Nvidia shares have dropped 15% since their May peak, even as projected revenue climbs. The company now trades cheaper than the S&P average, meaning investors pay less for every dollar of Nvidia’s expected profit than they do for a typical large American firm.
Capital continues to flow into AI infrastructure, but it is mostly buying memory stocks. Over the same period, Micron has nearly tripled in value. This maker of DRAM chips, found in standard computers and servers, has established memory as the new bottleneck for data centers and the hottest trade in AI.
The reason is straightforward. The GPU shortage that alarmed the industry last year has eased. At the same time, data centers require all the memory available.
For those who appreciate Nvidia’s technological achievements, the market reaction feels deflating. There is genuinely impressive technology behind the rise, both in developing CUDA, the programming platform that made Nvidia GPUs the default engine for AI research, and in pushing GPU development at a speed few thought possible. Nvidia’s success is the kind of thing you can write whole books about. The GPUs themselves are among the most complex devices ever produced, right at the bleeding edge of human capability.
For memory companies like Micron, the story is simpler. They build high-bandwidth memory chips, specialised components designed to move data in and out of processors as fast as possible. These components have improved incrementally for 20 years. Without the chips or the companies changing too much, the service they provide suddenly became very valuable. Since demand grows faster than anyone can scale up supply, prices have risen tenfold over the past year.
This, via Datatrack, is what the spot price for DRAM looks like since 2023:
You might think there was an amazing technical breakthrough in the summer of 2025. No. The industry as a whole just vastly underestimated how much memory it would need for the data center buildout.
In comparison, this (via the compute marketplace Ornn) is how the spot price for an hour of time on an Nvidia H100 GPU has changed over the last year:
Just like Nvidia’s stock price, there is a peak in May, around $3.20 an hour, followed by a steady drop-off. For better or worse, Nvidia’s value as a company is tied to the price of compute, and that price is falling. Micron and its cohort are tied to the price of DRAM, and that price keeps rising.
When I spoke to Ornn co-founder and CTO Wayne Nelms about the forces driving that disparity, he framed it as a simple issue of supply and demand. Google, Amazon, Microsoft, and even OpenAI have launched their own custom processors to lessen their dependence on Nvidia. Even if those chips are not as good as the latest model from Nvidia, they are good enough to drive down the price of compute.
“More GPU and accelerator players are entering the market. Everyone wants to make their own silicon, but no one is making their own DRAM,” Nelms said. “Until there’s a major technological breakthrough on HBM, a shift in supply and demand, or someone new enters the market in memory, I think things will more or less persist as we see today.”
What it means
For people making things with AI, the dynamic is shifting. The tool that once defined the field is becoming commoditised as others build alternatives. Meanwhile, the raw memory required to run those tools has become the scarcest and most expensive resource. Builders will find compute easier to source, but memory remains the primary cost driver.
