Apple's CFO allegedly blocked the procurement of GPUs

Apple's AI models are not considered to be as leading as those of OpenAI and co. A media report now provides a suspicion as to why this might be the case: Apparently, Apple's CFO Luca Maestri prevented a major GPU order around two years ago. These chips are essential for the rapid training of large AI models.

A New York Times report provides insight into a decision-making process that may have cost Apple a lot of rapid progress on its Apple Intelligence AI models, citing five people familiar with the process. In early 2023, John Giannandrea, head of Apple's AI department, is said to have asked Apple CEO Tim Cook to purchase additional GPUs to drive AI development forward. At that time, Apple had around 50,000 GPUs, all around five years old. No comparison to other leading AI companies such as Microsoft or OpenAI, which are in a much better position here.

Maestri threw a spanner in the works

Cook is said to have approved doubling the chip budget of Giannandrea's department. However, CFO Maestri then put a spanner in the works: he halved the budget increase and urged the department to optimize the efficiency of its existing GPUs.

The lack of in-house GPUs inevitably led to the Apple people having to procure additional capacity from external data centers. And because the Nvidia GPUs were in such high demand, Apple also had to switch to GPUs from Google to train its AIs. Overall, Apple was apparently left with significantly less performance than the competition.

It was the start of a whole series of problems in Apple's AI department. There was also trouble between the individual teams – and a battle over who was responsible for the Siri project. This in turn led to Apple having to postpone some of the Siri innovations that the manufacturer had already announced. This in turn brought the tech giant several consumer lawsuits in US courts. According to the New York Times report, the features will now be delivered in the fall of 2025.

(nen)