Data 2026-07
Sovereign AI
Announced commitments against government budgets and reported disbursements, by country.
Announced commitments against government budgets and reported disbursements, by country.
Sovereign programme record · South Korea · 2025–2030
Every figure below derives from the same record the dashboard reads — this page is the programme as a citable object: the reconciliation with its denominator stated, the decomposition with its residual, and the dated evidence behind each claim.
Pipeline stage: Deployed (primary-verified evidence) — 325,136 accelerators tracked. The stage is a Scrutica derivation (tier 4), computed from the figures above, never asserted by the source. Announced in KRW, converted at 1468.62 as of 2025-12-15 via BOK reference rate via FRED DEXKOUS (https://fred.stlouisfed.org/series/DEXKOUS).
| Component | Kind | Amount |
|---|---|---|
| Government: $381M for 5 sovereign AI model consortia (KRW 556B) | Government | $381M |
| Government: KRW 1.6341T ($1.1B) for NAICC compute | Government | $1.10B |
| Private sector (Samsung, SK, Hyundai, Naver, etc.) | Private | — |
| 3 GW DC in Jeollanam-do ($10B initial, expanding to $35B) | Private | $10.00B |
| KEPCO $45B power infrastructure (listed company, partially state-owned) | State enterprise | $45.0B |
The residualThe amount-bearing components sum to $56.5B against the $74.0B headline — a $17.52B gap the components do not account for. The decomposition is published as the record carries it; the residual is stated rather than smoothed.
| Model | Count | Status | Owner |
|---|---|---|---|
| Vendor unspecified (gov. 2028 target) | 52,000 | procurement_active | Ministry of Science and ICT |
| Vendor unspecified (gov. 2030 target) | 260,000 | announced | Ministry of Science and ICT |
| NVIDIA B200 | 10,080 | procurement_active | government (National AI Computing Center) |
| NVIDIA H200 | 3,056 | procurement_active | government (National AI Computing Center) |
| Partner | GPUs | Model | Stage | Award |
|---|---|---|---|---|
| NHN Cloud | 7,656 | NVIDIA B200 | awarded | Research database · 2025-07-01 |
| Naver Cloud | 3,056 | NVIDIA H200 | awarded | Research database · 2025-07-01 |
| Kakao | 2,424 | NVIDIA B200 | awarded | Research database · 2025-07-01 |
NVIDIA hardware: high · TSMC fabrication: high · US technology: high · BIS jurisdiction reach: low. Dependence levels are Scrutica classifications (tier 4) from the programme’s own disclosures — which hardware it buys, where it fabricates, and whose export jurisdiction its supply chain crosses.
Named partners: NVIDIA, Samsung Electronics, SK Group, Hyundai, MSIT, Naver, SK Telecom, LG, NCSoft, Upstage, KEPCO.
The ₩100 trillion headline conflates government R&D, private investment and industrial transformation. $381M is verified as government-only for the sovereign AI model consortia program: five were selected in August 2025 (Naver Cloud, SK Telecom, LG AI Research, NC AI, Upstage), narrowed to three at the January 2026 first cut (LG, SK Telecom, Upstage). NAICC's 10,080 B200 and 3,056 H200 total 13,136 GPUs, all in procurement, with no operational partition yet. Targets run to 52K GPUs by 2028 and 260K by 2030, alongside a 3 GW data center in Jeollanam-do. The AI Basic Act took effect in January 2026. NAICC's special-purpose company, 'Korea AI Computing Center', completed registration around 2026-06-09/13 (Samsung SDS 30%, government 29%, Naver Cloud 26.1%, chief executive Ahn Jung-tae); the Haenam groundbreaking remained pending as of 2026-07-14.
Machine surfacesThis record is one object under one id: the evidence-trail feed and the derivation-chain export serialize the same programme this page renders.