Skip to main content
Access as of 2026-07-17

Countries

Every country with a tracked compute footprint, read across facilities, ownership, export-control access and sovereign programmes.

Four ratios per country, each between a headline and what the substrate holds

Each country as four ratios: how much of its tracked plant is operating rather than announced, how much of that plant traces to an owner reaching a public filing, how much of its announced programme money has been disbursed, and how much of its headline programme figure is government money. Every axis is a quotient of two named substrate columns computed inside that country, so the figure carries shape and the printed capacity carries scale.

Plant inserviceOwner atfiling gradeProgramme disbursedGovernment shareof headline0%100%
Plant in service
Of the tracked plant that carries a build status, how much is operating rather than announced or under construction? Not a measure of how much compute a country has — a small country running everything it built scores above a large one mid-construction.
Owner at filing grade
Of the tracked capacity, how much traces to an owner reaching a public filing? Not a measure of how opaque a country is — it is how far OUR ownership walk got, and the un-investigated share is drawn separately rather than folded in.
Programme disbursed
Of the money a government announced for its AI programme, how much has it disbursed? Not a measure of programme success or speed — a young programme and a stalled one both read low.
Government share of headline
Of the headline figure a programme announces, how much is the government’s own money? Not a judgement about private investment — it states what the headline is made of, nothing about whether the money is good.
United States
38.3 GW tracked
United Arab Emirates
5.2 GW tracked
France
2.0 GW tracked
Saudi Arabia
2.0 GW tracked
India
1.5 GW tracked
China
632 MW tracked
Italy
551 MW tracked
Malaysia
387 MW tracked
United Kingdom
202 MW tracked
South Korea
130 MW tracked
Finland
106 MW tracked
Singapore
105 MW tracked
Japan
67 MW tracked
Brazil
61 MW tracked
Taiwan
41 MW tracked
Portugal
33 MW tracked
Israel
33 MW tracked
Norway
29 MW tracked
Germany
25 MW tracked
Switzerland
16 MW tracked
Thailand
9 MW tracked
Australia
8 MW tracked
Iceland
7 MW tracked
Sweden
7 MW tracked
Spain
6 MW tracked
Russia
6 MW tracked
Canada
6 MW tracked
Vietnam
4 MW tracked
Denmark
3 MW tracked
Netherlands
2 MW tracked
Poland
2 MW tracked
Luxembourg
687 kW tracked
Czech Republic
548 kW tracked
Slovenia
155 kW tracked
Plate 1. Each card is one country, four axes, every axis a quotient of two substrate columns inside that country. Cards are all the same size: the radar carries shape, and tracked capacity is printed beneath because it is a scale, not a radius. The outline breaks wherever an axis has no value rather than passing through the centre — a gap is missing data, and only a country measured on all four axes shows a closed shape. On the ownership axis a faint bar runs outward to the share of capacity whose ownership has not been investigated: the true value lies somewhere along it. A ring is 100%; a circled vertex sits past it. Ordered by tracked capacity, which reflects what this substrate has ingested as much as what exists.

34 of 109 countries in the substrate resolve at least 2 axes and appear above; 10 resolve all four, and 25 carry a sovereign programme with money figures attached. The remaining 75 have a footprint but resolve fewer than 2 axes — one spoke is not a shape, and drawing it beside the others would invite a comparison the datum cannot carry. They keep their rows on the Overview table.

The two ownership and plant axes read the same materialized view the Compute Visibility Index is built from, so a country’s ownership figure here and there are the same number. The two programme axes read the sovereign programme records behind Sovereign AI, and divide by government-only money wherever a government publishes it — the United States reads fully disbursed against a $30M government-only figure while its headline is $2.6B, which is why both the quotient and its two operands are printed rather than the quotient alone.

Export-control access is not an axis. It is a five-value ordinal rather than a quotient, and its conditionality — entity gating, expiry dates, per-chip carve-outs — lives in prose that a radius cannot carry, so it is stated as a classification on the Overview table and in full on Export Controls, with its caveats intact. Designation counts are likewise absent: they track enforcement attention rather than exposure, and fell to zero through 2026 while the rule-making that constrains access continued.

