Apple is hunting AI chip startups as its in-house silicon falls short

Apple is actively exploring acquisitions of AI chip startups, according to The Information. The company has already reached out to chip firms and hired financial bankers to explore deals — a striking shift for a company that has spent 20 years building almost everything in-house.

The gap in the data center

The problem is Apple's server chips. Its current data center hardware runs on M2 Ultra processors, which are capable enough for lighter AI tasks — but not for the heaviest inference workloads. For those, Apple still leans on Google Cloud's NVIDIA GPUs, including for parts of the Gemini-powered Siri backend. Internal efforts to move those demanding jobs onto Apple's own servers haven't worked well enough.

The company's next-generation M7 Ultra server chip won't arrive until 2029, per MacRumors. In the meantime, Apple plans to upgrade its data centers with M5 Ultra chips. A new internal server chip codenamed Baltra — originally planned for a 2026 launch — has also been delayed. That leaves a roughly three-year window where Apple's own silicon won't be competitive with NVIDIA in the data center.

The acquisition pivot

Apple has the cash and, now, the appetite. The company held $45.6 billion in cash and equivalents at the end of March 2026. More telling, CFO Kevan Parekh quietly dropped Apple's long-standing "net cash neutral" policy this year, freeing the company to pursue larger deals without the old constraint of returning surplus cash to shareholders.

The $2 billion acquisition of AI startup Q.ai earlier in 2026 — Apple's second-largest deal ever after the $3 billion Beats purchase in 2014 — signals that big checks are back on the table. PrismML, a startup working on model compression technology, has been named as one potential target, though no deal has been confirmed.

Apple also signed a $30 billion agreement with Broadcom to source chips made in the US, showing it is willing to work outside its own labs when the performance gap demands it.

What it means

For iPhone and Mac users, none of this changes anything immediately. But Apple Intelligence features — and Siri's ability to handle complex requests — depend on server-side AI that Apple currently can't run efficiently on its own hardware. An acquisition could close that gap faster than waiting for Baltra or the M7 Ultra. Until then, Apple's AI ambitions are partly built on someone else's chips.