The Microsoft CEO says data sovereignty and multi-model infrastructure will shape AI competitiveness, warning that dependence on a single model or platform can leave enterprises exposed.
Microsoft CEO Satya Nadella said companies that surrender control of their AI data risk losing their independence as businesses, and warned that relying on a single AI model or external platform could leave them unable to compete. In his interview with Fareed Zakaria, Nadella said enterprises should be careful about how much information they share with model providers, arguing that prompts, response data and especially metadata generated during model use should remain under company control so they can later train their own weights or adapt open models. He described data sovereignty as more important than the performance of any individual model and called for AI systems in which the model is separated from context, memory and core controls. Nadella said companies should build internal infrastructure such as an AI gateway to centrally manage prompts and data, allowing them to switch among multiple models based on their strengths and avoid being stranded if one provider disappears. He also cautioned against tying development workflows too closely to embedded coding tools such as Anthropic's Claude Code or OpenAI's ChatGPT Codex. The comments drew attention because Microsoft is OpenAI's largest shareholder in its for-profit entity with an approximately 27% stake, yet Nadella still argued against single-model dependency. On competition with Chinese AI models, he said U.S. companies retain an edge through trust in the American technology ecosystem and because open-weight Chinese models can run on U.S. hyperscaler cloud infrastructure, where they can be monitored, tested and post-trained. The remarks come after market shocks linked to DeepSeek and, more recently, Moonshot AI's Kimi K3, and echo a broader industry shift toward open-weight models, proprietary AI platforms and reduced vendor lock-in.