AI’s infrastructure cost is now measured in trillions. The harder question is what revenue that spending implies, because the math is backwards, with revenue uncertain, but capex increasingly locked in.
Various groups have estimates, like researchers at Columbia and analysts at Bain, but those are static. We wanted something that allowed us to play "what if" with existing forecasts, or create our own.
We have built a new interactive tool for premium subscribers, How Much Revenue Must AI Deliver?, to make that arithmetic less opaque. Start with the capital outlay, choose a return hurdle and cash margin, and see the implied annual revenue the investment needs to support it. The chart shows how that answer changes as spending rises.
You can start from published Columbia or Bain estimates, or choose Custom and enter your own aggregate or annual capex. Then change the assumptions. A larger buildout, a higher required return, slower depreciation, or a thinner margin can dramatically move the revenue target. And it is all put in context against other large markets, like global software, or, you know, U.S. GDP.
Video walkthrough
You can try the tool here. It is available to all premium subscribers.