Microsoft CEO warns: "These companies will not survive"
Satya Nadella has reiterated a sharp warning against relying solely on popular AI tools. We examine the implications of this claim and why it directly affects the investment portfolios of many Israelis.
In an interview with CNN, Microsoft CEO Satya Nadella reiterated and escalated a warning he had voiced previously: companies that rely entirely on large AI providers for all their artificial intelligence needs will not survive in the long run.
"Any company that doesn't have that control, I argue, won't be a company — because you've essentially outsourced your thinking," Nadella said. His point is simple: anyone who allows an external model to think for them is giving up their most valuable asset.
Nadella calls on companies to retain every piece of information that passes through a model — the data, the prompts, and all associated context. The idea is that this information should remain the property of the company, so that one day it can be used to train its own model or an open model that it controls.
He aims specific criticism at the labs' built-in code tools, known in the industry as a "harness." Anthropic's Claude Code and OpenAI's Codex are prominent examples. Instead of relying on them, Nadella recommends separating code tools, memory, and context from the model itself using an infrastructure layer called an "AI gateway." This way, he explains, it is possible to use several models simultaneously based on their specific strengths, and if a certain model disappears, the company remains independent.
It is important to remember that Microsoft is a major investor in both Anthropic and OpenAI, yet Nadella warns customers not to rely too heavily on their tools. The reason: Microsoft's cloud arm is exactly the one that sells the alternative "gateway" infrastructure he recommends. In other words, the more companies adopt his concept, the more money they will pour into Microsoft's infrastructure.
However, this does not mean he is wrong. Many companies already understand that they need a variety of models, especially cheaper alternatives, and many are turning to open models that they can run on their own hardware. Such a trend indeed requires ways to manage several models at once.
Beneath the surface lies a deeper fear: once a company has "outsourced its thinking" to an external model, what prevents that same AI company from launching a competing service? This is a fear that has haunted the startup world for years.
In May, when OpenAI CEO Sam Altman offered credits to companies in the latest Y Combinator batch, investor Jason Calacanis warned: there is a significant chance that the AI company will learn exactly what you are doing, copy the idea, and eliminate you.
Microsoft is one of the largest stocks in the S&P 500 index and, therefore, one of the most prominent holdings in the pension funds and advanced training funds that many Israelis hold without noticing.
The debate Nadella is raising — whether the value of AI will be concentrated in the hands of a few giants or flow broadly to the entire economy — is exactly the debate that will determine whether the current valuation of technology stocks is justified. Nadella paints a future where everyone wins, but it is worth remembering that it is also a future that is very convenient for Microsoft itself.
