Behind this strategic shift is the real pressure faced by Apple's AI chip layout. Previously, Apple's self-developed next-generation AI server chip Baltra encountered delays, while the existing M2 Ultra has hit a performance bottleneck in data center AI tasks, forcing the company to outsource some heavy workload work to Nvidia infrastructure on the Google Cloud platform.

This means that in the context of AI models iterating rapidly on a weekly basis, Apple's core AI infrastructure still has a dual dependence on external suppliers.

Apple's pace in mergers and acquisitions is quietly accelerating. In January of this year, Apple spent nearly $2 billion to acquire Israeli AI startup Q.ai, becoming the second largest merger in the company's history, second only to the $3 billion acquisition of Beats Electronics in 2014.

Q. AI focuses on the technology of interpreting speech through facial micro movements, and this transaction marks Apple's departure from its previous restrained style in strategic acquisitions.

Meanwhile, Apple has recently held talks with PrismML, a startup specializing in model compression technology. The company claims to be able to compress a 27 billion parameter large model to run locally on the iPhone without sacrificing performance, which is highly in line with Apple's device side AI vision.

It is worth noting that although the acquisition intention is clear, the competition in the AI server chip field is highly concentrated, with Nvidia occupying a dominant position. If Apple wants to substantially enhance its AI computing power through acquisitions, the potential target's technological maturity and strategic alignment face a high threshold.