Microsoft's Satya Nadella Outlines Vision for AI Sovereignty in Era of Cognitive Loops

At a time when the US administration’s restrictions on Anthropic’s Fable 5 and Mythos 5 AI models have been sparking debates on AI sovereignty, Microsoft’s Satya Nadella has some advice for corporations that look at AI as another digital system to enhance output. Nadella urges the people behind these corporations to use both their ‘human capital’ and ‘token capital’ to create a ‘cognitive loop’, which will eventually create value. In a detailed article shared on X, the Microsoft CEO spoke in depth about a strategic vision for the AI era, highlighting that true sovereignty for companies and nations lies not merely in data localisation or physical infrastructure. Instead, it can only be achieved by controlling proprietary knowledge embedded within AI systems. “This transition is different than any previous platform shift,” states Nadella. “In the past, we used digital systems to enhance human capital,” he adds, referring to the previous technological shifts in corporate world. Now, though, AI creates a “real cognitive loop” between people and digital systems, where human judgment and machine intelligence continuously reinforce each other, making a valuable system. In his knowledge piece, Nadella refers to two major assets, which he claims hold the true value. – The first is human capital, which encompasses knowledge, judgment, relationships, ingenuity, and pattern recognition. – The other one is token capital, i.e., the firm’s AI-derived digital intelligence, including custom models, workflows, and the traces of human-AI interactions. Nadella says that companies must build self-improving “learning loops” to capture institutional expertise, convert it into model weights and contexts, and compound advantages over time, rather than leaking value to external frontier AI models. “This loop becomes the new IP of the firm. I think of it as a hill-climbing machine. And unlike most assets, it compounds. Every improved workflow generates better training signal, which accelerates the accumulation of tacit knowledge unique to the firm,” says Nadella. “The companies that build this early will have an advantage that is hard to replicate, regardless of any new individual model capability,” he adds. The Microsoft CEO also argues against over-reliance on a handful of frontier AI labs, which risks “commoditising intelligence and concentrating power.” Instead, he advocates for broad ecosystems where every organisation, regardless of size, develops its own compounding AI capabilities hosted on platforms like Azure. Nadella’s approach preserves competitive moats, maintains national and corporate control over sensitive knowledge, and fosters a positive-sum economy. In this case, Microsoft positions itself as the enabler, offering sovereign cloud solutions, governance tools, and the ability to run multiple models while keeping data and fine-tuned weights under customer control. This proposal aligns with broader geopolitical trends, as governments worldwide are demanding greater control over critical AI infrastructure for national security, regulatory compliance, and economic resilience. Nadella’s proposition drew attention from one of his industry rivals, Elon Musk . In response to Nadella’s article, Musk simply posted “Interesting.” This comes after SpaceX’s recent IPO, which has drawn attention to issues of national security and export controls.