Micron, Nvidia face renewed pressure as AI spending fears swell

The AI trade had another rough day. On October 15, Micron shares fell 4% to 4.5%, while Nvidia slipped between 1.3% and 1.8%.

Those moves capped a brutal stretch for chipmakers. Over the course of the week, 20 major AI and semiconductor stocks lost over $1 trillion in combined market value.

What set off the latest sell-off

The pressure built sharply between October 13 and 15. The trigger was a Financial Times report on OpenAI’s finances.

According to that report, OpenAI’s annualized revenue run rate stood at approximately $50 billion as of September 30. The market had been expecting something in the range of $68 to $70 billion.

The damage spread across the sector. AMD declined approximately 2% to 3.5% in premarket trading on October 15. The Philadelphia Semiconductor Index, the benchmark for chip stocks, has been under pressure for multiple consecutive sessions.

Rates, debt and a competitor from China

Rising US inflation has fueled expectations of higher interest rates. CoreWeave, the AI cloud provider, has seen its bond yields soar as investors demand more compensation for holding its debt.

On October 12, Chinese memory maker ChangXin Memory Technologies, known as CXMT, announced advancements in DRAM technology. DRAM is the working memory that sits inside everything from laptops to AI servers. It is also Micron’s core business.

Strong results, nervous investors

Micron has achieved record gross margins. It has also secured significant contracts to supply high-bandwidth memory, or HBM, in 2026.

The spending forecasts themselves remain staggering. US hyperscalers are projected to spend around $800 billion on AI-related capital expenditures in 2026. That figure is projected to rise to about $1.1 trillion in 2027.

What this means for chip investors

The core tension is between near-term demand and long-term return. The OpenAI revenue gap suggests the monetization side may be running behind the build-out.

For Micron specifically, the setup is two-sided. Record margins and locked-in HBM contracts for 2026 offer visibility. Yet memory has historically been a cyclical business, which is why talk of future oversupply and CXMT’s DRAM progress hits the stock harder than its peers.

CoreWeave’s soaring bond yields are a signal of how credit markets view AI companies that have borrowed heavily against future demand. If financing conditions keep tightening, that stress could spread to their suppliers through slower orders or renegotiated commitments.

What to watch from here comes down to a few variables. US inflation data will shape rate expectations and, by extension, the cost of funding data centers. Hyperscaler capital spending guidance will show whether the projected $800 billion for 2026 holds up. And any further updates on AI revenue, from OpenAI or others, will test whether this week’s fear was an overreaction or an early warning.