Goldman Sachs Forecasts Agents Boost Tech Cash Flow

Reporting by PYMNTS summarizes a Goldman Sachs research note forecasting that the adoption of autonomous, agentic AI will drive a large increase in usage and revenue-related metrics for major technology firms. PYMNTS reports Goldman Sachs projects a 24-fold increase in global token consumption by 2030, reaching 120 quadrillion tokens processed per month. PYMNTS quotes Jim Schneider, a senior equity analyst covering U.S. semiconductor and IT services, saying lower compute costs and higher gross margins could produce a period of "margin inflection."

Reporting by PYMNTS notes the forecast couples rising token volumes with declining per-token compute costs, producing improved revenue-to-cost dynamics for cloud providers and AI platforms. PYMNTS reports Goldman Sachs' adoption forecasts include 12% of knowledge workers using agentic AI by 2030 and 37% by 2040, figures presented in the research note.

Editorial analysis: Industry observers following infrastructure economics will view the Goldman Sachs numbers as quantifying a large demand shock for compute and data throughput, which affects capacity planning, pricing, and procurement for hyperscalers and cloud customers. Editorial analysis: The research highlights a persistent hardware constraint, with PYMNTS reporting Schneider expects a shortage of high-end semiconductors to last 12 to 18 months as foundries scale production, which aligns with public supply-chain signals seen across the chip sector.

Reporting by PYMNTS identifies several near-term indicators that would test the research assumptions: measured growth in token-processing volumes from major cloud providers, visible per-token cost declines in published pricing or cost-of-goods disclosures, and semiconductor capacity announcements from major foundries. Editorial analysis: Practitioners and procurement teams should monitor announced wafer capacity, GPU/accelerator allocations, and public telemetry about inference volumes to gauge whether the forecasted "margin inflection" materializes for hyperscalers and AI platforms.