Nadella warns businesses may pay for AI twice, and leak expertise

Satya Nadella warned businesses on X that AI could force them to “paying for intelligence twice”: once in money, and again in the proprietary knowledge they feed into the system to make it useful. Nadella called the risk the “Reverse Information Paradox,” which comes from prompts, tool use by agents and, most of all, the human corrections that teach a model how a company really works.

Nadella framed the problem as the mirror image of Kenneth Arrow’s Information Paradox, in which a buyer cannot know the value of information until it is revealed. In Nadella’s version, the enterprise reveals its own playbook while trying to extract value from a product it has already paid for. Patents were society’s answer to Arrow’s original paradox, but AI reverses the dynamic by rewarding the provider with more of the customer’s institutional knowledge over time.

When companies build on closed models from labs such as OpenAI and Anthropic, they can end up binding internal workflows, decision-making patterns and exception handling to a system they do not control. The result is less portability, weaker pricing power and a growing dependence on the vendor, which may learn more about the customer than the customer learns about the vendor.

Business customers have been scrutinising AI spending more closely as usage-based pricing makes real-world costs climb even when token prices fall. Palo Alto Networks chief executive Nikesh Arora and Coinbase chief executive Brian Armstrong have argued that smaller and cheaper models can handle much of corporate demand, a view that gives purchasing teams another reason to limit how much sensitive material they hand to frontier systems.

In June 2026, Nadella warned about a future in which a few AI models “eat everything they see” and capture the economic returns while industries watch their expertise get commoditised.