AI Industry Hits Compute Crisis as Token Demand Surges, GPU Costs Spike 48%, and War Disrupts Supply Chains
Summary
A perfect storm of surging AI token demand, a 48% spike in GPU costs, and war-driven semiconductor supply chain disruptions are pushing the AI industry to a breaking point, with OpenAI's token demand exploding from 6 million to 15 billion per minute and Anthropic restricting access to its latest model.
Key Points
- The AI industry is facing a severe compute crisis driven by three converging forces: the war in Iran is disrupting semiconductor supply chains, including critical helium supplies, while fueling inflation across logistics and shipping.
- AI agent adoption is exploding, with OpenAI's token demand skyrocketing from 6 million per minute in October 2025 to 15 billion per minute in March 2026, creating unprecedented and often inefficient demand for inference compute.
- Anthropic is limiting access to its powerful new Mythos model and GPU costs have surged 48% in recent months, signaling that compute scarcity is now actively constraining AI growth and forcing labs into difficult decisions about deployment and expansion.