A fixed investment outlook for adoption, efficiency and the physical capacity required to serve useful work.
Cheaper intelligence expands use. The bottleneck moves to delivery and monetization.
Illustrative AI inference workload, 2026 = 100. These are fixed research assumptions, not measurements of the whole AI economy.
Every path is a fixed analytical scenario prepared for this edition. These are indexed inference-workload assumptions, not observed industry totals, company guidance or price targets.
Required compute index = useful task volume index × compute per task index ÷ 100.
Service revenue index = required compute index × compute-price index ÷ 100. Task volume already describes consumed work; utilization is not multiplied into revenue a second time.
Installed capacity index = required compute index ÷ (utilization ÷ 70). Lower utilization requires more installed capacity for the same consumed workload.
No EPS or stock return is mechanically inferred from these indices. Hardware mix, depreciation, operating cost, reinvestment and entry valuation remain separate underwriting questions.
Distribution, proprietary context and verification can retain value as model access becomes cheaper.
Paid conversion, repeat usage and accepted-task cost; usage alone is not revenue.
Usage expands while customer retention, willingness to pay or contribution profit deteriorates.
Our base case requires more compute despite lower compute per task. Internal chips can change who captures the spending.
Deployed throughput, custom-chip adoption, gross margin and customer concentration.
Efficiency exceeds task growth, or capacity utilization and service pricing fall together.
The original memory thesis focuses on HBM content and the subsequent equipment cycle; stronger demand still has to become qualified shipments and cash.
HBM4 qualification, bit supply, realized pricing and actual WFE orders.
Competing supply weakens pricing or customer equipment budgets turn down.
Original research ↗Electrical and thermal capacity must arrive on time. Announced megawatts are not energized megawatts.
Order conversion, commissioning dates, project margins and free cash flow.
Backlog cancellations, delivery delays or lower cash margins overwhelm order growth.
Original research ↗2025 to 2030
IEA 2026 outlook · global data centres, not AI-only
Quarter ended 26 Jul 2026
Company reported · fiscal Q2 2027
Original memory and equipment thesis ↗ · Original U.S. power research ↗ · IEA · Key Questions on Energy and AI ↗ · NVIDIA fiscal Q2 2027 results ↗ · Frontier efficiency and scaling research ↗