Apple on the Hunt for AI Chip Companies: Its M2 Ultra Servers Are Not Enough

Apple has managed to build a strong competitive advantage with the chips it designs in-house, from iPhones to Macs. However, that advantage seems to have stalled when it comes to the servers that need to run its artificial intelligence, to the point that the company is contemplating a large-scale operation: acquiring external companies. According to The Information, which first reported the scoop, Apple has initiated contacts with semiconductor startups asking if they would be willing to be acquired.

The reason relates to the data centers that Apple uses for some of its AI tasks, which currently run on chips designed by Apple, the M2 Ultra, and are showing limitations. The most demanding work, including the Gemini model behind the new Siri, relies on Nvidia chips hosted in Google's cloud. Apple has attempted to shift that load onto its own machines, but the infrastructure proved insufficient.

The Delayed Chip

There is a server chip designed by the company in the works: codenamed Baltra, expected this year. But according to Bloomberg, a processor powerful enough to compete with Nvidia might not arrive until 2029, with an upgrade to M5 Ultra chips serving as a temporary measure in the meantime.

On such a horizon, waiting for the right silicon means remaining dependent on third-party hardware for years, precisely at a time when AI is the company's declared priority. Hence the consideration to shorten the timeline by purchasing pre-existing expertise on the server side, where Apple's experience, built on consumer devices, is thinner.

A More Open Portfolio

Making large purchases would be a break from Cupertino's habits. The most expensive operation ever remains the purchase of Beats in 2014 for three billion dollars; the entire chip empire, on the other hand, originated from a much smaller expenditure of 278 million dollars paid for PA Semi in 2008. Even this year's purchases remain measured: about two billion for the AI startup Q.ai, the second-largest operation in the company's history.

Two elements suggest that this time Cupertino is ready to spend more. CFO Kevan Parekh told analysts that the company will abandon the long-held goal of maintaining cash equal to its debt: a choice that frees up resources. And there is no lack of cash, as Apple had a total liquidity of 45.6 billion dollars at the end of March.

Acquisitions, however, remain just one of the avenues. Apple is also in negotiations with PrismML, a startup that compresses large AI models to run them directly on iPhones, and has just confirmed a commitment of 30 billion dollars for Broadcom chips, with collaboration extended until 2031. The direction is the same: to break free from Nvidia and the stranglehold on memories and hardware that is reshaping the industry.