The AI Industry Is Running Out of Compute, with Outages, Rationing, and Rising GPU Prices

Why It Matters

The compute shortage threatens the reliability of leading AI services and forces firms to prioritize high‑margin products, accelerating cost pressures across the industry.

Key Takeaways

• Anthropic's Claude API uptime fell to 98.95% in 90 days.
• OpenAI shut down Sora to reallocate compute to coding tools.
• Nvidia Blackwell GPU hour price rose 48% to $4.08.
• Providers impose token limits and new pricing tiers for heavy workloads.
• Coreweave raised AI cloud prices over 20% and demands three‑year contracts.

Pulse Analysis

The rapid expansion of agentic AI—software that can act autonomously—has outpaced the hardware pipelines that power it. Providers like Anthropic are grappling with frequent outages, as evidenced by a 98.95% Claude API uptime over the last 90 days, well below the 99.99% benchmark expected of cloud services. OpenAI’s decision to retire its Sora video‑generation app underscores a strategic shift toward higher‑margin, compute‑intensive offerings such as the new Spud model, highlighting how scarcity forces firms to reallocate resources away from experimental products.

At the same time, the hardware bottleneck is inflating costs across the AI stack. Nvidia’s latest‑generation Blackwell GPU now commands $4.08 per hour, a 48% increase from two months earlier, according to the Ornn Compute Price Index. Cloud operators like Coreweave have lifted prices by more than 20% and are requiring three‑year contracts to lock in demand, while Vultr’s CEO warns that power and data‑center capacity are already pledged through 2026. These price hikes are accompanied by stricter usage policies: GitHub Copilot, OpenAI Codex, and Anthropic have introduced token caps, tiered pricing, and off‑peak incentives to smooth demand spikes.

The broader market implication is a recalibration of AI economics. As compute becomes a premium commodity, firms will prioritize revenue‑generating workloads, potentially slowing the rollout of exploratory tools. Investors may see increased capital allocation toward GPU manufacturing and specialized data‑center builds, while smaller players could face higher barriers to entry. In the short term, customers can expect more granular billing, longer contract commitments, and a shift toward hybrid strategies that blend on‑premise resources with cloud bursts to mitigate the ongoing shortage.