‘Paying for intelligence twice’: Satya Nadella warns firms over hidden cost of AI adoption

Microsoft CEO Satya Nadella has cautioned enterprises against blindly embracing AI, warning that in the race to adopt the technology, many businesses risk exposing their proprietary knowledge that could erode their long-term competitive edge.

In a short essay published on X that focuses on how firms should protect their core intellectual property (IP) in the age of intelligence, Nadella put forward the idea of the ‘Reverse Information Paradox’. He said that businesses are not just paying money to use AI tools but also sharing their internal expertise, workflows, and decision-making processes after buying the tools.

“The better you want the model to perform, the more of that knowledge you have to feed it! That is what I think of as the Reverse Information Paradox,” Nadella wrote in the post on Sunday, July 13.

Nadella’s remarks come at a time when business customers are increasingly scrutinising AI spending. Prices of tokens – the units used to measure AI usage – are falling, but the cost of completing a task is rising as ​AI firms shift from flat subscriptions to usage-based pricing.

This has, in turn, triggered a reassessment within companies that until recently encouraged heavy use of AI tools, often treating rising consumption as a proxy for productivity, dubbed “tokenmaxxing”. Now, those bills are starting to bite.

However, top executives such as Nadella, Palo Alto Networks’ Nikesh ‌Arora, and Coinbase Global’s Brian Armstrong have argued that smaller, cheaper AI models can handle a big share of corporate needs. In a similar post last month, Nadella warned against a future in which a handful of AI models “eat everything they see” and capture all the economic returns, with entire industries left to watch their expertise commoditised out from under them.

What is the Reverse Information Paradox?

Nadella’s latest theory is based on a traditional economic theory called the ‘Information Paradox’ that was introduced by Nobel Prize-winning economist Kenneth Arrow. This theory proposes that a paradox exists in the market of information, where the value of the information is not known to the buyer until he has that information and the seller risks giving away the knowledge in order to sell it.

Story continues below this ad

In the AI era, Nadella hypothesised, the problem is reversed. “In the AI age, the buyer risks giving away knowledge, just in order to use what they bought,” he wrote.

https://t.co/xv6csf1SbV

— Satya Nadella (@satyanadella) July 12, 2026

“You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!” he said.

“Over time, the information asymmetry becomes increasingly skewed. The seller learns more and more about you as you use what you purchased, while you learn very little about what the seller is learning in return,” he added.

Every prompt, correction, evaluation and piece of feedback given to an AI system helps improve it, Nadella said. “Every correction is distilled into institutional know-how. It’s the kind of knowledge a competitor could never buy, and the kind that leaks almost imperceptibly: trace by trace, correction by correction, eval by eval,” he further said.

Why is Nadella against restrictions on distillation?

The Microsoft chief also criticised AI model providers for restricting other players from distillation – a common training method that AI labs use on their own models to create smaller, cheaper versions of it. However, competitors can also use distillation to essentially copy the homework of other labs.

Story continues below this ad

Earlier this year, Anthropic accused three Chinese companies – DeepSeek, Moonshot AI, and MiniMax – of setting up more than 24,000 fake accounts with its Claude AI model to improve their own models through distillation.

“While the great innovation that comes from model providers having fair use rights to train models on public data is needed. I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation, and to reserve the right to learn from customer usage and interaction data,” Nadella said.

“If learning flows in only one direction, economic value converges toward the owners of the learning infrastructure rather than the creators of the knowledge itself. Therefore, it’s imperative that we distribute the learning infrastructure to every firm so that they can control their own learning loop,” he added.

How can enterprises safeguard themselves?

According to Nadella, there needs to be an equivalent of patents, which allow inventors to share ideas without losing ownership of them, for the AI era.

Story continues below this ad

“In the cloud era, enterprises accumulated data. In the AI era, they accumulate learning. The trust boundary must evolve accordingly, from protecting information to protecting the mechanisms through which organizations learn, adapt, and compound intelligence,” he said.

Nadella further outlined a few things every enterprise must do to ensure this:

-Control: Create your private evals, because evals define what “good” looks like inside the organisation. Also, retain ownership of your organisation’s memory, traces, feedbacks, decisions, and institutional context, and ability to use outputs of models from your own tasks and queries.

-Capability: Build your own proprietary learning environments within the tenant boundary to train or tune models, where models learn against real workflows without exposing the company’s knowledge.

Story continues below this ad

-Choice: Ensure the orchestration layer is decoupled from any single model. Ask yourself: If any one model you are using is taken away, do you still have the ability to operate and optimize for your evals using other models? Does your company “veteran” capability remain with you even if a given “generalist” model is taken away?

-Cost: By decoupling the orchestration layer, you are also able to bring together context, models, and tasks in the most efficient and cost-effective way without sacrificing quality.

-Compound: Bring these four together and you create your own continuous learning loop (i.e. hill climbing machine) that will allow your AI investments to compound the value of your firm.