Microsoft CEO Satya Nadella has issued a sharp warning to companies rushing to adopt artificial intelligence: the cost of AI may be far higher than the subscription bill, because businesses could be handing over the proprietary knowledge that makes them valuable.
Nadella Says AI Buyers Are Paying Twice
Nadella’s central argument is that enterprises are not just buying intelligence from AI providers. They are also teaching those providers how their businesses work.
TechCrunch reported that Nadella wrote companies “pay for intelligence twice,” first through token usage and second through proprietary knowledge they must reveal to make the model perform well.
The Times of India shared that Nadella warned AI flips Kenneth Arrow’s information paradox because, in the AI age, the buyer risks giving away knowledge just to use what they bought.
The concern is not only raw data. Nadella pointed to AI “exhaust,” including prompts, tools agents call and corrections employees make when the model gets something wrong.
Model Providers Face a Fairness Challenge
Nadella also questioned whether AI providers are applying different rules to themselves and their customers. Nadella criticized the irony of model providers relying on fair use to train on public data while imposing restrictive terms on distillation. The CEO warned value flows toward whoever owns the AI learning infrastructure if learning only moves in one direction.
That argument lands at a tense moment for the AI industry.
The Wall Street Journal reported that Nadella, who helped usher in the AI boom, is now joining a wider effort to challenge the power balance around AI giants such as OpenAI and Anthropic. Nadella’s vision for the next AI wave involves cheaper models, more user control and political messaging that wins the public’s trust.
The Five-Part AI Playbook
Nadella’s solution is not to avoid AI, but to redesign how companies use it. Nadella proposed a five-part framework: Control, Capability, Choice, Cost and Compound. He wants companies to retain ownership of prompts, feedback, memory and other data while building proprietary learning environments on cloud infrastructure.
A major part of the advice is avoiding dependence on one model provider. Nadella urged companies to decouple orchestration from any single model so losing one vendor does not mean losing capability. AI gateways and model-switching tools are becoming more popular as companies try to move between models instead of being locked into one provider.
Open Models and Enterprise Control Gain Momentum
The warning also points toward a broader shift in enterprise AI. Large companies are increasingly considering open-source models installed on their own premises because they cost less and give businesses more control. Solo.io founder and CEO Idit Levine shared that some customers ask whether they can run an open-source model on-premise if it can deliver almost 90% of what larger models do at much lower cost.
The trend is already visible in developer infrastructure. Open models accounted for 29% of all traffic routed through Vercel’s gateway last month.
AI Adoption Becomes an Ownership Question
Nadella’s warning matters because AI adoption is no longer only about productivity. It is about ownership. Companies using AI to write code, analyze customers, automate operations or summarize internal decisions may be creating valuable learning loops every day.
The danger, as Nadella frames it, is that businesses may train external systems on their own expertise without capturing the long-term benefit themselves. His message is not anti-AI. It is a warning that enterprises should not let their knowledge, feedback and workflows quietly become someone else’s competitive advantage.
For companies, the next AI question may no longer be simply which model performs best. It may be who gets to keep the intelligence created while using it.
