Satya Nadella says companies that trust one AI with everything may not survive

On Sunday, Microsoft CEO Satya Nadella doubled down on the shock warning he issued earlier this month to companies using artificial intelligence, going one step further this time. He predicts that companies that rely entirely on private AI labs to meet their AI needs will ultimately not be able to survive.

This is what he said on the “Fareed Zakaria GPS” program on CNN. When Zakaria asked Nadella to explain why a company shares so much with an AI model provider, Nadella said companies need to be careful of everything they provide, from their data to their claims.

Nadella called for setting up “every time you use the model, you keep all the metadata around it, so you can use all of that to train your own weights or your open model.” (The weights are the trained parameters of the model—essentially its brain. Nadella’s point: Companies should hold on to their own usage data until they can eventually build a model of their own.)

“Any company that doesn’t have that control, I would argue, will not remain a company because you have essentially outsourced your thinking,” he added.

In short: Companies that don’t have their own models — or that don’t have a layer of AI infrastructure known as AI gateways to separate their claims from the model itself — are going to be in trouble, Nadella says.

Specifically, he wants companies to stop relying on encryption tools built into AI labs, known as tools. (Examples include Claude Code from Anthropic and ChatGPT Codex from OpenAI.)

“By keeping the belt separate from the model and the context and memory separate from the model, you can certainly use multiple models for what they’re good at,” Nadella said. “At the same time, any one model can disappear, and you can still continue to control your destiny.”

Considering that Microsoft is an investor in the two largest AI labs, Anthropic and OpenAI. Programming agents are a particularly popular way for organizations to use AI models, and by all accounts model makers make a lot of money.

However, Nadella calls on companies not to rely on it too much. Microsoft would naturally take advantage of this warning, as its cloud business now also sells the kind of alternative infrastructure it recommends.

Despite the obvious self-serving fear tactic, he is not wrong. Increasingly, companies are realizing that they need many model options, especially the cheapest options, and are turning to open-weight models – models whose underlying code is publicly available – that they can configure and run on their own hardware. This in turn means that they will also need ways to manage multiple forms, as well as coding agents that are not tied to a specific form provider.

But Nadella’s observation is not just about runaway budgets. He predicts that once a company “outsources its thinking” for a model, there will be nothing stopping the AI ​​lab from eventually offering its own competing service. This risk is increasing as companies adopt AI agents and give them access to the internal parts of the company.

It’s the kind of warning the startup industry has been trembling for years: What’s to stop model makers from killing off startups by imitating and competing with them?

In May, for example, when OpenAI CEO Sam Altman offered to invest in every Y Combinator startup in its latest batch by giving them AI credits, seed investor Jason Calacanis issued a similar buyer’s warning, posting: “If you take these tokens, there’s a non-zero chance that OpenAI will look at exactly what your startup is doing, copy your idea, and put your app in their free offerings. This is the platform’s classic playbook — founders, be careful!” to publish.

Now Nadella is making the same argument to companies.

One caveat: Nadella’s concern about oversharing AI models only applies to companies, not individuals. When Zaccaria Nadella specifically asked how ordinary people can protect themselves, Nadella shrugged it off, saying data sharing is simply the price consumers pay for using a service, especially a free service.

“To some extent, there has to be an exchange of value in the consumer space, where you get something for free, perhaps in exchange for your data. That’s kind of how the advertising business model works,” Nadella said.