The pursuit of sovereign AI strategies is reshaping national industrial policies worldwide. Nations are moving beyond ambition and investment levels towards controlling critical AI infrastructure and regulations. As Vish Nandlall analyzes, the ability to integrate power, infrastructure, regulation, and demand is now paramount.
The global commitment to AI is staggering, nearing a trillion dollars in pledges. Yet, much of this total reflects intentions rather than actual expenditures. A prime example is Stargate’s pledge of up to $500 billion towards this AI mania.
Sovereign AI strategies aim for supply security, legal authority, and value capture, yet these objectives become conflated. Supply security ensures access to essential resources like compute power. Legal authority refers to having domestic laws govern data and infrastructure. Value capture looks to guarantee that domestic firms yield sustainable profits.
Many nations employ different AI models to achieve these outcomes. France exemplifies the “dealmaker” model. At a recent summit, it announced commitments of €109 billion for AI infrastructure. This was aided by foreign investments such as SoftBank‘s €75 billion contribution and land provision for AI campuses. Notably, France’s focus allows foreign capital to play a major role, offering supply security and negotiating leverage but not yet ensuring substantial domestic value capture.
The “procurement” model is visible in Korea and Japan. Korea has committed to acquiring 10,000 GPUs quickly to kick-start its national AI computing center. This model focuses on aligning silicon supply with industrial demand, accelerating access to advanced AI capabilities. Japan emphasizes demand-side management, subsidizing domestic developers for achieving delivery milestones.
The European Union explores the “subsidy” model through its InvestAI program, targeting substantial mobilized investments. With European-led consortia, the EU aims to ensure that facilities remain sovereign assets while dealing with high electricity costs, revealing the challenges of securing local technological value amidst foreign races.
Singapore and the UK showcase “governance” models. Singapore focuses on providing regulatory certainty and sectoral AI deployment programs. In the UK, planning approval and direct power purchase agreements are bundled to expedite AI project execution, illustrating the power of regulatory streamlining.
In contrast, many regions struggle with the “dependency” model, wherein sovereign AI strategies rely heavily on international cloud providers like Microsoft and AWS. While these arrangements offer data and access advantages, they often compromise domestic value capture.
The effective state strategies identified by Nandlall highlight the importance of focusing on utilities they control. Currency, power, and demand must be harnessed to generate genuine AI sovereignty rather than renting foreign-developed solutions. The true test remains whether these models can transition from theoretical promises to practical, effective implementations. The next frontier might not be another GPU farm but rather the development of an open and governable software control plane.
This raises critical considerations for middle powers, suggesting that collaboration might trump standalone capital investments. This realization could redefine the AI sovereignty race and determine the true technological leaders of tomorrow.
