Nvidia shares fall as AI data centers spend more on memory chips

In short: Nvidia’s stock has slipped even as AI spending continues, because the tightest supply for data centers is shifting from GPUs to memory chips.

What's going on

Nvidia’s stock price has fallen about 15% since its peak in May, even though projected revenue is still expected to grow. TechCrunch notes that by some measures, investors are now paying less for each dollar of Nvidia’s expected profit than they pay for the average large US company.

At the same time, money flowing into “AI infrastructure” stocks is increasingly going to memory makers. Micron, a major maker of DRAM (a common type of computer memory), has nearly tripled in value over the same period, according to the report.

The reason is basic supply and demand. The GPU shortage that worried buyers last year has eased, which has helped push down the spot price of renting GPU time. TechCrunch cites Ornn data showing the spot price for an hour on an Nvidia H100 GPU peaked around May and then declined, from roughly $3.20 an hour.

Memory is moving the other way. TechCrunch cites Datatrack data showing DRAM spot prices jumping about 10 times starting around August 2025. High-bandwidth memory, which is extra fast memory used in AI servers (like a wider on-ramp to a highway), has become a key pinch point.

What to watch

More large tech companies are building their own AI processors to reduce reliance on Nvidia, including Google, Amazon, Microsoft, and OpenAI, according to the article. Ornn CTO Wayne Nelms said no one is making their own DRAM at the same pace, so memory prices could stay high unless supply grows or there is a major technical change.

Source: TechCrunch AI