Apple is looking to acquire AI chip companies as part of an effort to reduce its reliance on Nvidia for demanding AI workloads, according to The information.
Apple currently handles some AI processing in its own data centers using its own chips, but relies on Nvidia hardware hosted in Google Cloud for more demanding tasks, an arrangement that includes the Gemini model that powers the revised version of Siri. Apple’s AI server chip, internally dubbed Baltra, was supposed to ship this year but has been delayed, according to people familiar with the project.
Apple has historically limited its acquisitions to deals worth hundreds of millions of dollars and avoided large purchases, but that approach appears to be changing. In January, Apple completed the acquisition of Q.ai, an Israeli company specializing in speech interpretation using facial micro-motions, for nearly $2 billion, second only to the $3 billion Apple paid for Beats Electronics in 2014.
Apple also signaled a change in financial strategy during its latest quarterly earnings conference call, when CFO Kevan Parekh told analysts that the company would no longer aim for “net cash neutral” status, a policy under which it had kept its cash reserves roughly in line with its total debt. Apple has not explained the reasoning behind this change, although the added flexibility could free up capital for larger acquisitions.
Apple is already reportedly undertaking acquisitions of AI companies that could help scale down AI models for more effective use on iPhone.
It’s also worth noting that Apple’s in-house chip design capabilities came about through acquisition. In 2008, the company acquired PA Semi for $278 million, a deal that laid the foundation for the custom processors now used in many of its product lines.
Looming leadership transitions could also lead to a more aggressive approach to negotiating, with hardware chief John Ternus set to succeed Tim Cook as CEO in September, and chip chief Johny Srouji given expanded responsibility over all of Apple’s hardware engineering, in addition to semiconductors.
Apple’s chip design team has traditionally focused on battery-powered mobile devices rather than the high-performance server chips needed to compete with Nvidia, the leading supplier of AI server hardware. This limitation became apparent during the development of the revamped Siri, when engineers reportedly attempted to run Google’s Gemini models on Apple’s own server infrastructure, but discovered that the chips, designed for Mac workloads, could not handle a model of this scale. As a result, Apple had to process part of the workload for the new Siri using Nvidia chips within Google’s cloud infrastructure.
Acquiring external chip expertise would complement the work already underway internally. Apple is currently developing a server chip based on the M5 Ultra chip, according to Bloombergwhile a future M7 Ultra chip is said to be intended to significantly improve AI performance to a level that could start to rival Nvidia’s Blackwell chip. The M7 Ultra is expected to support up to 1.5TB of memory, approximately double the capacity of the M5 Ultra. Bloomberg added that an M7 Ultra-based server chip is unlikely to be ready before 2029.
Acquisitions represent just one path Apple is taking to reduce its dependence on Nvidia. The information first reported Apple’s collaboration with Broadcom on an AI server chip in 2024, and Broadcom confirmed in a securities filing last week that the companies had extended that partnership through 2031.
