This model is not a capex forecast. It is a demand model expressed in MW-equivalent. The dollar output is the implied capex if 100% of demand were satisfied. Compared to top-down consensus, the difference is the SUPPLY GAP — the size of the physical execution constraint (power, permits, chips, labor) that the AI economy is now hitting. Hyperscalers themselves describe markets as supply-constrained, not demand-constrained (Moody's Apr 2026, Futurum Feb 2026).
Five knobs drive the demand build. All headline numbers in this page reflect the knob settings shown below. Default settings are bolded.
Bottom-up token demand and the demand-implied capex it would require at current chip-mix throughput, workload realization, and $/MW. Cumulative capex through 2030 lands at $7.8T.
| Year | Demand (T tokens) | BU capex ($B) | Cumulative ($B) | YoY Δ |
|---|---|---|---|---|
| 2026 | 271,799 | $1,103 | $1,103 | — |
| 2027 | 541,862 | $1,254 | $2,357 | +14% |
| 2028 | 1,036,972 | $1,411 | $3,768 | +12% |
| 2029 | 1,986,408 | $1,409 | $5,177 | -0% |
| 2030 | 3,986,990 | $2,649 | $7,826 | +88% |
Bottom-up demand-implied capex compared to the mean of eight authoritative top-down forecasts (Goldman, Dell'Oro, BCG, Bain, Moody's, Futurum, McKinsey, Deloitte). A positive gap means BU demand exceeds what consensus expects will actually be built.
| Year | BU capex ($B) | TD consensus mean ($B) | Supply gap ($B) | Gap as % of supply |
|---|---|---|---|---|
| 2026 | $1,103 | $573 | $530 | 92% |
| 2027 | $1,254 | $692 | $562 | 81% |
| 2028 | $1,411 | $837 | $574 | 69% |
| 2029 | $1,409 | $1,005 | $404 | 40% |
| 2030 | $2,649 | $1,208 | $1,442 | 119% |
The gap widens dramatically in 2030 (119% of planned supply) as the BU demand-implied capex steps up while TD consensus follows a smoother trajectory.
The agentic chain is the model's load-bearing demand source. It decomposes across 11 occupational buckets, each with its own workforce size, penetration curve, and per-worker annual token volume. Sorted by 2030 agentic contribution.
| Function | Workforce 2030 (M) | Penetration 2030 | Tokens / worker / yr | 2030 demand (T) |
|---|---|---|---|---|
| F1 Software engineering | 26.9 | 72% | 44.8B | 867,364 |
| F5 Healthcare | 35.9 | 58% | 26.4B | 548,935 |
| F4 Knowledge ops | 43.6 | 42% | 18.6B | 340,759 |
| F8 Arts & media | 7.5 | 78% | 52.5B | 308,763 |
| F11 Pharma R&D | 1.0 | 82% | 284.4B | 237,851 |
| F9 Sciences | 5.7 | 60% | 45.0B | 152,550 |
| F3 Sales & marketing | 16.9 | 55% | 16.2B | 150,597 |
| F7 Engineering (non-SWE) | 12.1 | 50% | 21.0B | 126,591 |
| F6 Education | 28.4 | 40% | 4.2B | 47,536 |
| F2 Customer support | 9.9 | 62% | 7.5B | 46,221 |
| F10 Government | 11.1 | 30% | 4.3B | 14,424 |
| Total agentic 2030 | 2,841,592 |
Tokens-per-worker varies ~65× across functions, from F10 Government (~4.3B/yr) to F11 Pharma R&D (~284B/yr — reflecting literature-search and trial-design workloads that scale very differently from chat). SWE, Healthcare, Knowledge Ops, and Arts/Media together contribute over 70% of 2030 agentic demand.
Demand: 4 workload chains — consumer chat (user-based), embedded chat (seat-based), agentic (labor-based across 11 function buckets via a 6-factor formula), ambient (top-down) — summed and × induced-demand multiplier (Knob 3).
CapEx: token demand → MW_Bridge (chip-mix throughput × workload realization × derating) → required inference MW → refresh treadmill on installed base × $/MW = inference capex. Training capex layered via Deloitte inference-share trajectory × same $/MW. Total = inference + training.
Knob 5 (workload realization) drives the chip-throughput derate factor: Bear 75% (chat-like workloads), Base 59% (research-derived 60% reasoning + 25% summarization + 15% chat mix), Bull 45% (heavy agentic). Top-down consensus inputs sourced from Z_Sources S190-S200.
The one remaining check failure (Pharma R&D × High intensity stress-corner exceeds the 5T/yr per-worker ceiling at 5.69T) is a documented calibration finding, not a model bug — the check's rationale anticipates the breach. All other 18 integrity rules pass at default knobs.