Power ranges: AI faces supply crunch

The AI industry is facing a serious supply crunch. While demand for artificial intelligence is surging exponentially—OpenRouter, a marketplace for AI models, reported that the number of tokens processed each week quadrupled between January and March alone—the physical infrastructure needed to run these models is not keeping pace. AI companies are now actively throttling access to their tools, and the root cause is a bottleneck in the hardware supply chain that cannot be solved quickly, even with hundreds of billions of dollars.

The core problem: processing power can't keep up

Every time a user interacts with an AI tool, the model performs "inference"—it processes a query and generates a response. This requires powerful processors called GPUs (graphics processing units), which are housed in massive data centers. As more people use AI, more computing resources are needed. The industry is simply running out of processing power to meet current demand. This has already forced companies to change their behavior. Anthropic, a major AI model maker, recently altered its service terms to discourage users from accessing capacity at peak times. OpenAI shut down Sora, its video generation tool, to reallocate scarce computing resources toward more lucrative ventures. These are not signs of a temporary hiccup; they are evidence of a structural mismatch between software demand and hardwa