What just happened
On 14 and 15 October 2026, the AI trade fell apart. Tech shares tumbled across Asia, Europe and the United States, and the numbers were brutal:
- The Philadelphia Semiconductor Index — the main gauge of chip stocks — erased more than $1.6 trillion in market value during an 11.6% slide over just six trading sessions.
- The Nasdaq dropped more than 3% over the same stretch.
- On 14 October alone, four AI-linked companies — CoreWeave, Nebius, Oracle and Dell — lost roughly $80 billion in combined market value. What made this selloff unusual is that it wasn't triggered by bad AI technology. Nobody discovered that AI had stopped working. Instead, investors started asking an uncomfortable question: how is all of this being paid for?
CoreWeave: the company at the centre of the storm
To understand the crash, you need to know one company: CoreWeave. It is a "neocloud" — a new type of cloud company that rents out computing power from Nvidia's AI chips to customers including Microsoft and OpenAI. Instead of building their own data centres, AI companies can simply rent GPU power from CoreWeave. The business is booming in revenue terms — analysts expect nearly $13 billion in revenue this year. But there is a catch: CoreWeave has never made a profit, and it has piled up roughly $50 billion in debt and obligations to fund its expansion. In the second quarter of 2026, it earned $2.58 billion in revenue while spending $6.42 billion on new infrastructure. It raised its 2026 capital-spending forecast to between $35 billion and $39 billion. In mid-October, the debt markets started flashing red:
- The yield on CoreWeave's bonds due in 2031 — already rated "junk" — spiked to about 16%, a distress-level number.
- The cost of insuring its debt against default peaked above 1,140 basis points on 15 October, settling near 1,038 — anything above 1,000 reads as distress to credit markets.
- Its co-founder told the Financial Times the company was seeking an investment-grade rating that would open the door to "unlimited" capital — and investors took the remark as a sign of just how much money the company needs. CoreWeave's CEO, Mike Intrator, pushed back in an internal memo, saying the company's $104.2 billion revenue backlog more than covers its debt and that markets can "misunderstand and act irrationally". But the damage was done.
The three fears driving the selloff
1. The debt mountain
CoreWeave is the extreme case, but the whole AI buildout runs on borrowed money. US tech giants are expected to spend more than $700 billion on AI infrastructure this year alone, and research firm Gartner estimates AI providers will spend nearly $1 trillion on AI infrastructure in 2026 — then another $6 trillion between 2027 and 2030. Gartner added a chilling warning: about 80% of neoclouds could shut down by 2030, and more than $3.5 trillion of planned AI spending could face impairment if providers keep investing at the current pace while enterprise and consumer AI revenue fails to keep up.
2. Interest rates
On 14 October, hotter-than-expected US inflation data raised the odds that the Federal Reserve would raise interest rates. Higher rates are poison for companies carrying huge debts, because borrowing gets more expensive. Yields on the debt of CoreWeave and Oracle maturing in six years surged to record levels of around 15%.
3. Chip competition and oversupply
Memory chip makers wobbled on fears of rising competition from China's CXMT, plus worries about an oversupply of memory chips. Broadcom's chief executive poured fuel on the fire, saying many AI investors "have no clue what they're investing in". Earlier in the month, the Financial Times reported that OpenAI's annualised revenue was $20 billion lower than previously signalled.
The selloff in one table: 14 October's worst-hit
| Company | 13 Oct close | 14 Oct close | One-day fall |
|---|---|---|---|
| CoreWeave | $77.81 | $65.78 | −15.46% |
| Dell Technologies | $572.47 | $507.93 | −11.27% |
| Nebius | $206.86 | $186.01 | −10.08% |
| Oracle | $143.20 | $130.55 | −8.83% |
| These are closing figures; intraday swings were even sharper. And a falling share price is not just a number on a screen — it makes future fundraising more expensive, because a company must issue more shares to raise the same amount of money. |
Is this the AI bubble bursting?
Honest answer: nobody knows. But history offers a useful parallel. The internet changed the world, yet many internet stocks crashed in 2000 and 2001 before the survivors recovered. A technology can keep growing while its stocks go through a brutal correction — excitement and share prices don't move in a straight line. The key distinction: the October selloff was driven by worries about debt and financing, not by a collapse in AI itself. AI usage keeps climbing. The question is whether the revenues will arrive fast enough to justify the trillions being borrowed and spent.
What it means for you
If you own a broad index fund, you already have some exposure to these companies, spread across hundreds of others. If you piled into AI-themed funds or individual AI stocks, you felt this much more sharply. This is a good moment to ask: what share of my money rides on a single theme, and would I be comfortable if it fell 30%? The lesson is boring but true: keep any single theme to a portion of your savings you can live with, and let diversification do its job.
Takeaway checklist
- What happened: AI and chip stocks crashed in mid-October 2026; the chip index alone lost over $1.6 trillion in six sessions.
- Why: fears about the massive debts funding the AI buildout, higher interest rates, and chip competition — not broken AI.
- CoreWeave is the flashpoint: $50bn debt, 16% bond yields, record default-insurance costs — despite $104bn of future revenue booked.
- The big unknown: Gartner estimates $1 trillion of AI infrastructure spending in 2026 alone; the debate is whether revenues will justify it.
- History says: technologies can boom while their stocks crash (see: the internet in 2000). The two are related, but not the same.
- Practical move: check how much of your portfolio depends on one theme, and diversify. Related reading: Semiconductors Explained: The Tiny Chips Running the World · Samsung's $80 Billion Quarter, Explained
