Apple hunts chip startups to escape reliance on Google Cloud for AI

Apple is quietly pursuing semiconductor acquisitions to bolster its AI server capabilities, according to a report from The Information. The company has been in active talks with investment bankers and semiconductor startups over the past few months to evaluate potential buyouts. The motivation is straightforward: Apple has run into a performance wall.

Today, Apple powers some of its AI data center work using in-house M2 Ultra chips, but those processors have hit a performance ceiling. When Apple needs to handle heavy-lifting workloads—such as running a version of Google's Gemini model to support the next-generation Siri—it has been forced to outsource to Google Cloud's Nvidia-powered infrastructure. Compounding the problem, Apple's own next-generation AI server chip, code-named Baltra, was expected to launch this year but has been delayed.

Apple's historical acquisition strategy favored smaller, hundred-million-dollar deals. That approach changed sharply in January when Apple spent nearly $2 billion（約3200億円） on Q.ai, an Israeli startup developing technology that interprets speech through facial micro-movements. The deal became Apple's second-largest acquisition in history, surpassed only by the $3 billion（約4800億円） purchase of Beats Electronics in 2014. That willingness to open its wallet suggests Apple is now serious about plugging AI infrastructure gaps—and fast.

The precedent is instructive. Apple built its modern chip empire on the back of the PA Semi acquisition in 2008 for $278 million（約440億円）. That smaller, focused buy created the foundation for the iPhone processor that would dominate consumer devices for nearly two decades. Now, nearly twenty years later, Apple appears ready to apply that same playbook to AI server chips, but at a much larger financial scale. If Apple can acquire the right semiconductor talent and technology, it could sharply reduce its dependence on Google Cloud and Nvidia, reclaiming control over a critical piece of its AI infrastructure.

Apple's shift toward semiconductor acquisitions marks a significant strategic pivot. Historically, the company preferred smaller, hundred-million-dollar buys and built much of its competitive advantage through internal chip design—epitomized by the 2008 PA Semi acquisition for $278 million（約440億円） that eventually underpinned its entire iPhone processor line. That model worked for consumer devices, but the AI era is forcing a different calculus. Apple's admission that it must outsource heavy AI inference tasks to Google Cloud reveals an uncomfortable gap: its proprietary M2 Ultra chips cannot match the scale or performance demanded by modern large language models powering services like Siri.

The January acquisition of Q.ai for nearly $2 billion（約3200億円） demonstrates that Apple is now willing to spend at a magnitude comparable to its 2014 Beats deal ($3 billion（約4800億円）). This is not merely a talent grab—it signals that Apple views semiconductor capability in AI infrastructure as existential competitive terrain. The reported delay of Baltra, its next-generation server chip, only sharpens the urgency. Rather than wait for internal R&D, Apple is shopping for proven teams and ready-made technology. If successful, such acquisitions could reduce Apple's dependence on Google's cloud and Nvidia's hardware, shifting the cost and control of AI inference back in-house.