Nvidia is a victim of the compute marketplace it created

2026-07-09 · Source: TechCrunch · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Intermediate, short

Summary

Nvidia's stock price has fallen 15% since its May peak, despite continued revenue growth, making it cheaper relative to projected earnings than the S&P average. This decline contrasts sharply with the performance of memory companies like Micron, which has nearly tripled in value over the same period. The shift is driven by an easing GPU shortage and a new bottleneck in data centers: high-bandwidth memory (HBM). The spot price for DRAM has increased tenfold since August 2025 due to underestimated demand for data center buildouts. Concurrently, the spot price for an hour on an Nvidia H100 GPU peaked around \$3.20 in May and has since steadily dropped. This market disparity is attributed to increased competition in the GPU and accelerator market, with major players like Google, Amazon, Microsoft, and OpenAI developing their own custom processors, while the memory market lacks new entrants or significant technological breakthroughs.

Key takeaway

For CTOs and Directors of AI/ML evaluating infrastructure investments, recognize that the compute market dynamic has fundamentally shifted. Your focus should expand beyond GPU acquisition to critically assess high-bandwidth memory supply and pricing. The tenfold increase in DRAM prices and falling GPU compute costs indicate memory is now the primary bottleneck. This directly impacts your budget and scaling strategies for data center buildouts.

Key insights

The compute market, once dominated by GPUs, is now bottlenecked by high-bandwidth memory, shifting value to memory manufacturers.

Principles

• Market value shifts to the scarcest resource.
• Increased competition drives down compute prices.

In practice

• Monitor DRAM spot prices for data center buildout cost projections.
• Evaluate custom accelerator development against HBM supply constraints.

Topics

• NVIDIA
• GPU Market
• High-Bandwidth Memory
• DRAM Pricing
• Data Center Infrastructure
• Compute Economics

Best for: Investor, CTO, Director of AI/ML

Related on AIssential

• The AI Memory Tax & the Bifurcation of AI Scaling Laws — Memory has transitioned from a commodity to a strategic asset due to AI's high bandwidth demands.
• You Wouldn’t Buy a Car From One Dealer Without Checking Prices Elsewhere. — GPU procurement requires diligent benchmarking across providers to avoid significant cost overruns due to market inefficiencies.
• 📈 Data to start your week: AI boom, nowhere near the ceiling — AI compute demand significantly outpaces supply, driving up GPU prices and leading to rationing by providers.
• This chip startup just raised $135M on a bet that AI’s biggest bottleneck isn’t compute — it’s memory — AI inference is increasingly a memory scaling problem, not solely a compute challenge.
• Why GPU Server Prices Are Unpredictable — Limited component supply and high demand make GPU server prices unpredictable, impacting cloud provider costs and customer pricing.
• Beyond the NVIDIA's Tax — The AI silicon market is diversifying as hyperscalers and startups challenge NVIDIA's GPU dominance with specialized hardware.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by TechCrunch.