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

Earlier this week, I wrote an article about startups that are spending money on AI compute (tokens on tools like Claude and OpenAI’s products) rather than hiring human employees. There are all sorts of ways this business strategy could fail, and we are beginning to see signs that one of the most obvious ones could be coming to pass: AI companies can’t endlessly subsidize their AI products by charging users less than it costs to actually run them.This is the AI compute crunch, and the signs are all around us: GitHub announced it is pausing new signups for Copilot, tightening usage limits, and removing access to several more expensive AI models. Anthropic has tightened access to Claude Code, and tested removing access to Claude Code entirely in its $20 per month plan (keeping access in its $100 per month plan)As noted in The Verge, Anthropic restricted Claude access to users of OpenClaw because the heavy usage was unsustainableOpenAI’s CFO Sarah Friar has been talking endlessly about how the company does not have enough compute, which has manifested in decisions like deciding to shut down SoraSoftware that has AI tools embedded in them have increased between 20 and 37 percent according to some analysts; this has included increases in prices for Microsoft 365, Notion’s Business plan, Salesforce, and Google Workspace pricesThere is a general rationing of AI products and servicesMeta is laying off 10 percent of its workforce in part because it sounds like the company wants to spend some of the savings on AI infrastructure: The layoffs are “to allow us to offset the other investments we’re making,” the company told its remaining employees. Its main recent investments have been data centers and the tech to run data centers.But it’s not just that AI companies are restricting access to their products, shutting down products altogether, and beginning to increase prices. The broader impact of the current unsustainability of AI can be seen across various sectors of the economy. RAM, graphics cards, and hard drive / solid state storage for consumers have skyrocketed in price and are sold out in many stores. The same 2TB external SSD I bought late last year cost me $159 at the time, cost $449 a month ago, and costs $575 today.Similarly, the general cost of consumer electronics is increasing as chip manufacturers and production lines shift their focus to building more AI capacity. The largest consumer electronics manufacturer in the world, Apple, says it is having trouble securing chipmaking capacity for upcoming iPhones. Home electric bill costs have skyrocketed in some states with high concentrations of AI data centers, leading in part to a widespread, concerted effort by some towns and states to reject and restrict new data centers entirely. There is a fear among experts that similar shortages and price increases could come for water supplies as well.What this means is that the age of cheap, underpriced AI appears to be ending, or at least the compute crunch means the venture capitalists and investment firms funding OpenAI and Anthropic are going to have to be willing to burn even more cash in order to continue subsidizing their products. On the podcast this week, I compared this situation to Uber (and any number of fast-scaling startups that sought to lock in customers then jack up prices). This comparison is only useful in that, like Uber, what AI companies are doing to this point is wildly unsustainable and is being subsidized by investors. For…

This excerpt is published under fair use for community discussion. Read the full article at 404 Media.