Political considerations of NVIDIA’s “sovereign AI” concept

In late 2023, Jensen Huang, CEO of Nvidia, introduced the concept of “sovereign AI”—the idea that each country should build its own artificial intelligence systems tailored to national data, values, and infrastructure. Framing these systems as “AI factories,” Huang promoted the initiative as a means for governments to enhance digital sovereignty while appealing to their interest in domestic manufacturing and technological control.

The idea has quickly gained traction among world governments. The European Commission proposed a €20 billion fund to develop up to five AI “gigafactories,” with an article by The Economist recalling that countries like France, Germany, Saudi Arabia, South Korea, and the United Arab Emirates (UAE) have announced similar initiatives. Nvidia claims at least 20 countries are now pursuing sovereign AI systems.

For Nvidia, which recently became the first company to hit a $4 trillion market cap, this global interest presents a massive business opportunity. Analysts at Jefferies estimate sovereign AI projects could generate around $200 billion in cumulative revenue for Nvidia, with potential long-term spending reaching $1 trillion. Moreover, these initiatives help Nvidia hedge against risk, as major tech clients like Amazon, Google, Meta, and Microsoft are developing their own in-house AI chips and could become less reliant on Nvidia over time.

However, whether sovereign AI will deliver good value for taxpayers is less clear. While such projects promise greater control over data and wider access to AI tools, they may not significantly reduce dependence on US companies—and could lead to substantial financial waste if not properly executed.

Different countries are pursuing sovereign AI in varying ways. France’s strategy centers on Mistral, a domestic AI firm, which has partnered with Bpifrance, Nvidia, and Emirati investors to build Europe’s largest AI data campus. In contrast, Gulf nations are taking more centralized approaches: Saudi Arabia created a national AI firm called Humain, and the UAE designated G42—an AI company backed by the sovereign wealth fund Mubadala—to lead its efforts.

Governments justify sovereign AI projects on several fronts. Some, like the European Commission, want to catch up with US AI leadership. Others, such as India, are concerned about cultural and linguistic biases embedded in AI models trained on foreign, especially English-language, data. They argue that national models are needed to better serve local populations, especially in sensitive areas like education and public services.

Attraction beyond money

Another major motivation is data protection. In sectors like healthcare, there are concerns about sharing patient data with foreign-owned AI platforms. Sovereign systems promise to keep such sensitive data within national borders while still allowing it to be used for AI development.

Governments also want to democratize access to AI tools. Nadia Carlsten, head of Denmark’s national AI supercomputer project Gefion, highlights how smaller companies and research institutions often struggle to access powerful AI platforms dominated by tech giants. Gefion, launched in 2024, aims to level the playing field by providing local and affordable AI resources for tasks such as drug discovery and weather forecasting.

Despite these goals, true AI self-sufficiency remains elusive. Advanced AI models require cutting-edge chips—most of which are made by Nvidia, which controls around 90% of the global AI chip market. Its only real rival, AMD, is also American. The servers used to run AI models are similarly dominated by US firms like Dell and Supermicro. Even China, with its massive investment in self-reliant AI infrastructure, still lacks top-tier alternatives to American hardware.

While sovereign AI may help countries secure data and support under-resourced users, they are unlikely to fully escape dependence on US cloud giants and chipmakers. In fact, firms like Amazon and Microsoft are already offering “sovereign cloud” services with localized data control and infrastructure. National AI models could potentially be developed on top of these existing platforms—raising questions about the true value of starting from scratch.

By Nazrin Sadigova