Machinepower Index / Dossiers / United Kingdom
Rank 4–8 across the scoring rules tested. How this is tested
United Kingdom
Specialist Power — For nations among the world's best at something the frontier needs but could find elsewhere: keep earning the lead before others catch up, and trade expertise for access.
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 Kingdom controls 64,000 H100-equivalents, wherever the sites are; hosts 126,000 for foreign firms. 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 | 5.2 |
| Firm Power Capacity | Dataset | 5.4 | |
| Sovereign Compute | Dataset | 6.7 | |
| Hardware & Silicon Access | Dataset | 5.8 | |
| Weights | Frontier Model Output | Assessed | 5.2 |
| Alignment Capability | Assessed | 9.2 | |
| Regulatory Standing | Dataset | 9.2 | |
| Talent Density | Dataset | 7.2 | |
| Will | Policy Capacity | Dataset | 10.0 |
| Public Sector Adoption | Dataset | 6.9 | |
| Workforce Transition | Dataset | 8.6 | |
| Public Trust | Dataset Sources differ | 5.3 |
Assessed marks a measurement we judged rather than took from a dataset: 3 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 →Britain's grid connection queue grew so long that it was overhauled in 2025, and the government has designated data centres critical national infrastructure and created AI Growth Zones to speed planning and power; the queue, not demand, remains the constraint.
View source →Google DeepMind's London headquarters leads much of the research behind Gemini, putting frontier work on British soil but under US ownership; British-owned labs such as Stability AI build credible open models below the frontier.
View source →Isambard-AI Phase 2 (Bristol) is the UK's most powerful public AI system, debuting at #11 globally in June 2025 and anchoring the AI Research Resource.
View source →The July 2025 Compute Roadmap commits over £1B to expand the AI Research Resource twenty-fold by 2030, with five AI Growth Zones designated by January 2026.
View source →The UK ranks third globally in the 2025 index, behind the United States and France, reflecting the AI Security Institute (renamed from AI Safety Institute in February 2025) and strong research capacity.
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 →Bottlenecks
These hold the United Kingdom below its Potential. Its two lowest scores cost it the most: planning & permitting at 5.2 (part of Watts) and frontier model output at 5.2 (part of Weights).
Assessment
The UK punches far above its weight in intelligence power — elite research institutions, DeepMind roots, and the AI Security Institute — but severe energy and compute bottlenecks hold its Score well below its Potential. The question is not whether Britain can matter in AI; it is whether it will build the physical foundations its talent deserves.
The art of the possible
Britain's mistake would be trying to rebuild everything. The frontier gap is real — one notable British model in 2025 — but the way back is through what Britain already leads rather than a standing start on every layer. Arm designs much of the world's silicon from Cambridge and Pragmatic Semiconductor and Fractile are building novel chips; the alignment and evaluation bench is the deepest anywhere, and it is the capability the frontier labs actually need; and London still clears a disproportionate share of the world's specialty insurance and commercial contracts. Deepen those until they are hard to route around, then trade them — evaluation capacity for compute guarantees, on terms set while the position holds. The Compute Roadmap's twenty-fold expansion to 420 AI exaflops by 2030 and the SMR programmes matter because they price that trade higher, not because Britain needs to be self-sufficient.
Strategic play — Specialist Power
Stop trying to rebuild every layer and specialise in the one the frontier labs need most: evaluation and alignment, where Britain's bench is the deepest anywhere. Deepen it alongside Arm and London's legal and insurance weight, then trade evaluation capacity for guaranteed compute while the position still holds.
Leverage — the nearest position. No chokepoint today. Arm sits in nearly every chip on earth, but Britain holds no lever over where its designs go. The nearest thing to one is older: English law and London insurance sit beneath a large share of world commerce, and if AI liability settles anywhere, that becomes a position. September 2026 showed the difference between leverage and a favour: when Washington turned its frontier-model chokepoint, the AI Security Institute reportedly lost its early access to the newest models overnight.
Open United Kingdom in the live Index → — every source behind every number, and the comparison tool.