Microsoft CEO Satya Nadella has redefined corporate AI strategy around what he calls "token capital"—a company's own AI systems and models. He insists firms must pair this with human expertise to avoid becoming dependent on a handful of dominant AI providers. The warning comes as enterprise AI spending surges past corporate forecasts.
Token Capital Defined
Nadella described token capital as proprietary AI capability a company builds and owns, including its systems and models. This isn't just about using AI tools—it's about creating integrated systems that grow with the business. Human capital becomes essential here: employee knowledge, relationships, and pattern recognition must guide the technology to deliver real value.
The Learning Loop Imperative
Companies shouldn't chase the strongest AI model, Nadella argued. The real opportunity lies in building learning loops where human and token capital compound. These loops turn workflows and human judgment into self-improving systems through private evaluations and reinforcement learning on a firm's own data. Continuous improvement happens when people and AI collaborate daily.
Hollowing Out Risks
Nadella warned that an AI economy dominated by a few major models would mirror globalization's hollowing out of industrial economies. That future lacks societal permission, he said. People won't accept an economy where only a few players capture AI's value while others become mere data suppliers. Control remains the core issue.
Sovereignty as the Standard
Firms must build agentic systems allowing them to swap base models without losing accumulated expertise, Nadella emphasized. This ability to maintain intellectual property control is his key test of sovereignty. Companies that fail this test risk ceding long-term control to external AI providers no matter how advanced their current tools are.
Urgent Spending Choices
Enterprise AI spending already outpaces corporate forecasts, making ownership decisions urgent. Companies now face a stark choice: build proprietary token capital or deepen dependence on major AI providers. The wrong path could leave firms vulnerable as the AI economy evolves. Companies must choose now: build proprietary AI or deepen dependence on major providers.
