Self-developed chips fail to live up to expectations; Apple turns to acquisitions to break through the AI server dilemma

```

Apple's success with consumer-grade chips has not extended to AI servers—a shortcoming that is now forcing it toward acquisitions.

Apple is seeking to acquire chip companies to make up for its lack of AI server capabilities. According to the latest report from The Information, over the past few months, Apple has held talks with bankers about potential deals and proactively reached out to several semiconductor startups to inquire about their willingness to sell.

The immediate trigger for this move: Apple's self-developed M2 Ultra server chip cannot run Google's Gemini large model, causing some functions of the new version Siri to rely on Google Cloud and run on Nvidia GPUs.

Apple and Nvidia have historically had tense relations. According to previous reports, some Apple executives consider Nvidia "difficult to work with." Yet now Apple has to rely on Nvidia's chips, a passive situation for Apple's management.

M2 Ultra Can't Handle It, Baltra Delayed Again

Apple's AI servers currently run on the M2 Ultra chip, which was originally designed for Macs and is inadequate when faced with ultra-large-scale models like Gemini.

Sources told the media that earlier this year Apple tried to deploy Google's Gemini model on its own server infrastructure to drive the upgraded Siri, but "these chips simply can't handle such a huge model." Ultimately, Apple was forced to shift the relevant workloads to Nvidia chips in Google Cloud.

The root cause is that Apple's chip team has long focused on mobile power efficiency optimization, while the core requirements for AI server chips are completely different—high power consumption, high concurrency, large memory bandwidth. These are two distinct technical approaches.

Apple originally hoped its next-generation AI server chip, codenamed "Baltra," would solve this problem, with a planned release this year, but according to insiders, the project has been delayed.

Won't Catch Up Until 2029, How Will the Gap Be Filled?

According to Bloomberg, Apple is currently pushing for a server upgrade based on M5 Ultra as a transitional solution. The more competitive M7 Ultra is highly anticipated—it will have a maximum memory capacity of up to 1.5TB, about twice that of M5 Ultra, with hopes that its performance will rival Nvidia's Blackwell chip.

But the server version of M7 Ultra won't be ready until 2029.

This means Apple faces at least three years of a technology gap. The AI competition won't wait during this period. Acquiring a startup with server chip technology is the most direct way to fill this gap.

Apple is also seeking other paths. The Information previously reported that Apple has partnered with Broadcom in 2024 to develop AI server chips. Last week, Broadcom disclosed in securities filings that the two sides are "extending their long-term technical cooperation to 2031," but did not specify the details of the collaboration.

Acquisition Door Opening

Apple has always been restrained in large acquisitions. Its largest deal in history was the $3 billion acquisition of Beats Electronics in 2014.

But this year, this convention has loosened.

In January, Apple acquired Israeli startup Q.ai for nearly $2 billion—the second largest acquisition in Apple's history. Q.ai's technology can interpret voice content through facial micro-expressions.

There are signals in financial policy as well. On the Q2 earnings call, CFO Kevan Parekh announced that Apple would abandon its long-standing "net cash neutral" policy—i.e., keeping cash reserves balanced with total debt. Apple didn't explain the reason for the policy change, but The Information suggests one possibility is to use some cash for major acquisitions.

Apple's path to self-developed chips actually began with an acquisition. In 2008, Apple bought PA Semi, a startup, for $278 million, laying the foundation for over a decade of chip self-development.

Leadership Changes May Bring a More Aggressive Acquisition Style

Apple is in a CEO transition period. This September, hardware chief John Ternus will succeed Tim Cook as CEO. Meanwhile, chip head Johny Srouji will be promoted to oversee all hardware engineering, with semiconductor business still under his purview.

The arrival of the new leadership may bring a more active acquisition orientation.

Apple's daily acquisition work is managed by Steve Smith, reporting to Vice President of Corporate Development Adrian Perica. Besides server chips, Apple is also seeking to acquire AI companies capable of compressing model size to better suit iPhone-side operation, The Information previously reported.

Risk Warning and DisclaimerThe market has risks, investment needs caution. This article does not constitute personal investment advice, nor does it consider individual users' special investment goals, financial situation, or needs. Users should decide whether any opinions, viewpoints, or conclusions in this article fit their particular circumstances. Investing accordingly is at your own risk. ```