Machinepower Index / Dossiers / United States
Rank 1–7 across the scoring rules tested. How this is tested
United States
Total Sovereignty — For the two nations able to attempt every layer alone: own the whole stack, from energy and chips to frontier models and the rules, while depending on each other somewhere.
How the Score, Potential and bottleneck cost are worked out: methodology.
The three layers
Energy, grids, data centres and silicon — the physical capacity to run models at scale.
Compute tracked by Epoch AI: the United States controls 13.14 million H100-equivalents, wherever the sites are; hosts 134,000 for foreign firms; has 873,000 on its soil with the owner not reported. Hosted compute is shown, not scored.
Models, alignment capability, standards influence and the talent that produces them.
The capacity to wire AI into the state and the economy while holding public consent.
Where the score comes from
| Layer | Sub-dimension | Basis | Score |
|---|---|---|---|
| Watts | Planning & Permitting | Assessed | 7.8 |
| Firm Power Capacity | Dataset | 8.8 | |
| Sovereign Compute | Dataset | 10.0 | |
| Hardware & Silicon Access | Dataset | 10.0 | |
| Weights | Frontier Model Output | Dataset | 10.0 |
| Alignment Capability | Assessed | 8.8 | |
| Regulatory Standing | Dataset | 8.2 | |
| Talent Density | Dataset | 9.7 | |
| Will | Policy Capacity | Dataset | 9.3 |
| Public Sector Adoption | Dataset | 9.3 | |
| Workforce Transition | Dataset | 9.3 | |
| Public Trust | Dataset Sources differ | 4.9 |
Assessed marks a measurement we judged rather than took from a dataset: 2 of the 12 here.
Sources differ marks a cell where two sources disagree by more than a point. Public Trust: Ipsos and Melbourne/KPMG differ by more than a point (4.8 against 6.1); the cell is their average.
Sources
Share of adults agreeing that products and services using AI have more benefits than drawbacks; the Index scores public trust as 1 + share ÷ 10. Survey of 23,532 adults across 32 countries, March to April 2026.
View source →The United States builds data centres faster than anywhere else and a July 2025 executive order set out to accelerate federal permitting for them, but grid interconnection queues run to well over a terawatt of projects with multi-year waits.
View source →El Capitan (LLNL) leads US sovereign installations at 1.809 exaflops on the June 2026 TOP500 list, second globally behind China's LineShine.
View source →The US released 59 notable AI models in 2025 (vs China's 35) and attracted $285.9B in private AI investment — roughly 23x China's recorded private total.
View source →The US ranks first in the 2025 Government AI Readiness Index across 69 indicators covering government, technology sector, and data & infrastructure pillars.
View source →Used for the workforce measurement. The IMF combines schooling, pay, labour mobility and active labour-market policies; the Index rescales it so the weakest of 173 economies scores 0 and the strongest 10.
View source →A cross-check for planning and permitting, which stays assessed: it scores building permits and electricity connections for ordinary firms, not grid-scale connections for data centres.
View source →A cross-check for alignment, which stays assessed. The Research Almanac counts English-language research only, so it understates Japanese, Korean and Chinese work.
View source →Fossil, nuclear, hydro and other dispatchable plant; averaged with generation for firm power.
View source →Second source for talent, averaged with Stanford's count.
View source →Second source for public trust, averaged with Ipsos.
View source →Capital is shown beside the Score, not scored.
View source →Half of public trust: support for building more data centres, scored as 1 + share ÷ 10 and averaged with trust in AI.
View source →On the register — what the world cannot get anywhere else
The United States holds four of the register's chokepoints: positions nobody else can route around, which its government could actually shut off. See the whole chain.
Almost no advanced chip is designed without American electronic design automation tools. Withdrawing them stops a design programme before any silicon exists, and Washington has done exactly that.
US Bureau of Industry and Security →Nvidia designs the overwhelming majority of AI accelerators, and Washington has repeatedly restricted which nations may buy them. The design sits in California even when the manufacturing does not.
US Bureau of Industry and Security →The Foreign Direct Product rule lets Washington restrict foreign-made goods if they were built using American technology. Because ASML machines and Japanese tooling both contain American parts and software, this reaches into chokepoints the United States does not own. In 2023 both the Netherlands and Japan brought in their own export controls that lined up with American ones. This is the strongest entry on the list: it is a law rather than an industry, and it sits on top of everyone else's position.
US Bureau of Industry and Security →On 12 June 2026 the Commerce Department ordered Anthropic to obtain a licence before giving any foreign person access to its most capable models, after researchers showed their safety guardrails could be bypassed for offensive cyber work. The company answered that it could not separate foreign users at short notice, and switched the models off worldwide. Two weeks later Washington exempted a set of trusted partners, and a second developer released its next models to a government-vetted group first. On 30 June Commerce withdrew the licensing requirement altogether, and access returned worldwide the next day. This is the newest lever on the page and the least settled: the legal authority is novel, a customer is already challenging it in court, and it reaches named models rather than the frontier as a class. But it has been pulled, and the world’s most capable models went dark when it was.
Mayer Brown — US Commerce extends export controls to AI models →Bottlenecks
These hold the United States below its Potential. Its two lowest scores cost it the most: public trust at 4.9 (part of Will) and planning & permitting at 7.8 (part of Watts).
Assessment
The global leader in intelligence power, the US maintains a dominant position through its control of the primary model weights and semiconductor design. However, bottlenecks in energy grid permitting and domestic state diffusion remain key challenges to total sovereign capacity.
The art of the possible
Fixing grid interconnection queues is the single highest-leverage move available to any nation in the Index: the US already leads on models, chips and capital, and every megawatt it unlocks compounds an existing lead. Faster federal permitting plus public-sector adoption would push the leader further from the pack.
Strategic play — Total Sovereignty
Keep the lead by building the power it runs on: clear the interconnection queues and permitting backlog so new capacity reaches the grid as fast as chips arrive, and use the export lever sparingly, because every turn of it teaches the rest of the world to route around American parts.
Leverage — the nearest position. The strongest chokepoint on earth is a law. American export rules reach any machine built anywhere with American parts or software inside it. The cost of that power is overuse: every time it turns, the rest of the world gets another reason to build around it.
Open United States in the live Index → — every source behind every number, and the comparison tool.