The AI Compute Crunch Is Here (and It's Affecting the Economy)

Summary

The era of cheap, subsidized artificial intelligence is coming to an end as companies struggle with the mounting costs of running AI models at scale. Leading AI companies including Anthropic, OpenAI, GitHub, Meta, Google, Microsoft, and Salesforce are responding to severe compute constraints by restricting access to their products, removing expensive features, pausing new signups, and increasing prices. Anthropic has tightened access to Claude Code, OpenAI's leadership has publicly stated the company lacks sufficient compute and decided to shut down Sora, and GitHub has paused new signups for Copilot while removing access to more expensive models.

The compute crunch extends far beyond product restrictions and price increases—it's reverberating through the broader economy. Consumer electronics are experiencing dramatic price increases as chip manufacturers redirect capacity to AI data centers, with Apple reportedly struggling to secure chipmaking capacity for upcoming iPhones. Storage devices have seen particularly steep price hikes, with one analyst noting a 2TB external SSD jumping from $159 to $575 over the course of a year. The infrastructure demands of AI are also straining electricity grids in regions with high concentrations of AI data centers, leading towns and states to actively reject new data center proposals due to concerns about electrical capacity and water usage.

Unlike the early promise of AI as accessible technology, the current business model of AI companies—subsidizing products with venture capital to lock in users—is proving fundamentally unsustainable. Meta has laid off 10% of its workforce partly to redirect savings toward AI infrastructure investment, and other companies are grappling with how long investors will continue burning cash to maintain artificially low prices. The article draws parallels to Uber's years of investor-funded subsidies, which eventually gave way to price increases and declining service quality once market dominance was secure.

• Infrastructure constraints from data centers are straining electricity grids and water supplies, prompting regions to restrict new AI infrastructure projects
• The age of cheap, widely accessible AI is ending, with pricing and feature restrictions becoming industry-wide across Anthropic, OpenAI, GitHub, Meta, Google, and Microsoft

Editorial Opinion

This analysis presents a sobering but logical inflection point in the AI industry's growth trajectory. The computational and financial economics of Large Language Models were always going to reach a breaking point where subsidized pricing could no longer be sustained, and the evidence now suggests we're at that moment. The coming correction—rising prices, reduced access, and harder trade-offs about compute allocation—will likely reshape the AI market significantly, potentially limiting adoption among smaller players while consolidating power among companies with the deepest pockets and most efficient infrastructure.