AI Revenue Finally Beats Infrastructure Costs—But Just Barely

Google, Amazon, and the rest have hit a milestone that sounds less heroic than you'd expect: AI revenue has started to outpace the depreciation costs of their infrastructure investments. On paper it looks great, but really it's like crossing the break-even point on a balance sheet—technically fine, but with zero room for mistakes.

The real problem is that this balance depends entirely on two fragile things: keeping prices high and never stopping growth. Meta, Google, and Microsoft bet the farm on this equation, pouring tens of billions into infrastructure betting that AI demand would never slow down. So far so good, but margins are so thin that any dip in user growth or a price war between players would be ugly.

Immediate consequence: these companies can't afford to slash AI service prices anytime soon, even if competitors start circling. Depreciation costs will keep climbing as they build new data centers to keep up with demand. So those AI subscriptions and APIs you use today? They're basically locked in at current prices for the next 18-24 months. After that, it's whoever figures out how to actually optimize costs instead of defaulting to "more GPUs = more AI" wins.