Goldman Sachs Research (Jim Schneider, senior equity analyst) models agentic AI driving token consumption from current levels to 120 quadrillion tokens per month by 2030, a 24-fold increase, with enterprise adoption reaching 12% of knowledge workers by 2030 and 37% by 2040. Semiconductor providers are delivering inference cost reductions of 60-70% per year, which GS expects to produce a gross margin inflection for hyperscalers within 3-12 months as cost declines outpace price declines, easing FCF compression from elevated capex. A chip shortage is forecast to persist for 12 or more months before new fabrication capacity catches up. For financial institutions, the central structural claim is that the cost-demand crossover, not just demand growth alone, determines whether hyperscaler FCF recovers, making the pace of inference cost deflation relative to capex commitment the key variable to track.
487 other tracked developments touching the same claims.