Economist Estimates AI Buildout Would Need $3.55 Trillion in Annual Revenue
aiinsider ·
A working paper by Columbia Business School finance professor Stijn Van Nieuwerburgh estimates that the U.S. AI data center buildout would require approximately $3.55 trillion in mature annual revenue to sustain its investment costs. This figure is based on a scenario where about 188 gigawatts of additional capacity become operational between 2025 and 2032, generating nearly $9 trillion in total investment. The calculation assumes a six-year economic life for computing equipment, a 10% annual return before borrowing, and a 50% operating cash-flow margin. If the useful life of GPUs is shortened to three years due to rapid technological changes, the required annual revenue rises significantly to about $5.7 trillion. The study highlights that financing structures involving leases, joint ventures, and outside borrowing distribute financial exposure but can create complex obligations for technology companies. Using Meta’s Hyperion project as a case study, the paper illustrates how residual-value guarantees can keep tenants exposed to project performance despite external ownership. These estimates represent modeled revenue requirements rather than forecasts of actual sales or the broader economic value created by AI applications.
aiinsider