Twenty-three of the forty campuses in the ledger are in the United States. Eighteen of them carry an announced or estimated capacity, 29,682 MW in all, and five announce no figure. The record evidences 4,820 MW across the twenty-three, at the end-of-August 2026 reading.
11 not yet due · 5 never dated · 4 slipped · 2 kept · 1 missed, if ‘early 2026’ meant the first quarter (the book counts it as a miss)
From The AI Compute Race. Read the book
The campuses
| Campus | Where | Announced or estimated capacity (IT load) | Evidenced power (IT load) | The book’s verdict |
|---|---|---|---|---|
| PORTS-Pike | Pike County, Ohio, US | 8,000 MW | 0 MW | Not yet due |
| Hyperion | Richland Parish, Louisiana, US | 5,000 MW | 0 MW | Not yet due |
| Fairwater Wisconsin | Mount Pleasant, Wisconsin, US | 2,263 MW | 369 MW | Missed, if ‘early 2026’ meant the first quarter (the book counts it as a miss) |
| Rainier / New Carlisle | New Carlisle, Indiana, US | 1,925 MW | 910 MW | Never dated |
| Stargate Doña Ana / Project Jupiter | Doña Ana County, New Mexico, US | 1,750 MW | 0 MW | Not yet due |
| Colossus 2 | Memphis, Tennessee, and Southaven, Mississippi, US | 1,531 MW | 946 MW | Slipped |
| Stargate Shackelford / Frontier | Shackelford County, Texas, US | 1,400 MW | 0 MW | Not yet due |
| Paducah AI/HPC campus | Paducah DOE reservation, Kentucky, US | 1,200 MW | 0 MW | Not yet due |
| Stargate Michigan / The Barn | Saline Township, Michigan, US | 988 MW | 0 MW | Never dated |
| Stargate Port Washington / Lighthouse | Port Washington, Wisconsin, US | 902 MW | 0 MW | Not yet due |
| Stargate Milam County | Milam County, Texas, US | 857 MW | 0 MW | Not yet due |
| Prometheus | New Albany, Ohio, US | 854 MW | 562 MW | Kept |
| Stargate Abilene | Abilene, Texas, US | 843 MW | 421 MW | Slipped |
| Fairwater Atlanta | Fayetteville, Georgia, US | 636 MW | 636 MW | Never dated |
| Helios | Dickens County, Texas, US | 526 MW | 133 MW | Slipped |
| Google New Albany | New Albany, Ohio, US | 453 MW | 453 MW | Never dated |
| Colossus 1 | Memphis, Tennessee, US | 340 MW | 340 MW | Kept |
| Stargate Lordstown | Lordstown, Ohio, US | 214 MW | 0 MW | Not yet due |
| Project Matador / HyperGrid | Amarillo (Carson County), Texas, US | not on file | 0 MW | Not yet due |
| Project Camellia | Effingham County, Georgia, US | not on file | 0 MW | Not yet due |
| AWS Northern Indiana program | Northern Indiana, US | not on file | 0 MW | Not yet due |
| IREN Sweetwater | Sweetwater County, Texas, US (as the August 2026 compilation gives it) | not on file | 0 MW | Never dated |
| Nebius Vineland | Vineland, New Jersey, US | not on file | 50 MW | Slipped |
Who is involved
- OpenAI 9 of 23
- Oracle 5 of 23
- Microsoft 3 of 23
- Amazon AWS 2 of 23
- Anthropic 2 of 23
- Blue Owl 2 of 23
- Meta 2 of 23
- SB Energy 2 of 23
- SoftBank 2 of 23
- Vantage Data Centers 2 of 23
- xAI (SpaceXAI) 2 of 23
- Blue Owl–Meta joint venture 1 of 23
- Brookfield 1 of 23
- CoreWeave 1 of 23
- Crusoe 1 of 23
- Crusoe asset SPV 1 of 23
- DataOne 1 of 23
- Fermi America 1 of 23
- Galaxy Digital 1 of 23
- Google 1 of 23
- IREN 1 of 23
- Nebius 1 of 23
- NextEra 1 of 23
- QTS (Blackstone) 1 of 23
- Related Digital 1 of 23
- Stack Infrastructure 1 of 23
- TensorWave 1 of 23
The dated tests that name these campuses
- · Chapter 7, the frontier’s deployment. An audited frontier model trained end to end on domestic Chinese silicon, at a compute budget within a factor of two of the American frontier, weakens the clause that denial slows the frontier’s deployment in China and confirms the clause that durable denial accelerates the ecosystem beneath it. The audit counts only if a party outside the developer publishes it.
