Machinepower Index / Dossiers / Spain
Rank 11–18 across the scoring rules tested. How this is tested
Spain
Strategic Autonomy — For economies big enough to cover most layers without leading any: build enough compute, open models and a resilient grid to depend on no single supplier and bargain from strength.
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: Spain controls 4,500 H100-equivalents, wherever the sites are. 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 | 6.0 |
| Firm Power Capacity | Dataset | 5.4 | |
| Sovereign Compute | Dataset | 5.9 | |
| Hardware & Silicon Access | Dataset | 4.8 | |
| Weights | Frontier Model Output | Assessed | 3.8 |
| Alignment Capability | Assessed | 6.6 | |
| Regulatory Standing | Dataset | 8.8 | |
| Talent Density | Dataset | 5.9 | |
| Will | Policy Capacity | Dataset | 9.2 |
| Public Sector Adoption | Dataset | 7.8 | |
| Workforce Transition | Dataset | 7.2 | |
| Public Trust | Dataset | 6.5 |
Assessed marks a measurement we judged rather than took from a dataset: 3 of the 12 here.
Sources
Requests for new grid access ran eighty times the previous plan's, and oversubscribed nodes now go to tender.
View source →Runs the EU's only standalone AI agency, AESIA, and the first AI Act sandbox; signed Bletchley and Seoul.
View source →A cross-check for planning and permitting, which stays assessed: it scores permits and electricity connections for ordinary firms, not grid-scale connections for data centres.
View source →Largest: MareNostrum 5.
View source →ALIA-40B, 40 billion parameters, trained from scratch on MareNostrum 5 on about 9.4 trillion tokens; no Spanish model met Stanford's notability bar in 2025.
View source →Talent is a log scale on the count.
View source →Feeds governance and public-sector adoption; policy capacity is listed separately.
View source →The Index scores public trust as 1 + share ÷ 10. Survey of 23,532 adults across 32 countries, March to April 2026.
View source →Feeds workforce, rescaled across 173 economies.
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 →Bottlenecks
These hold Spain below its Potential. Its two lowest scores cost it the most: frontier model output at 3.8 (part of Weights) and hardware & silicon access at 4.8 (part of Watts).
Assessment
Spain comes closest of the new nations to a full set: a public supercomputer, open models of its own and a state AI regulator. What holds it back is frontier work and a grid that cannot connect every data centre that wants to come.
The art of the possible
The grid is the lever. Spain's solar surplus is real; if it can connect that power to the campuses queuing in Aragón, it becomes Europe's cheapest place to run AI at scale, with its own models and machines to stop it being only a landlord.
Strategic play — Strategic Autonomy
Pair cheap power with public compute: connect first the campuses that share capacity with MareNostrum and ALIA, so foreign money feeds a national stack rather than replacing it.
Leverage — the nearest position. No chokepoint. The nearest thing is cheap solar power in a country that also builds its own models and computers, which no other southern European state does.
Open Spain in the live Index → — every source behind every number, and the comparison tool.