AI Compute Ownership Race Driven by Sovereignty Concerns

The concept of "the cloud" as a detached, abstract utility for computing has been fundamentally challenged by the rise of artificial intelligence, leading to a significant global effort to build AI compute infrastructure that owners can reliably control. This shift is driven by concerns over "sovereign AI," a term that, while lacking a precise universal definition, broadly refers to the desire for nations and organizations to maintain autonomy over their critical AI processing capabilities. The urgency of this movement was highlighted in May 2025 when Karim Khan, chief prosecutor of the International Criminal Court (ICC), experienced an email account lockout attributed to U.S. sanctions against Israeli officials, which implicated Microsoft, a U.S. company. Although Microsoft president Brad Smith denied cutting off ICC services, the incident underscored how geopolitical actions can directly impact access to essential digital infrastructure, regardless of its physical location. This event served as a stark illustration that the control over computing infrastructure ultimately rests with the entity that holds legal authority over the company providing the service, irrespective of where the data is physically stored.

AI models, particularly frontier models, are resource-intensive and depend on specific hardware, power grids, and operate under the laws of their host countries. This physical and legal dependency means that whoever controls the entire stack of infrastructure – from the physical machines to the power supply and legal jurisdiction – can exert significant influence through taxation, subpoenas, or outright shutdowns. The substantial investments required for building this infrastructure, involving concrete and power lines, result in systems that are fixed in place, making the race to secure reliable ownership of AI compute a paramount concern. This has spurred a scramble for what is increasingly termed "sovereign AI," a concept that gained considerable traction at the U.N.'s AI for Good Global Summit in Geneva.

Discussions at the summit revealed a spectrum of interpretations for sovereign AI. Some definitions prioritize keeping sensitive data within national borders, aiming to prevent foreign access and control. Others focus on the development of indigenous AI models that can compete with or surpass those developed by leading nations like the United States and China. Hakim Hacid, chief researcher at the UA, was among those discussing these evolving concepts. The pursuit of sovereign AI is not merely a technical challenge but a geopolitical imperative, as nations seek to secure their economic competitiveness, national security, and digital autonomy in an era where AI is becoming increasingly central to global power dynamics. The implications extend to research institutes, standards bodies, and startups, all of whom are grappling with how to define and achieve AI sovereignty in a complex and interconnected world. The trend suggests a move away from complete reliance on a few dominant cloud providers towards a more distributed and nationally controlled AI infrastructure landscape.