Machinepower

Machinepower Index / Dossiers / Sweden

Rank 20 of 50 · Tier 3

Rank 20–25 across the scoring rules tested. How this is tested

Sweden

Compute Rentier — For nations whose main asset is the ground itself, meaning cheap clean power, land and stability: host other nations' compute and turn geography into rent, jobs and relevance.

Score5.39
Potential6.20
Bottleneck cost−0.81
Weakest cell: Frontier Model Output at 3.0 Binding layer: Weights Direction: Improving Edition: Q3 2026 · updated October 2026

How the Score, Potential and bottleneck cost are worked out: methodology.

The three layers

WattsMaterial power
5.4 / 10

Energy, grids, data centres and silicon — the physical capacity to run models at scale.

Compute tracked by Epoch AI: Sweden controls 400 H100-equivalents, wherever the sites are; has 4,500 on its soil with the owner not reported. Hosted compute is shown, not scored.

WeightsIntelligence power · holds the weakest cell
5.6 / 10

Models, alignment capability, standards influence and the talent that produces them.

WillPolitical power
7.6 / 10

The capacity to wire AI into the state and the economy while holding public consent.

Where the score comes from

LayerSub-dimensionBasisScore
WattsPlanning & PermittingAssessed6.4
Firm Power CapacityDataset4.7
Sovereign ComputeDataset5.4
Hardware & Silicon AccessDataset4.9
WeightsFrontier Model OutputDataset3.0
Alignment CapabilityAssessed6.0
Regulatory StandingDataset8.2
Talent DensityDataset5.3
WillPolicy CapacityDataset8.9
Public Sector AdoptionDataset7.4
Workforce TransitionDataset8.5
Public TrustDataset5.5

Assessed marks a measurement we judged rather than took from a dataset: 2 of the 12 here.

Sources

Svenska kraftnät (summarised by a grid-queue analysis)2026
Planning & permitting (assessed)
6.4/10, assessed

Connections are first come, first served, but stalled projects lose their place; the north builds fast.

View source →
Swedish Government (via EU AI Act trackers)2025–2026
Alignment (assessed)
6.0/10, assessed

Its privacy and telecoms regulators enforce the AI Act; signed neither frontier-safety statement.

View source →
World BankDecember 2025
Business Ready 2025: business location and utility services
Business location 71.8/100 · utility services 77.5/100

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 →
Ember2025
Electricity Generation
170 TWh (2025)

Feeds firm power.

View source →
TOP500.orgJune 2026
High-Performance Compute Capacity
134 PFlop/s across 9 systems

Largest: Arrhenius.

View source →
Stanford AI Index 2026April 2026
Top AI authors and inventors
8,522 (Zeki Data)

Talent is a log scale on the count.

View source →
Oxford InsightsMay 2026
Government AI Readiness Index 2025
#23 of 195 · 69.5/100

Feeds governance and public-sector adoption; policy capacity is listed separately.

View source →
Ipsos AI Monitor 2026April 2026
Public Trust in AI
47% say AI products 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 →
IMF2023 data, published June 2024
AI Preparedness Index: human capital and labour market policies
0.171 of a possible 0.25 · 8.5/10 in the Index

Feeds workforce, rescaled across 173 economies.

View source →
Ember2024
Firm generating capacity
32 GW (2024)

Fossil, nuclear, hydro and other dispatchable plant; averaged with generation for firm power.

View source →
GitHub Innovation Graph; Oxford Insights2026
Developer base
1.0 million developers

Second source for talent, averaged with Stanford's count.

View source →
University of Melbourne; KPMGJanuary 2025
Trust in AI at work
39% trust AI at work

Second source for public trust, averaged with Ipsos.

View source →
Oxford InsightsMay 2026
Policy capacity
89/100

Feeds policy capacity.

View source →
Stanford AI Index 2026April 2026
Private AI investment, 2025 (shown, not scored)
$1.0bn

Capital is shown beside the Score, not scored.

View source →

Bottlenecks

These hold Sweden below its Potential. Its two lowest scores cost it the most: frontier model output at 3.0 (part of Weights) and firm power capacity at 4.7 (part of Watts).

Weights (Intelligence)Frontier Model Output

Assessment

Sweden has the north's advantages — hydro, nuclear, cold air and room to build — and a deep research base, but no frontier models and little compute it owns outright.

The art of the possible

Its power and land are winning large campuses. The gain comes if it keeps a share of that capacity for Swedish research and firms, as the Berzelius and Arrhenius machines show it can.

Strategic play — Compute Rentier

Rent the north on terms that reserve capacity for Swedish science and industry, before the queue in the south decides for it.

Leverage — the nearest position. No chokepoint. Its position is cheap, clean power in the north, which counts for more each time Denmark, Ireland or the Netherlands closes its grid to data centres.

Open Sweden in the live Index → — every source behind every number, and the comparison tool.