Satya Nadella told businesses on CNN this Sunday that relying on a single AI provider for all their needs is a path to failure.
The Microsoft CEO expanded on a warning from earlier this month during an interview with Fareed Zakaria. He argued that firms which hand over everything from raw data to specific prompts to a model provider will not survive the long term.
Nadella insisted on a setup where the user retains all metadata generated during interactions. This data should be stored locally to train private weights or build independent open models. By weights he means the trained parameters that function as a model’s brain. His point is that companies must hold their own usage data to eventually construct their own models.
Any firm lacking this control will not remain a firm because it has essentially outsourced its thinking, he stated.
Companies without their own models or a layer of AI infrastructure known as gateways to separate prompts from the model itself will face significant trouble, according to Nadella.
He wants enterprises to stop relying on built-in coding tools provided by AI labs. These are harnesses that Anthropic and OpenAI offer through their respective products.
By keeping the harness separate from the model and memory separate from the model, you can use multiple models for what they are best at. At the same time, if any one model disappears, you can still be in control of your own destiny, Nadella said.
Microsoft is an investor in the two largest AI labs, Anthropic and OpenAI. Coding agents are a popular way for enterprises to use AI models and are earning the model makers substantial revenue.
Yet, Nadella is telling enterprises not to rely too heavily on them. Microsoft would benefit from that warning, as its cloud business is now selling the kind of alternative infrastructure he recommends.
Despite the obvious self-serving fear tactic, he is not wrong. Enterprises are increasingly realising that they need many model options, particularly cheaper ones. They are turning to open-weight models whose underlying code is publicly available so they can fine-tune and run them on their own hardware. This means they will also need ways to manage multiple models and coding agents that are not tied to a specific model provider.
Nadella’s observation is not just about runaway budgets. He anticipates that once a company has outsourced its thinking to a model, there is little to stop the AI lab from eventually offering a competing service of its own. This risk grows as enterprises adopt AI agents and give them access to the innards of the company.
It is the kind of warning the startup industry has been shuddering about for years. What stops model makers from wiping out startups by copying and competing with them?
In May, OpenAI CEO Sam Altman offered to invest in every Y Combinator startup in its latest cohort by offering them AI credits. Seed investor Jason Calacanis issued a similar buyer-beware. He posted: If you take these tokens, there’s a non-zero chance that OpenAI will study exactly what your startup is doing, copy your idea and put your app into their free offering. This is the classic platform playbook — be careful, founders! he added.
Now Nadella is making that same case to enterprises.
One caveat: Nadella’s concern about oversharing with AI models applies only to businesses, not individuals. When Zakaria specifically asked Nadella how everyday people could protect themselves, Nadella shrugged it off. He said sharing data is simply the price consumers pay for using a service, especially a free one.
To some degree there’s got to be some value exchange in the consumer space where you’re getting something for free, maybe for your data. That’s sort of how the advertising business model has worked, Nadella said.
What it means
For companies building products, the shift is clear. You cannot treat a third-party AI as a permanent utility. You must build the plumbing to route requests through your own infrastructure. This allows you to swap models without rewriting code and ensures you own the data generated by your work. The cost of inaction is losing your intellectual property to the very tools you depend on.
