The biggest bet in history

Per LessWrong, five hyperscaler companies—Amazon, Google, Meta, Microsoft, and Oracle—have collectively committed two point five trillion dollars to AI infrastructure, with roughly half already spent. That's the largest capital expenditure in human history. But the financial math is the revealing part: the hardware depreciates by about a quarter every year, meaning to satisfy shareholder returns and cover that depreciation, these companies need to generate roughly six hundred seventy billion dollars in annual profit from their AI investments. LessWrong runs the numbers against the eight point five trillion in yearly US labor costs that current AI can address, and finds that capturing just eight percent of the productivity gains those systems unlock would make the bet pay off. That implies a twenty-four percent economy-wide productivity boost within four years. The article argues that this math suggests hyperscalers' finance teams have been convinced AGI is coming soon enough to justify the commitment. It closes by noting there's no comfortable middle ground: either AGI arrives and the capability gap holds, governments freeze open-weight competition in place, or competition collapses margins entirely—much like the fiber-optic bubble of the nineteen-nineties, where investors were right about the internet but still lost money.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton