Microsoft's boss, Satya Nadella, said something you rarely hear from a man whose company makes billions off artificial intelligence: companies that pay for someone else's AI models are actually paying twice - once with money, and a second time with something far more expensive.

„In effect you pay for intelligence twice - once with money, and again with something even more valuable: your own knowledge, which you have to give up", Nadella warned, referring to businesses that use the closed models of companies like OpenAI and Anthropic.

The logic is simple and uncomfortable. Every interaction with such models - every question, every correction, every piece of feedback - teaches the model's maker how your business works. „The models learn from the exhaust fumes - from the questions people write to them... Every correction distils into institutional knowledge", Nadella says. In other words, while you pay the subscription, your firm is unknowingly handing competitors knowledge they „could never buy".

Here it's worth pausing to ask: who really benefits from the whole AI race? Companies praise the models as a productivity miracle, but rarely mention that every enterprise using them also becomes a free teacher for the model. And once all your data is already inside, getting out is hard - you're locked in with a single supplier.

Nadella's advice is predictable for someone selling his own AI infrastructure: keep full ownership of your data, build „your own learning environments" on your own hardware, insert layers that let you switch between multiple suppliers, and consider open-source models installed locally. Is this an honest warning or just marketing for Microsoft? Probably both. But the point stands no matter who says it - in the world of artificial intelligence, data is currency, and most companies spend it without counting it.