Working detail — axis derivations and normalization

1 · Plant in service

Of the tracked plant that carries a build status, how much is operating rather than announced or under construction?

Numerator. capacity_mw_operational
Denominator. capacity_mw_operational + capacity_mw_construction + capacity_mw_announced
Source. mv_compute_visibility_by_country

Megawatt quotient within the country. Decommissioned and cancelled plant falls in no bucket and is excluded from both sides, so the denominator is status-classified capacity rather than all tracked capacity.

This axis is a part over its own whole and cannot exceed 100%.

2 · Owner at filing grade

Of the tracked capacity, how much traces to an owner reaching a public filing?

Numerator. tier_pct_capacity.Tier_1 (capacity in Tier 1)
Denominator. capacity_mw (all tracked capacity)
Source. mv_compute_visibility_by_country

Capacity-weighted share, already emitted as a percentage by the view and divided by 100 here. Weighted by megawatts rather than by facility count, so a gigawatt site counts for more than a megawatt one.

This axis is a part over its own whole and cannot exceed 100%.

3 · Programme disbursed

Of the money a government announced for its AI programme, how much has it disbursed?

Numerator. disbursed_usd
Denominator. announced_govt_only_usd, falling back to announced_usd only where no government-only figure exists
Source. sovereign_programs

USD quotient. The denominator is government-only money by design: execution can only be measured against what the government committed, not against private capital announced beside it.

A value above 100% is a real reading here and is drawn past the ring rather than clipped.

4 · Government share of headline

Of the headline figure a programme announces, how much is the government’s own money?

Numerator. announced_govt_only_usd
Denominator. announced_usd
Source. sovereign_programs

USD quotient, computed only where the headline has been decomposed. It never falls back to the headline for its own numerator, so an undecomposed programme is absent rather than reported as fully government-funded.

This axis is a part over its own whole and cannot exceed 100%.

