Leading AI company reaches $40B annual compute costs, 6GW peak power and $200B active compute
Tracker summary: OPENBRAIN'S COMPUTE COSTS $40B 2026 ANNUAL · OPENBRAIN POWER REQUIREMENT 6GW PEAK POWER · CAPITAL EXPENDITURE $200B COST OF OWNERSHIP OF OPENBRAIN'S ACTIVE COMPUTE
Primary disclosures show rapid single-lab compute growth and a large secured pipeline. The $40B annual expense, 6GW peak power and $200B active hardware combination is not yet established.
At a glance
- Assessment: Emerging
- Confidence in assessment: 60%
- Outcome: unresolved
- Timing: pending
- Evidence: proxy
- Predicted timing: End of 2026
- Primary source: AI 2027 archived scenario: Late 2026: AI Takes Some Jobs, p. 8 (Key Metrics 2026 dashboard)
What AI 2027 Predicted
Tracker summary: The leading lab reaches three end-2026 quantities: $40B annual compute expenditure, 6GW peak power and $200B in active computing hardware. These describe one lab’s resources and costs, not the global buildout. Archived scenario dashboard · Compute supplement
Scenario source: archived version and page-number reference: Late 2026: AI Takes Some Jobs, p. 8 (Key Metrics 2026 dashboard).
How We Track This
Labs tested: OpenAI and Anthropic are assessed separately as candidate leading labs. The deciding result is a complete same-lab conjunction: either named lab must establish all three quantities for the relevant 2026 period. Neither lab is preselected as a definitive identification of the fictional OpenBrain, and expense from one cannot be combined with power or hardware from the other. Neither currently supplies a complete qualifying result in the evidence below; the present rating is unchanged.
Assess each quantity separately before evaluating the combination. Contracted future capacity is distinct from operating capacity; available capacity is distinct from measured peak draw. Multi-year commitments are not annual expense. Cloud payments cannot simply be reclassified as a customer’s owned hardware.
Current Evidence
Anthropic’s October 26 S-1 reports unaudited H1 2026 compute/infrastructure expense of $8.945 billion, operating loss of $693 million and operating cash inflow of $3.929 billion. Q2 operating income was $974 million. H1 net loss of $80.748 billion includes $72.388 billion in financing-instrument fair-value changes, not compute expense. The Microsoft minimum spend of $44.9 billion covers November 2026-May 2033, not H1 expense. Cash inflow is not profit; these comparator accounts establish neither annual compute costs nor operating power or active hardware for the tracked single-lab combination. S-1, operations, cash flows and infrastructure agreements
Anthropic’s November 2 amended S-1 reports $18.808B of compute and infrastructure expense for the nine months through September 30, including $9.863B in Q3. These interim actuals extend the H1 account but do not establish $40B for the full year. The category includes infrastructure, and the filing does not supply a matched 6GW peak-power measurement or $200B active-compute valuation. Amended S-1, operations and quarterly results
OpenAI’s CFO reported available compute of approximately 0.2GW in 2023, 0.6GW in 2024 and 1.9GW in 2025. This is substantial realized access growth. Reaching 6GW from 1.9GW would require a further 3.16× increase. Repeating the preceding approximate threefold pace would give 5.7GW, an illustrative scale comparison rather than a forecast or measurement of peak power. OpenAI business account
The secured pipeline is larger. OpenAI’s April update said it had surpassed its objective of securing 10GW of US infrastructure by 2029, including 3GW added to the pipeline in 90 days. Its August Ports Pike announcement describes another large prospective campus. These announcements support continued expansion; they do not establish that the capacity is operating in 2026. April infrastructure update · Ports Pike announcement
| Scenario quantity | Anthropic | OpenAI |
|---|---|---|
| $40B annual compute expense | November 2 amended S-1: $18.808B of compute/infrastructure expense through September 30, extending the $8.945B H1 account but not a full-year result; no full-year qualifying amount established. Multi-year commitments are not annual expense. | The cited primary disclosures describe capacity and infrastructure commitments, not a comparable booked $40B calendar-2026 compute expense. No qualifying annual amount established. |
| 6GW peak power | The S-1 does not provide a reconciled operating peak-draw total establishing this threshold for Anthropic; contracted cloud capacity is not a peak-power measurement. | CFO’s account: 1.9GW of available compute in 2025; the later secured pipeline is not measured operating peak draw in 2026. No qualifying 6GW result established. |
| $200B active hardware | No comparable reconciled inventory or access valuation establishing $200B of active hardware for Anthropic is available in the cited evidence. Cloud-spend commitments are not an owned-hardware balance. | No comparable reconciled single-lab active-hardware valuation establishing $200B is available in the cited evidence. Infrastructure project totals cannot be assigned wholly to OpenAI. |
Counterevidence & Limitations
The disclosed baseline predates the deadline by a year, while much of the pipeline extends beyond it. Capacity labels can refer to facility power, IT load or contracted access. Hardware serving multiple customers cannot be assigned in full to each lab.
