In a widely circulated essay published on July 12, Microsoft (NASDAQ: MSFT) CEO Satya Nadella identified a fundamental imbalance in how companies deploy artificial intelligence (AI).
He argued that enterprises essentially pay for intelligence twice, first through tokens and again with the proprietary knowledge they must reveal to make AI useful
Every prompt and correction an organization feeds a model becomes “exhaust,” a trail of insights about how an organization operates that leaks out, “trace by trace.” To counter this “Reverse Information Paradox,” Nadella argued that companies must take control of their data by running AI within their own “tenant boundary.” The cost of protecting enterprise IP Today, companies use three primary paths to access frontier AI models. They can use large cloud platforms such as Azure OpenAI Service or Amazon’s (NASDAQ: AMZN) AWS Bedrock, a direct application programming interface (API) from model providers, or a hybrid of both. The cloud path doesn’t require expensive hardware, and the direct route is well established by companies like Anthropic, which built a multibillion-dollar business on direct enterprise contracts.
Regardless of the path, the risks Nadella described remain. The model provider’s infrastructure inevitably captures the usage patterns and query data that reveal how an enterprise operates. In response, sovereign nations and regulated entities are increasingly investing in their own hardware.