Nvidia Loses Ground: Memory Shortage Reshapes the AI Computing Market

Nvidia on July 9, 2026, found itself in an unusual position: despite growing projected revenue, its stock fell 15% from its May peak, while memory manufacturer Micron nearly tripled its market value. The reason is a weakening shortage of graphics processors and a new shortage of memory for data centers, which is reshaping the distribution of funds in artificial intelligence infrastructure.

Why Nvidia is Getting Cheaper

The market has begun paying less for computing on Nvidia graphics processors, even though demand for artificial intelligence infrastructure remains strong. According to data cited by TechCrunch, the company's shares lost 15% from their May high; simultaneously, its expected profit is valued by the market lower than the average large American company in the S&P index. This does not mean revenue is falling: forecasts for it continue to grow. But investors see that the main resource of the AI boom—access to GPUs—is ceasing to be as scarce as it was a year ago.

• On July 9, 2026, TechCrunch reported a 15% decline in Nvidia stock from its May 2026 peak.
• Nvidia H100's hourly spot price reached around $3.20 in May 2026, then steadily declined.
• Micron's market value nearly tripled over the same period.
• Memory manufacturers have been able to raise DRAM prices roughly tenfold over the past year.

Nvidia created the technological foundation of today's market: its CUDA software platform made its graphics processors the standard tool for AI model research and training. The company also rapidly updated GPU lineups, which rank among the most complex mass-produced electronic devices. However, this very success of the strategy attracted competitors: cloud platforms and laboratories began seeking ways to reduce dependence on a single supplier.

What Became the New Bottleneck

Memory, not the accelerators themselves, became a new constraint for data centers. Micron manufactures DRAM and high-bandwidth memory HBM—components that rapidly transfer data between processor and memory. These technologies have been developing for a long time, but the industry underestimated how much memory would be needed for mass construction of AI data centers. Demand exceeded production expansion rates, so spot market prices rose sharply.

This is an important shift: the cost of renting Nvidia H100 computing is falling, while DRAM prices are rising. Data center owners can acquire more accelerators or rent them cheaper, but still hit up against memory availability. For Micron and comparable manufacturers, familiar technology and limited supply are sufficient to profit from the new wave of AI infrastructure capital spending.

Spot price reflects immediate-delivery transactions, not long-term contracts. In the Datatrack chart presented in the material, the DRAM price spike begins in August 2025. The author emphasizes: it is not tied to a sudden breakthrough in memory technology. The market simply evaluated data center demand too late.

Why Computing is Getting Cheaper

The supply of accelerators is expanding in two directions at once: more GPU manufacturers and accelerators are entering the market, and largest buyers are developing their own processors. Google, Amazon, Microsoft, and OpenAI have already released custom chips to reduce dependence on Nvidia. Such solutions may lag behind Nvidia's latest products, but their capabilities are sufficient for some tasks and to press computing prices down.

Wayne Nelms, co-founder and technical director of the Ornn platform, describes the situation as ordinary supply and demand dynamics. According to him, companies want to develop their own silicon, but almost nobody manufactures their own DRAM. Until there is a notable technological breakthrough in HBM, a shift in supply and demand balance, or a new major player in memory emerges, the disproportion may persist.

"Everyone wants to make their own silicon, but nobody makes their own DRAM," —

Wayne Nelms, co-founder and technical director of Ornn.

For Nvidia, this does not negate technological leadership. Rather, the market is turning GPUs from a scarce resource into a more competitive commodity. When there are more alternatives, superprofits shift to the next scarce element in the system—memory. This also shows that valuation of AI companies and infrastructure suppliers increasingly depends not only on chip quality, but also on the specific constraint in the supply chain.

What This Means

The AI boom has not ended, but its financial center is shifting: Nvidia remains a key supplier of accelerators, while short-term profits are increasingly going to memory manufacturers. For developers and data center operators, this means that the cost and availability of HBM and DRAM are becoming as important as GPU selection.