Emerging reflects positive scale evidence with an unresolved conjunction. It does not mean that announced capacity has no value. It means the conversion from contracts to the three operational quantities is not yet sufficiently measured.
What Would Change Our Assessment
- Strengthen: Dated commissioning and usage disclosures bridge the existing baseline to 2026 operating resources.
- Confirm: Comparable single-lab accounts establish all three quantities within the window.
- Weaken: Disclosed delays or reconciled deadline-period accounts demonstrate a material shortfall.
Update History
| Date | Update |
|---|---|
| 2026-11-07 | Assessment reviewed. Status: Emerging; confidence: 60%. The amended Anthropic accounts report $18.808B in compute and infrastructure expense through September, supporting substantial realized scale without establishing $40B for the full year. OpenAI’s 1.9GW available-compute baseline is a separate company and period, not measured 2026 peak draw. Neither source establishes all three quantities for one lab; expense scope, operating power and active-hardware valuation remain material limitations. The $46B projection and training/inference breakdown are excluded as unverified supporting evidence, as are the unsupported ancillary growth and combined-capex figures. Commitments, projections and annualized revenue are not realized annual results. Anthropic and OpenAI are evaluated separately; one lab must meet the expense, peak-power and active-hardware conditions together. Added the amended nine-month expense and Q3 actuals without treating them as full-year expense or combining labs. |
| 2026-11-02 | Added Anthropic comparator accounts separating compute expense, operating results, cash flow, fair-value losses and future Microsoft commitments. The single-lab resource combination remains unresolved; status and confidence unchanged. |
| 2026-09-07 | Added the primary 1.9GW available-compute baseline and evaluated the growth required to approach 6GW. Secured future capacity supports expansion but cannot establish annual expense, peak draw or active-hardware value. 2026-11-07 · Correction: status Emerging (55%) → Emerging (60%). The available-compute disclosure does not establish the three end-2026 operational quantities. |
| 2026-09-06 | Assessment revised from on-track (0.55) to emerging (0.55). Annual compute expense, operational peak power and active-compute value are separate unresolved quantities for the same company. |
| 2026-08-31 | Reuters reported that Anthropic agreed to a six-year, roughly $45 billion Nscale compute rental covering about 460 MW of planned capacity. This adds direct lab-linked compute-cost and power evidence, while leaving 2026 operation and the much larger tracked thresholds unverified. Status and confidence remain unchanged. |
| 2026-08-24 | NVIDIA named OpenAI as the customer for a planned 8 IT-GW PORTS-Pike campus under a 20-year lease, with an initial 4.25 IT-GW phase. This adds direct future capacity evidence, while leaving 2026 delivery, utilization, and actual OpenAI peak load unverified. Status and confidence remain unchanged. |
| 2026-06-15 | Axios reported a $35B Anthropic-linked compute-financing platform involving Apollo, Blackstone, Broadcom, and Fluidstack, designed to support more than 20 GW of compute capacity through 2028. This strengthens evidence for large compute-financing structures, while not proving Anthropic has reached the end-2026 $40B compute-cost, 6 GW power, or $200B active-compute thresholds. Confidence adjusted 0.50 → 0.55. |
