Can Nvidia persuade governments to pay for “sovereign” AI?

In late 2023, Jensen Huang, the CEO of Nvidia, introduced a bold concept: every nation should develop its own artificial intelligence (AI) system, tailored to local data, values, and infrastructure. He referred to these systems as 'AI factories,' emphasizing their potential to generate homegrown intelligence. This vision, dubbed 'sovereign AI,' is capturing the attention of policymakers globally. Recent initiatives have shown growing interest from governments. Earlier this year, the European Commission announced a €20 billion ($23 billion) plan to establish up to five 'AI gigafactories.' In just a couple of months, nations such as France, Germany, Saudi Arabia, South Korea, and the UAE have engaged in agreements to foster local AI development. Nvidia claims that around 20 countries are exploring the sovereign AI model. Huang's enthusiasm is understandable, as governments represent a significant business opportunity for Nvidia, which recently became the first company to achieve a $4 trillion market valuation. Jefferies, an investment bank, estimates that these sovereign initiatives could yield approximately $200 billion in cumulative revenue for Nvidia in the coming years, with total spending potentially reaching $1 trillion. Sovereign AI also serves as a strategic safeguard for Huang, particularly as major clients like Amazon, Alphabet, Meta, and Microsoft are venturing into chip production, which could diminish their dependence on Nvidia. Notably, Saudi Arabia anticipates acquiring several hundred thousand of Nvidia's advanced processors over the next five years, while the UAE plans to import half a million annually. However, the question remains: will taxpayer investments in sovereign AI be justified? Despite the rising enthusiasm, the concept is still being defined. While sovereign AI could help nations create tailored models, safeguard sensitive data, and expand technology access, it may not significantly reduce reliance on American technology, and could lead to wasted investments. Different countries are taking varied approaches to implementing sovereign AI. France is focusing on Mistral, a domestic model developer, which has partnered with Bpifrance, an investment firm, and Nvidia to create a significant AI data campus. Meanwhile, Saudi Arabia has launched Humain to oversee AI infrastructure, while the UAE’s G42 is tasked with similar responsibilities. Governments justify these projects for various reasons. Some aim to catch up to the U.S. in AI development, while countries like India are concerned that AI systems trained on foreign data may not accurately reflect local languages and values. There is also a push to maintain control over domestic data, especially in sensitive areas like healthcare. Yet, these initiatives are unlikely to achieve true self-sufficiency. Nvidia dominates the advanced AI chip market, holding a staggering 90% share of commercially available AI processors. Even China, which has made strides toward an independent AI ecosystem, has not yet developed a domestic equivalent to the most advanced American chips. While sovereign AI projects may help nations enhance data security and broaden access to technology, U.S. cloud giants could deliver these benefits more efficiently due to their scale and supplier negotiations. Despite Nvidia's contrasting views, companies like Amazon and Microsoft are already promoting sovereign cloud solutions with improved data control and local infrastructures. Kevin Xu from Interconnected Capital cautions that many sovereign AI projects risk becoming extravagant endeavors rather than effective solutions, potentially misusing taxpayer funds—though they may align well with Nvidia's interests.