Paul Kedrosky has released an interactive model for estimating the yearly revenue required to support different AI infrastructure budgets, giving paid subscribers a way to test how financial assumptions change the target.
In a post on his website dated Oct. 23, Kedrosky said users can enter capital spending, set a return hurdle and choose a cash margin. The model calculates the implied annual revenue and displays it alongside measures including software revenue, U.S. gross domestic product and white-collar labor.
Subscribers can use estimates associated with researchers at Columbia or analysts at Bain as a starting point. The model also accepts user-entered spending estimates, either annual or cumulative.
Users can vary the scale of the buildout, required return, depreciation pace and cash margin. Kedrosky wrote that changing these assumptions can substantially shift the revenue target.
The post frames the exercise around a gap in visibility: AI infrastructure costs are measured in trillions, while the revenue those investments may support remains uncertain. Kedrosky said estimates from Columbia and Bain are static, and described the model as a way to test different scenarios. It is available to paid subscribers.
