Apple hunts for AI chip acquisitions as its own AI servers struggle to keep up
What’s the deal? Apple is hunting for AI chip acquisitions, approaching chip startups and talking to bankers about deals — a rare move for a company that almost never buys big. The driver: its own AI servers, running on internally designed M2 Ultra chips, are struggling. The heavy lifting behind the new Gemini-powered Siri runs instead on Nvidia chips inside GoogleDealroom has a profile for this one. Try Dealroom →’s cloud.
The chip that slipped: Apple’s own server chip, codenamed Baltra, was due in 2026 but has slipped — a chip powerful enough to rival Nvidia may not land until 2029, with an M5 Ultra upgrade filling the gap.
A more open wallet: Two signals suggest Apple is ready to spend. Finance chief Kevan ParekhDealroom has a profile for this one. Try Dealroom → told analysts the company would drop its long-held goal of holding as much cash as debt (it sat on $45.6bn at the end of March 2026). And leadership is shifting to engineers — Tim CookDealroom has a profile for this one. Try Dealroom → hands the CEO role to hardware boss John TernusDealroom has a profile for this one. Try Dealroom → in September 2026, while chip chief Johny SroujiDealroom has a profile for this one. Try Dealroom → now oversees all hardware.
For context: Apple’s biggest deal ever remains the $3bn Beats purchase (2014); it built its chip empire off the $278m PA Semi buy (2008). Per the report, Apple paid roughly $2bn for Israeli AI startup Q.ai, its second-largest deal on record. It has also committed to $30bn of Broadcom chips through 2031 and is in talks with model-compression startup PrismML.
The signal: Apple wants off Nvidia and clear of the wider memory and hardware squeeze reshaping the industry. Designing its own silicon is the long game; buying someone else’s may be the shortcut. For a company that guards its independence, needing either is the real story.
Read more:
The Next Web · The Information
