Nadella warns of 'reverse information paradox' as AI data risks take centre stage

Microsoft chief executive Satya Nadella has issued a stark warning regarding the hidden costs of artificial intelligence, arguing that businesses risk surrendering their most valuable assets when they adopt generative models. In a blog post titled "Sn Scratchpad," Nadella described a "reverse information paradox" where companies effectively pay for intelligence twice: once in cash and again by revealing the internal data, workflows, and corrections that make the technology useful.

Nadella contends that for AI systems to perform effectively, they require high-quality internal context, including employee prompts, operational procedures, and corrections made when models err. He noted that every correction is distilled into institutional know-how, meaning that in consuming intelligence, users are simultaneously creating it. "What you create should belong to you," Nadella wrote, highlighting the tension between utilising purchased tools and retaining competitive advantage.

This perspective aligns closely with recent warnings from Palantir chief executive Alex Karp, who has argued that enterprises are inadvertently donating intellectual property to model providers. Karp advocates for Palantir’s Ontology platform, an application layer designed to connect models to company operations while controlling data access. He asserts that such a layer prevents large language models from caching customer data or replicating business logic, ensuring that proprietary knowledge remains within the organisation.

The debate over data ownership and valuation comes as the broader artificial intelligence sector faces intense scrutiny. Reuters reports that Amazon, Microsoft, Alphabet, and Meta are projected to spend approximately $630 billion on data centres and AI chips in 2026, a figure more than four times their 2023 guidance. Despite this massive capital expenditure, prominent investors including Ray Dalio, Jeremy Grantham, and Michael Burry have expressed concerns about AI valuations, with Burry describing semiconductor valuations as a "pure form of overvaluation."

Market sentiment reflects this growing anxiety. A Bank of America survey found that 45 per cent of fund managers view an AI bubble as the market's biggest tail risk. For Palantir, which trades at 88 times non-GAAP forward earnings compared to a sector median of around 25 times, Nadella’s comments may validate the need for its data-protection strategies. However, the stock has declined 27 per cent over the past six months and more than 26 per cent year-to-date, underscoring the challenges of translating strategic concerns into durable contracts and measurable returns.