Breaking

Nvidia Shares Down 15% From May Peak as Custom Chips, Memory Costs Reshape Market

July 9, 2026 at 13:08 EDT

Infra & Chips
Funding & M&A

Nvidia, which pioneered the market for AI-oriented GPU compute, is now being pressured by shifts in that very market. Its shares have fallen 15% from a May peak, and its valuation against expected earnings has slipped below the S&P average, TechCrunch reported on July 9, 2026.

July 9, 2026 · AI Compute Market

Nvidia, victim of the compute marketplace it created

The value Nvidia proved drew in rivals building custom chips. As GPU supply grew and prices fell, the bottleneck — and the profits — began shifting toward memory.

−15%

Nvidia shares off their May peak

~3×

Micron & memory makers surged in the same period

~50%

Cost cut targeted by OpenAI's "Jalapeño" vs standard GPU

H100 GPU spot price is sliding

Hourly spot price ($/hr) — supply eased as custom chips flooded in

$3.20

May peak

→

$2.46

recent high

$1.70

recent low

Hyperscalers build their own silicon

Google

TPU (7th gen+)

For its own cloud

Amazon

Trainium

Over 1M deployed on AWS · training + inference

Microsoft

Maia 200

Azure · for GPT-5.2 and Copilot

OpenAI

Jalapeño

Inference-focused · ~50% cost cut vs standard GPU

Nvidia's moat holds

The CUDA ecosystem remains dominant, and developers still cite the high cost of migrating away from Nvidia's environment.

Compute is commoditizing

Falling prices ease new entry, and GPU compute futures contracts are being planned — treating raw compute as a tradable commodity.

The turning point

The bottleneck is shifting from GPUs to memory.

No one makes the HBM whose data-center demand is exploding in-house, so supply can't keep up — the imbalance persists until an HBM breakthrough or a shift in supply and demand.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.