While Satya Nadella’s warning about the risks of proprietary AI is grounded in valid concerns regarding data ownership, the proposed solution of sovereign AI may be out of reach for many organizations. Building, maintaining, and securing private AI infrastructure requires significant capital, specialized technical talent, and massive computational resources that are typically only available to the largest corporations. For small and medium-sized enterprises, the trade-off between using powerful, pre-trained models and the risk of data exposure is often a matter of survival. These firms may lack the capacity to develop their own models and must rely on third-party providers to remain competitive.
Furthermore, the argument that companies are paying twice for AI ignores the immense value that these models provide in terms of productivity and efficiency. For many, the ability to leverage a world-class AI model—even with the associated data-sharing risks—is what allows them to scale operations and compete in a global market. The cost of not using these tools could be far higher than the potential loss of some institutional knowledge. For these businesses, the focus should perhaps be on better contractual protections and improved privacy-preserving technologies rather than a wholesale shift toward sovereign AI.
There is also a risk that the push for sovereign AI could lead to increased fragmentation in the technology landscape. If every company attempts to build its own siloed AI, the industry may lose the benefits of collective learning and standardized innovation. A more balanced approach might involve demanding greater transparency and stricter data-usage policies from AI providers, rather than encouraging a retreat into private, isolated systems. The challenge is to find a middle ground that protects corporate interests without stifling the accessibility that has made AI a transformative force for businesses of all sizes.