Microsoft CEO Satya Nadella says companies adopting AI face costs beyond subscription or compute fees, arguing that they effectively “pay for intelligence twice.” In a post shared on social media, Nadella frames the issue as a “Reverse Information Paradox,” describing how enterprises must provide prompts, evaluations and corrective feedback to get reliable performance from AI systems. He argues that this process turns operational and institutional know-how into “intelligence exhaust” that gradually helps model providers improve. Nadella contrasts this with economist Kenneth Arrow’s original “information paradox,” where information can lose value to a seller once a buyer sees it, saying AI reverses the direction of that risk.
He also notes that patents historically offered a partial solution by enabling disclosure without fully giving ideas away for free, and says the AI sector needs a similar safeguard that goes beyond data protection to cover how organizations learn and adapt using AI. Nadella proposes a five-part “five Cs” approach—Control, Capability, Choice, Cost, and Compound—focused on keeping evaluation, learning loops, and institutional memory within enterprise boundaries and reducing dependence on a single provider. He adds that he supports fair use for public data but argues restrictive terms for enterprises are inconsistent if providers still retain learning from customer usage.
