No Massive NVIDIA Clusters for Apple?

Apple says its AI models were trained on Google’s custom chips

Apple is using chips designed by Google in building its advanced AI models, according to a paper published on Monday.

CNBCKif Leswing

The fact that Apple was using Google's TPUs to train their AI systems was known right after WWDC in June thanks to another paper that was published around the unveiling of Apple Intelligence. But it went little-remarked upon with everything else being talked about – notably the OpenAI partnership.¹ And now there are seemingly some new details about the TPU usage:

Apple said on Monday that the artificial intelligence models underpinning Apple Intelligence, its AI system, were pre-trained on processors designed by Google, a sign that big tech companies are looking for alternatives to Nvidia when it comes to the training of cutting-edge AI.

Apple’s choice of Google’s homegrown Tensor Processing Unit (TPU) for training was detailed in a technical paper just published by the company. Separately, Apple released a preview version of Apple Intelligence for some devices on Monday.

Nvidia’s pricey graphics processing units (GPUs) dominate the market for high-end AI training chips, and have been in such high demand over the past couple years that they’ve been difficult to procure in the quantities needed. OpenAI, Microsoft, and Anthropic are all using Nvidia’s GPUs for their models, while other tech companies, including Google, Meta, Oracle and Tesla are snapping them up to build out their AI systems and offerings.

The question is if Apple went with Google because their TPUs best suited their needs – or if they simply couldn't get access to the NVIDIA chips required. OR if they didn't want to use NVIDIA chips due to a sort of simmering relationship that goes back decades. Or some combination of the above?

It's also possible – and probably probable? – that Apple is using NVIDIA chips in some capacity. But this paper doesn't indicate that – while neither Google nor NVIDIA are specifically mentioned, there's only one explanation for all the talk of very specific TPUs...

At the same time, there's obviously a reason NVIDIA dominates the market right now and such dominance is highlighted by the fact that a sub-industry has emerged for tech companies to rent out NVIDIA chips to their rivals in some cases. Everyone uses NVIDIA chips – including Google! So either Apple went out of their way not to use their chips or couldn't get access to the chips.

¹ And, of course, Apple's head of AI, John Giannandrea, came from Google...