spending plans would confirm that boards are recalibrating.
- Credit spreads and refinancing costs for neocloud and data-center pure plays. Rising yields already signal investor skepticism; further widening would tighten the funding loop.
- Actual AI software and services revenue growth versus the infrastructure build. The wider the gap persists, the more pressure mounts on valuations.
A full-blown “burst” that erases the majority of AI-related market value is not the base case for most serious observers. A multi-year period of slower growth, rotating leadership, and selective write-downs is more plausible. Some analysts describe it as a series of “rolling bubbles” across different parts of the stack rather than one dramatic pop.
The practical takeaway
AI capability will keep advancing. The companies that own the models, the distribution, and the most efficient infrastructure will still capture large economic value. What is under pressure is the idea that every dollar spent on AI compute today will produce proportional returns on the same timeline investors have priced in.
For founders and operators, the shift already favors capital efficiency and clear paths to revenue over pure scale. For investors, the distinction between cash-flow-positive platforms and leveraged infrastructure plays has become sharper. For everyone else watching the market, the Chinese correction and the recent U.S. volatility are useful reminders that even genuine technological revolutions produce financial excess—and that excess eventually gets corrected.
The technology is real. The question of 2026 is how much of the current capital structure survives contact with reality.
Sources
- NPR / 1A, 28 Oct 2026: China AI bubble and U.S. deal activity
- Reuters / Tribune, late Oct 2026: Chinese STAR50 decline and market value loss
- Economic Times, Sep 2026: Michael Burry commentary
- The Telegraph, 24 Oct 2026: CoreWeave and neocloud sell-offs
- Various reports on OpenAI valuation talks and Anthropic IPO filing (Oct 2026)
- BIS annual report and IMF commentary on AI investment risks
0 Comments