Sovereign AI Compute: EU Gigafactories, Gulf, India, Korea — 2026-09-18

Top developments

South Korea’s Record AI Budget Prioritizes GPU Sovereignty

South Korea’s Ministry of Science and ICT (MSIT) proposed a record-high budget of 29.64 trillion won ($21.5 billion) for 2027, representing a 24.5% increase from the previous year. The budget allocates 74% of its total to AI and advanced technologies, with a specific line item of 3.85 trillion won dedicated to purchasing 10,000 GPUs for the National AI Computing Center. This move aims to secure "top-tier" AI models and reduce dependency on foreign compute, positioning Korea as a major sovereign AI player in Asia.

Yotta Commits $12B for 80,000 Nvidia Vera Rubin GPUs in India

Indian data center operator Yotta Data Services has committed to a $12 billion expansion to deploy 80,000 Nvidia Vera Rubin and GB300 GPUs. The infrastructure will be spread across two campuses in India, marking one of the largest single-country sovereign AI compute commitments to date. This build-out supports India’s ambition to become a global AI hub while navigating US export controls on advanced chipsets.

EU Experts Call for €1.3 Trillion to Reach 45 GW AI Capacity by 2030

A new expert strategy report, covered by German media, argues that the EU must mobilize approximately €1.3 trillion to achieve 45 GW of AI compute capacity by 2030. This target is critical to closing the gap with the US and China. The proposal complements the EuroHPC Joint Undertaking’s ongoing tender for up to seven AI Gigafactories, which aims to leverage €30 billion in public and private investment.

NVIDIA: Gulf AI Success Depends on "AI Factories," Not Just Data Centers

In an interview with Asharq Al-Awsat, NVIDIA executives clarified that Saudi Arabia’s AI success will be measured by the value generated from "AI factories"—integrated systems of compute, energy, and industrial applications—rather than traditional data centers. This aligns with the Kingdom’s Vision 2030 goals and the Humain initiative, emphasizing that raw compute power must be coupled with efficient inference and local application development to be truly sovereign.

Local view

South Korea: The Dong-A Ilbo reports that despite global debates on slowing AI development, government investment in "Sovereign AI" has increased 26-fold over the past three years. The article notes that nations are prioritizing domestic capability over caution due to geopolitical risks.

Germany: Tagesschau highlights that German actors are currently competing for contracts under the EU’s new AI Gigafactory initiative, which offers billions in funding. The focus is on ensuring Germany secures a share of the seven planned gigafactories to maintain industrial competitiveness.

Gulf Region: Al Khaleej newspaper discusses how the UAE is entering a new phase where AI capabilities are being stress-tested for cybersecurity and national security implications, moving beyond mere adoption to critical infrastructure integration.

Context & numbers

• South Korea Budget: 29.64 trillion won (approx. $21.5 billion) for 2027; 3.85 trillion won specifically for GPU procurement.
• India Yotta Investment: $12 billion for 80,000 Nvidia Vera Rubin/GB300 GPUs.
• EU Compute Target: 45 GW of AI compute capacity by 2030, requiring ~€1.3 trillion in total mobilization.
• EU Gigafactory Tender: Up to 7 gigafactories; €10 billion public funding aiming to leverage €30 billion total investment. Deadline for proposals is November 12, 2026.

On the radar

• EuroHPC Submission Deadline: November 12, 2026, is the cutoff for proposals for the EU AI Gigafactory consortia.
• Huawei Announcement: Huawei is expected to reveal new AI chip technology this week, potentially challenging Nvidia's dominance in the Chinese market and affecting global supply chain dynamics.
• Nvidia-Anthropic Investment Rumors: Reports suggest Nvidia may invest up to $10 billion in Anthropic’s upcoming IPO, which could further consolidate compute-model partnerships.