Goldman Sachs Global Institute published a scenario-based framework estimating $765B in annual AI capex in 2026 rising to $1.6T by 2031, totaling approximately $7.6T cumulatively, anchored to NVIDIA forward data center revenue estimates. The analysis argues the headline figure is highly sensitive to four supply-side variables: AI silicon useful life (where a shift from 4-year to 6-year depreciation moves cumulative spend by hundreds of billions), next-generation data center cost and complexity, chip architecture mix, and physical bottleneck elongation. Demand-side adoption questions, which dominate public debate, are framed as secondary to these supply-side structural assumptions in determining total capital deployed. For financial institutions, this framework provides a structured basis for stress-testing AI infrastructure equity and credit exposures against plausible assumption shifts rather than a single consensus number.
310 other tracked developments touching the same claims.