| 2026-03 | OpenAI closes $122B funding at $852B valuation. Stargate program now encompasses ~10GW planned capacity. Assessment: $40B compute costs likely met (OpenAI projects $46B). 6GW in planned capacity exceeded but operational capacity harder to verify. $200B single-company capex approached by Amazon but serves entire AWS business. Status: Emerging → On Track for compute costs; power and capex remain Emerging. Confidence 0.40 → 0.50. |
| 2026-02 | Alphabet guides 2026 capex at $175-185B (nearly double $91.4B in 2025). Amazon guides $200B for 2026 (+53%). Big 5 combined: $660-690B. OpenAI revises compute spending to ~$600B by 2030 (down from $1.4T), projects $46B total compute costs for 2026 ($32B training + $14.1B inference) — exceeding the $40B threshold. Anthropic closes $30B Series G at $380B valuation. 2026-11-07 · Source clarification: Alphabet’s Q4 2025 earnings call supports $175-185B 2026 guidance and $91.4B 2025 capex; Amazon, February 5, 2026 supports about $200B 2026 capex; Anthropic, February 12 supports the $30B raise and $380B post-money valuation. Reuters, February 20, 2026 reports a roughly $600B compute-spending target through 2030 from an unnamed source, not realized annual spending. The earlier $1.4T commitment covered a different horizon; the subtraction is not like-for-like. The $46B total/$32B training/$14.1B inference breakdown, Amazon +53% comparison and $660-690B combined estimate were not verified against the original reporting and are withdrawn as supporting evidence. Historical text is retained, not treated as measured spending. |
| 2026-01 | Meta guides 2026 capex at $115-135B (nearly double 2025). Stock surges 10.4%. |
| 2025-12 | Analysts project hyperscaler capex exceeding $600B in 2026 (+36% over 2025). SoftBank completes $22.5B second tranche to OpenAI. Michigan regulators approve 1.4GW for Stargate. Full-year 2025 capex: Amazon ~$131.8B, Alphabet $91.4B, Meta $72.2B. |
| 2025-11 | Sam Altman states OpenAI has ~$20B ARR and ~$1.4T in datacenter commitments over 8 years (~30GW capacity). Peak ambition before later revision. 2026-11-07 · Source clarification: CNBC, November 6, 2025 reports a forecast of more than $20B annualized revenue at year-end, not $20B realized annual revenue. Reuters, February 20, 2026 retrospectively supports the $1.4T/30GW commitment account. The eight-year horizon is reported by Reuters, November 7, 2025; that exact source was available only in the indexed excerpt during this check. These are plans and commitments, not spending already incurred. |
| 2025-10 | AMD and OpenAI announce 6GW GPU partnership (~$90B cumulative). OpenAI completes for-profit restructuring. Q3 capex: Alphabet $24B, Amazon $34.2B, Meta $19.4B, Microsoft $34.9B (+74% YoY). |
| 2025-09 | Anthropic closes $13B Series F at $183B valuation. First Stargate data center operational in Abilene, TX (200MW+). CoreWeave expands OpenAI deal by another $6.5B. Five new Stargate sites announced with combined ~7GW planned capacity. |
| 2025-07 | OpenAI and Oracle announce 4.5GW Stargate expansion exceeding $300B over 5 years. Alphabet raises 2025 capex guidance to $85B. Amazon reports Q2 capex of $31.4B, implying full-year ~$125B+. |
| 2025-05 | Q1 2025 earnings: Alphabet $17.2B, Meta ~$19.4B, Amazon $24.3B, Microsoft $21.4B quarterly capex. All confirm AI infrastructure acceleration. CoreWeave expands OpenAI agreement by $4B (cumulative ~$16B). |
| 2025-03 | OpenAI signs $11.9B five-year deal with CoreWeave. OpenAI closes $40B funding round at $300B valuation — largest private tech fundraise in history. |
| 2025-01 | Microsoft announces $80B capex for FY2025. Stargate Project announced at the White House — OpenAI, SoftBank, Oracle target $500B over 4 years with $100B immediate deployment. DeepSeek R1 released at claimed $5.6M training cost; Nvidia loses $589B in market cap. |