Long an industry leader, Nvidia has had a couple of bad months. Bloomberg it has the ugly detailsbut the result is that the company’s share price is down 15% from its peak in May, even as projected revenue continues to rise. Compared to expected earnings, the company is now cheaper than the S&P average. Investors are paying less per dollar of Nvidia’s projected earnings than they would for the typical major U.S. company.
Money is still flooding into AI infrastructure stocks, but mostly going into memory companies. In the same period, Micron — one of the world’s biggest makers of DRAM, the standard type of memory chip found in computers and servers — has nearly tripled in value, establishing memory as the new bottleneck for data centers and the hot new trade in artificial intelligence. The main reason is simple: The lack of GPUs that seemed so alarming last year has lessened a bit. At the same time, data centers need all the memory money can buy.
For anyone who appreciates Nvidia’s technological achievements, this might feel a little deflated. There’s a lot of really impressive technology behind Nvidia’s rise, both in the development of CUDA, its widely adopted programming platform that made Nvidia GPUs the engine of choice for AI research, and in pushing the pace of GPU development to a speed few thought possible. Nvidia’s success is the kind of thing you can write entire books about, and GPUs themselves are among the most complex devices ever made, right at the edge of human capability.
For memory companies like Micron, the story is much simpler. They make high-bandwidth memory chips — specialized components designed to move data in and out of processors as quickly as possible — that have been steadily improving for 20 years. Without the brands or companies changing too much, the service they provide has suddenly become valuable — and since demand is growing faster than anyone can increase supply, they’ve been able to raise prices tenfold in the past year.
This, via Datatrack, is the exact price for DRAM—the price buyers pay for chips on the open market, as opposed to long-term contract rates—as of 2023:
You might think there was some amazing technical breakthrough in the summer of 2025, but no, the industry as a whole vastly underestimated the amount of memory that would be needed to build the data center.
In comparison, this (via the Ornn computer market) is how the spot price for an hour on an Nvidia H100 GPU has changed over the last year:
Just like Nvidia’s stock price, there is a peak in May (about $3.20 an hour) and then a steady decline. For better or worse, Nvidia’s value as a company is tied to the price of computing, and that price is falling. Micron and its cohort are tied to the price of DRAM, and that price continues to rise.
When I spoke with Ornn co-founder and CTO Wayne Nelms about the forces driving this disparity, he framed it as a simple matter of supply and demand. Google, AmazonMicrosoft and more OpenAI have released their own custom processors to reduce their reliance on Nvidia. even if these chips are not as good as Nvidia’s latest model, they are good enough to reduce the price of computing.
“More GPU and accelerator players are entering the market. Everyone wants to make their own silicon, but nobody makes their own DRAM,” Nelms told me. “Until there is a major technological breakthrough at HBM [high-bandwidth memory]a change in supply and demand or someone new [enters the market in memory]I think things will more or less persist as we see today.”
It’s a frustrating situation for Nvidia, and largely a product of its own success. Having proven how valuable computing can be, the company is at the center of a market that everyone wants to be in — while simpler technologies and less interesting companies get rich on the sidelines.
When you purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