Working detail — the table the plate is drawn from

Country radar axis values, with the numerator and denominator of every quotient and the stated reason for each axis that has no value.
CountryTrackedPlant in serviceOwner at filing gradeProgramme disbursedGovernment share of headline
United States38.3 GW36.3% 13.8 GW of 38.1 GW56.3% 21.5 GW of 38.3 GW100.0% $30.0M of $30.0M against government-only money1.2% $30.0M of $2.6B
United Arab Emirates5.2 GW0.1% 3 MW of 5.2 GW100.0% 5.2 GW of 5.2 GWGovernment publishes no disbursement figure0.7% $3.5B of $518.7B
France2.0 GW4.9% 100 MW of 2.0 GW1.6% 33 MW of 2.0 GW1.6% $43.6M of $2.7B against government-only money1.3% $2.7B of $201.3B
Saudi Arabia2.0 GW0.7% 14 MW of 2.0 GW0.0% 0 MW of 2.0 GWGovernment publishes no disbursement figure77.0% $77.0B of $100.0B
India1.5 GW5.4% 83 MW of 1.5 GW10.9% 168 MW of 1.5 GW3.8% $48.0M of $1.3B against government-only money100.0% $1.3B of $1.3B
China632 MW100.0% 632 MW of 632 MW26.7% 169 MW of 632 MWGovernment publishes no disbursement figure100.0% $55.7B of $55.7B
Italy551 MW6.8% 37 MW of 551 MW94.6% 522 MW of 551 MW55.8% $261.6M of $469.0M against the headline figure — no government-only figure is publishedHeadline is not decomposed into a government-only share
Malaysia387 MW90.4% 350 MW of 387 MW28.4% 110 MW of 387 MWGovernment publishes no disbursement figure10.2% $440.0M of $4.3B
United Kingdom202 MW55.4% 112 MW of 202 MW49.8% 101 MW of 202 MW6.7% $318.0M of $4.8B against government-only money100.0% $4.8B of $4.8B
South Korea130 MW9.5% 12 MW of 130 MW81.7% 106 MW of 130 MWGovernment publishes no disbursement figure2.0% $1.5B of $74.0B
Finland106 MW20.5% 22 MW of 106 MW0.0% 0 MW of 106 MW57.8% $157.5M of $272.5M against government-only moneyNo announced figure to measure against
Singapore105 MW19.3% 20 MW of 105 MW19.0% 20 MW of 105 MW27.0% $200.0M of $740.0M against government-only money100.0% $740.0M of $740.0M
Japan67 MW49.7% 33 MW of 66 MW7.1% 5 MW of 67 MW5.9% $470.0M of $7.9B against government-only money12.3% $7.9B of $64.0B
Brazil61 MW11.1% 7 MW of 61 MW0.0% 0 MW of 61 MWGovernment publishes no disbursement figure36.9% $1.6B of $4.2B
Taiwan41 MW10.1% 4 MW of 41 MW2.1% 887 kW of 41 MWGovernment publishes no disbursement figure100.0% $6.2B of $6.2B
Portugal33 MW100.0% 33 MW of 33 MW0.0% 0 MW of 33 MWNo sovereign AI programme trackedNo sovereign AI programme tracked
Israel33 MW8.9% 3 MW of 33 MW100.0% 33 MW of 33 MW102.9% $140.0M of $136.0M against government-only money100.0% $136.0M of $136.0M
Norway29 MW100.0% 29 MW of 29 MW0.0% 0 MW of 29 MW6.6% $20.9M of $317.0M against government-only money100.0% $317.0M of $317.0M
Germany25 MW100.0% 25 MW of 25 MW0.0% 0 MW of 25 MW100.0% $3.7B of $3.7B against government-only money61.7% $3.7B of $6.0B
Switzerland16 MW100.0% 16 MW of 16 MW0.0% 0 MW of 16 MW100.0% $113.0M of $113.0M against government-only money100.0% $113.0M of $113.0M
Thailand9 MW38.2% 3 MW of 9 MW0.0% 0 MW of 9 MWNo announced figure to measure againstNo announced figure to measure against
Australia8 MW24.8% 2 MW of 8 MW0.0% 0 MW of 8 MWGovernment publishes no disbursement figure8.6% $24.0M of $280.0M
Iceland7 MW100.0% 7 MW of 7 MW4.8% 354 kW of 7 MWNo sovereign AI programme trackedNo sovereign AI programme tracked
Sweden7 MW100.0% 7 MW of 7 MW0.0% 0 MW of 7 MWGovernment publishes no disbursement figureNo announced figure to measure against
Spain6 MW100.0% 6 MW of 6 MW0.0% 0 MW of 6 MW10.1% $165.0M of $1.6B against the headline figure — no government-only figure is publishedHeadline is not decomposed into a government-only share
Russia6 MW100.0% 6 MW of 6 MW0.0% 0 MW of 6 MWNo sovereign AI programme trackedNo sovereign AI programme tracked
Canada6 MW100.0% 6 MW of 6 MW0.0% 0 MW of 6 MWGovernment publishes no disbursement figure100.0% $2.6B of $2.6B
Vietnam4 MW3.2% 137 kW of 4 MW0.0% 0 MW of 4 MWGovernment publishes no disbursement figure0.5% $38.5M of $7.0B
Denmark3 MW100.0% 3 MW of 3 MW0.0% 0 MW of 3 MWNo sovereign AI programme trackedNo sovereign AI programme tracked
Netherlands2 MW100.0% 2 MW of 2 MW60.5% 1 MW of 2 MWGovernment publishes no disbursement figure65.0% $141.7M of $218.0M
Poland2 MW100.0% 2 MW of 2 MW0.0% 0 MW of 2 MWNo announced figure to measure againstNo announced figure to measure against
Luxembourg687 kW100.0% 687 kW of 687 kW0.0% 0 MW of 687 kWNo sovereign AI programme trackedNo sovereign AI programme tracked
Czech Republic548 kW100.0% 548 kW of 548 kW0.0% 0 MW of 548 kWGovernment publishes no disbursement figure50.0% $23.0M of $46.0M
Slovenia155 kW100.0% 155 kW of 155 kW0.0% 0 MW of 155 kWNo sovereign AI programme trackedNo sovereign AI programme tracked