Apple Seeks to Acquire AI Chip Companies to Address Server Shortcomings
0xBroomberg
Apple is actively approaching semiconductor startups and investment banks to acquire its way out of an AI server-chip gap — its M2 Ultra cannot handle large-model workloads, forcing parts of Siri onto Nvidia chips in Google's cloud, a strategic embarrassment for a company that has long avoided Nvidia dependence.
Where exactly does Apple's AI server chip fall short?
Apple tried deploying Google's Gemini model on its own servers, but the M2 Ultra couldn't keep up — it was designed for the Mac, not for large-model inference.
The result: parts of the new Siri now run on Nvidia chips inside Google Cloud. This means → Apple depends on the very supplier it least wants to depend on for its most important AI capability.
In plain terms = Apple's own silicon can't power Apple's own AI, so it rents a rival's compute. That is the starting point for everything that follows.
How is Apple trying to fix this?
According to *The Information*, Apple has contacted multiple investment banks to discuss potential deals and proactively approached semiconductor startups to gauge willingness to sell. Apple declined to comment.
Apple is also looking for AI companies that can help compress AI models to run better on-device on iPhones. This means → the acquisition search runs on two tracks: stronger chips for the cloud, smaller models for the edge.
The M&A strategy is led by corporate-development VP Adrian Perica, with Steve Smith handling day-to-day execution.
Has Apple's spending appetite actually changed?
Apple has historically favored sub-billion-dollar tuck-in deals. But in January it paid close to $2 billion for Israeli startup Q.ai — its second-largest acquisition ever, behind only the $3 billion Beats deal in 2014.
The bigger signal: CFO Kevan Parekh said on the Q2 earnings call that Apple is gradually abandoning its long-standing "net-cash-neutral" policy — keeping cash reserves roughly equal to total debt. This means → Apple's balance sheet is clearing room for large-scale M&A.
Apple's custom-chip story itself began with an acquisition: the $278 million purchase of PA Semi in 2008, which seeded every generation of iPhone processors since. This reflects a proven playbook — when Apple faces a critical capability gap, it buys a company to build the capability.
Where does the in-house chip roadmap stand?
The next-generation AI server chip, codenamed "Baltra," was planned for this year but has been delayed.
Per Bloomberg, Apple is currently upgrading servers around the M5 Ultra; the longer-term M7 Ultra aims to sharply boost AI performance and could compete with Nvidia's Blackwell chips.
In plain terms = the near-term plan is to hold the line with M5, the long-term bet is M7 catching Nvidia — but the near-term plan is already slipping, which is exactly why Apple is hunting for acquisitions now.
How will the leadership transition shape this?
In September, hardware chief John Ternus will succeed Tim Cook as CEO; chip chief Johny Srouji will be promoted to oversee all hardware engineering.
This means → the ultimate decision-maker on chips is about to change. Whether the new leadership pursues M&A more aggressively will directly determine how fast Apple can break free from Nvidia dependence.
This is the story's core inflection point: Apple's AI chip problem is a technology question (the in-house roadmap), a tempo question (acquisition acceleration), and a leadership question (the new team's appetite for bold moves). Whether all three lines converge will become clearer after September.
Content is for reference only, not financial advice.