Marcus Mendes, reporting for 9to5Mac:

According to The Information, Apple has been exploring potential acquisitions of semiconductor startups to “boost its efforts to build server chips for running AI.”

From the report:

In recent months, the iPhone maker has talked with bankers about possible deals. It has also approached semiconductor startups to gauge their interest in selling themselves, the people said. Apple’s hunt for chip acquisitions comes as the company struggles with the performance of its own internal AI servers, which currently run on internally designed M2 Ultra chips.

The Information says Apple’s next-generation server chip, code-named Baltra, has slipped past its planned 2026 debut. In the meantime, Apple is using M2 Ultra-based systems for some of its own AI processing, while the more demanding tasks are handled by the Gemini-based model powering the new Siri, which runs on Nvidia GPUs in Google Cloud.

A big acquisition would be a departure from Apple’s usual M.O. of buying smaller startups from time to time, but it wouldn’t be unprecedented.

It would indeed be unusual for Apple to purchase a Silicon Valley start-up specializing in artificial intelligence processors, when Apple arguably makes the best consumer AI chips. As Bloomberg has reported extensively, Apple’s consumer processor strategy — for high-end Mac models, at least — has shifted considerably toward AI. People like buying Mac Studios and Mac minis to do local AI inference — they’ve become popular as cloud computers and mini data centers. Per The Information’s reporting, Apple’s internal processors are also on that trajectory, albeit delayed for unclear reasons. Apple’s processors aren’t just fast and capable of AI inference, but they’re extremely efficient, which makes them relatively inexpensive to use in the long run. It’s no wonder why Apple has retaken the No. 1 market capitalization, surpassing Nvidia on Friday.

Regardless, this news is quite interesting because it proves Apple is intent on shipping more powerful, capable models in the future — perhaps ones that can write code or perform agentic tasks. It wants people to be able to run those models on their own devices — hence why the company is testing Mac Studios with 1.5 terabytes of memory and skipping high-end M6-generation chips entirely — and to eliminate its reliance on Google Cloud, or maybe even Google altogether. If Apple were to buy such a company, it could probably pre-train its own models: a lack of compute was one of the key limitations that led to the Google partnership. Apple’s new leadership doesn’t see AI as a fleeting trend. It is finally taking it seriously, something we couldn’t say about the AI team under John Giannandrea, the machine learning chief who was stripped of his responsibilities and replaced by Mike Rockwell, the executive previously in charge of Apple Vision Pro.

I think Apple should lean into efficient, low-cost AI inference on both the hardware and software fronts. Siri AI clearly emphasizes on-device processing, and the company’s future Apple silicon plans do, too. Readers of this blog know that my single biggest qualm with the AI industry for now is that its economics aren’t rooted in logic. Anthropic is compute-constrained, so much so that they’re under heavy competition from Chinese AI labs that can offer inexpensive tokens and relatively the same performance as Anthropic’s best models. OpenAI is deathly unprofitable, leading some pundits to believe that its economics will spell its demise. Compute is the biggest story of Silicon Valley this summer, whether it’s skyrocketing memory prices for consumer products or unusual deals like Meta selling compute to Anthropic.

Whatever happens here — if Apple chooses to buy a compute start-up or not — the fact that it is even entertaining the idea is proof that it sees Apple silicon as its golden ticket in the AI space. As open-weight models get smaller and more powerful, people will buy more powerful Macs marketed as the best for AI inference; as the frontier AI labs become more compute constrained — and their flawed economics catch up to them — people will look to Apple’s models and products. Siri AI is merely the beginning of the Apple AI story.