Nvidia: Trapped by its own compute market.

Nvidia, long revered as a trailblazer in its industry, has experienced a challenging period recently. Despite its projected revenue continuing to climb, the company's stock price has declined by 15% since its peak in May, as detailed by Bloomberg. This downturn means that, relative to its expected earnings, Nvidia is now more affordable than the average S&P company, with investors paying less per dollar of anticipated profit compared to a typical large American corporation.

Concurrently, while capital continues to flow into AI infrastructure, the primary beneficiaries are memory companies. Over the same timeframe, Micron, a leading producer of DRAM—the standard memory chip in computers and servers—has seen its value nearly triple. This surge has positioned memory as the critical new bottleneck for data centers and a highly sought-after investment in the AI sector. The underlying reason is straightforward: the previously alarming GPU shortage has somewhat alleviated, while data centers now require an unprecedented volume of memory.

For those who admire Nvidia's technological prowess, this market shift can feel somewhat disheartening. Nvidia's ascent is underpinned by genuinely impressive innovations, from the development of CUDA—its widely adopted programming platform that established Nvidia GPUs as the foundational engine for AI research—to its relentless acceleration of GPU development, achieving speeds many once deemed impossible. Nvidia's achievements are profound enough to fill entire books, and its GPUs represent some of the most intricate devices ever created, pushing the boundaries of human engineering.

In contrast, the narrative for memory manufacturers like Micron is considerably simpler. These companies produce high-bandwidth memory (HBM) chips, specialized components engineered to facilitate the fastest possible data transfer to and from processors. While these chips have undergone incremental improvements over two decades, their core function and the companies producing them haven't drastically changed. Yet, the service they provide has suddenly become immensely valuable. With demand outstripping the industry's capacity to scale supply, these companies have managed to increase prices tenfold over the past year.

Data from Datatrack illustrates the significant rise in the spot price for DRAM—the open-market price for chips, distinct from long-term contract rates—since 2023.

One might infer that a groundbreaking technical innovation occurred around mid-2025, but in reality, the industry collectively underestimated the sheer volume of memory required for the extensive data center buildout.

Conversely, the compute marketplace Ornn reveals a different trend for the spot price of an hour of time on an Nvidia H100 GPU over the past year.

Mirroring Nvidia's stock performance, the H100 GPU spot price peaked in May at approximately $3.20 per hour before experiencing a consistent decline. For better or worse, Nvidia's corporate valuation is intrinsically linked to the price of compute, which is currently falling. Meanwhile, Micron and its peers are tied to the price of DRAM, a commodity that continues its upward trajectory.

When discussing the factors driving this divergence with Wayne Nelms, co-founder and CTO of Ornn, he characterized it as a fundamental issue of supply and demand. Major tech giants such as Google, Amazon, Microsoft, and even OpenAI have launched their own custom processors to reduce their reliance on Nvidia. Even if these proprietary chips don't quite match the performance of Nvidia's latest models, they are sufficiently capable to exert downward pressure on compute prices.

"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 stated. He further predicted, "Until there’s a major technological breakthrough on HBM [high-bandwidth memory], 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."

This situation presents a frustrating paradox for Nvidia, largely a consequence of its own formidable success. By demonstrating the immense value of compute, the company now finds itself at the core of a highly competitive market that everyone wants to enter, while businesses focused on simpler technologies and seemingly less innovative products are accumulating substantial wealth on the periphery.