Microsoft CEO Satya Nadella has taken his previous warning to businesses using artificial intelligence technologies a step further. Nadella predicts that companies relying entirely on private AI laboratories for all their AI needs will not survive in the long run.

Speaking on CNN’s Fareed Zakaria GPS program, Nadella emphasized that businesses should be cautious about data and command sharing. He stated that it is critical for companies to keep their usage information and metadata under their own control so that they can train their own models.

Strategic Control and Independence

Nadella says that companies should have their own models or infrastructures called AI gateways. This approach allows companies to create an independent structure by separating their commands from the model itself.

Microsoft CEO emphasizes that companies should avoid excessive reliance on tools like Anthropic’s Claude Code or OpenAI’s ChatGPT Codex. He argues that companies that don’t build their own infrastructure are outsourcing their thinking processes, jeopardizing the future of those companies.

It’s noteworthy that Nadella is making this warning despite Microsoft’s investment in large labs like Anthropic and OpenAI. Companies are tending to work on their own hardware, moving towards open-source models to reduce costs and provide greater flexibility.

Competition Risk and Data Security

Another concern of Nadella is the possibility that AI labs could use company information to develop their own competing services. This risk is said to increase as businesses allow AI spies access to critical internal information.

Previously, investor Jason Calacanis issued a similar warning, stating that companies providing excessive information to platforms like OpenAI could lead to the copying of their ideas. Nadella argues that this platform game is now also valid at the enterprise level.

Nadella points out that this data sharing concern only applies to businesses and that the situation is different for individual users. He says that individuals sharing their data in exchange for free services is a component of the advertising business model.

Do you think that companies maintaining their independence in their AI strategies is really that critical for their long-term success?