In order to reduce its dependence on NVIDIA hardware when handling high-intensity artificial intelligence (AI) tasks, Apple is currently actively seeking to acquire related AI chip companies. Although Apple currently uses its own chips in its own data centers to handle some AI tasks, when facing larger and more complex AI workloads, it still needs to rely on Nvidia hardware deployed in Google Cloud.
For example, the Gemini model that powers the new and upgraded version of Siri uses this approach. According to people familiar with the matter, Apple originally planned to launch a self-developed AI server chip code-named "Baltra" this year, but the project has been delayed.
Historically, Apple has generally tended to make small acquisitions worth hundreds of millions of dollars, rarely engaging in huge mergers and acquisitions. However, this strategy appears to be shifting significantly. In January this year, Apple completed the acquisition of Israeli company Q.ai, which specializes in analyzing speech through small facial movements, for nearly $2 billion. This is also the second-largest acquisition in Apple’s history after spending $3 billion to acquire Beats Electronics in 2014.
At the same time, Apple Chief Financial Officer Kevin Parekh also released a signal of financial strategy adjustments in the latest quarterly earnings call. He told analysts that the company would no longer pursue a goal of being "net cash neutral," where cash reserves are roughly equal to total debt. Although Apple did not clearly state the reason for this change, it will undoubtedly free up sufficient capital space for larger acquisitions. In addition, there are reports that Apple is also actively looking for AI startups that can help compress AI models to run efficiently on the iPhone.
Looking back at the development history of Apple's self-developed chips, its core capabilities originated from a key acquisition. In 2008, Apple acquired PA Semi for US$278 million. This transaction laid the foundation for the custom processors that are now widely used in its various hardware products. The upcoming personnel changes may also push Apple to adopt a more aggressive strategy in the M&A market. It is reported that John Ternus, the head of the hardware business, will succeed Tim Cook as CEO in September this year, while chip business executive Johnny Sluigi will expand his scope of authority to take charge of all Apple's hardware engineering in addition to being responsible for the semiconductor business.
Apple's chip design team has long focused on battery-powered mobile devices rather than high-performance server chips capable of competing with AI server hardware giant Nvidia. This limitation was exposed during the development of the new version of Siri: Engineers tried running Google's Gemini model on Apple's own server infrastructure, only to find that chips designed for Mac workloads simply could not handle such a large model. Therefore, Apple had to use Nvidia chips in Google's cloud infrastructure to handle some of the work of the new Siri.
The acquisition of an external chip technology team will undoubtedly complement Apple's ongoing internal projects. According to Bloomberg, Apple is currently developing a server chip based on M5 Ultra, and the future M7 Ultra chip is said to significantly improve AI performance and is even expected to compete with Nvidia's Blackwell chip. The M7 Ultra is expected to support up to 1.5TB of memory, about twice as much as the M5 Ultra, but Bloomberg also added that server chips based on the M7 Ultra may not come out until after 2029.
In addition to seeking mergers and acquisitions, Apple is also pursuing hardware independence through other means. Previous reports claimed that Apple had begun cooperating with Broadcom on AI server chips as early as 2024. Just last week, Broadcom confirmed in a securities filing that the two parties had officially extended the partnership until 2031.