- · Chapters 1 and 12, the first route. Supported if one laboratory holds an independently measured frontier lead through three or more complete improvement cycles, which at a three-month training run is nine months or more without a break. Undermined if American leadership has gone on rotating between laboratories since 1 January 2027 at a median laboratory tenure under thirteen weeks, less than one such cycle and near the present rate of about ten weeks. Both limbs count laboratories: a laboratory’s consecutive leading models make one tenure. A record that meets neither limb, a median of thirteen weeks or more with no lead held for nine months, scores neither. If Epoch has published no count of lead periods for that window by that day, the limb is unscored.
- · Chapters 1 and 12, the second route. Supported if the leading American-controlled system, or the allied stack behind it, sustains a lead of more than fourteen months over the strongest Chinese system with the gap widening, both at once. Undermined if the lag is fourteen months or less. The lag is a proxy for the machine, not a measure of it. It stands in because the lead the wager needs is a lead in what the models can do, and a superior machine serves the wager only if it yields one.
- · Chapters 1 and 12, the wager. Falsified if both of two things hold. Rivals are still reproducing each advance within fourteen months, the top of the historical four-to-fourteen-month envelope, so that a faster catch-up counts too. The public record still shows the leader’s generation time, the time one generation of models takes to produce the next, at months rather than weeks, which chapters 1 and 12 read as more than eight weeks between successive frontier releases. The lag alone undermines the second route, and the generation time alone falsifies nothing.
- · Chapter 2, compute moved rather than stopped. Falsified if the frontier of Epoch’s capability index keeps its slope in calendar 2027 and 2028, advancing at least 14 points in each year, the pace Epoch reported on 1 September 2026, while the growth of compute slows sharply everywhere at once in both years. That means Epoch’s frontier training-compute series growing less than twofold in each year, no offsetting rise in any post-training or inference estimate Epoch publishes for those years, and accelerator demand flat, NVIDIA’s reported Data Center revenue no higher in either of its fiscal years 2028 and 2029 than in the year before it. Every one of those conditions must hold together, and any one alone, frontier training compute slowing on its own included, leaves the claim standing. If Epoch has not published its index or its training-compute series for those years by the reading date, the test is unscored.
- · Chapter 14, the campus’s financing. The verdict that the campus’s financing is secure stands while the monitors show the plants on schedule and S&P’s reports show the project’s coverage above 1.10 times. It is weakened, toward a campus short of power, if either Franklin Farms unit slips past 1 June 2029.
- · Chapter 14, the hour, on a proxy. No filing reports Hyperion’s utilization or what an hour is worth to Meta, so the chapter’s question, and its reading that the campus can run as a strategic engine, are read on a published rent against the break-points of chapter 14’s model of Hyperion’s four futures, on SemiAnalysis’s $3.1 million price for the rack-scale server. Above $3.01 an hour the reading stands, with the strategic engine first on this model. Between $2.66 and $3.01 the strategic engine yields first place to the rationed utility, and the reading is weakened. Below $2.66 a strategic hour no longer covers its rent on this model, and the reading is withdrawn. A market rent is not Meta’s own value, which is why this limb is a proxy, and if neither series is published on the day, the limb is unscored.
- · Chapter 14, the campus’s financing. The verdict is also weakened, toward a campus short of power, if S&P reports the project’s downside-case coverage below 1.10 times on a sustained basis, in two consecutive reports on Hyperion’s notes.
Every figure is the campus page’s own, read at the end of August 2026; each campus page carries its sources. Places · All 40 campuses.