AI woes prompt Apple to pivot and spend $1 billion on NVIDIA GB300 NVL72 super-accelerators

Loop Capital analyst Ananda Baruah said in a note to investors on Monday that Apple has placed orders for about $1 billion worth of NVIDIA GB300 NVL72 systems. That’s about 250 systems, priced at $3.7 million to $4 million each, the analyst said.

Barua also said that Apple has engaged Dell and Supermicro to work on its own large server cluster to support generative AI applications. The analyst said that there is little information about this project at the moment, but it seems that Apple intends to use the cluster to work with large language models (LLM). The company has previously partnered with Google Cloud and used TPUs to train its AI models. For secure inference, the company uses servers based on its own M-series chips. The company is also rumored to be working on its own server AI accelerators with the support of Broadcom.

Image source: Laurenz Heymann/unsplash.com

Ananda Baruah believes that Apple’s abrupt change in strategy is due to problems in bringing to market its updated AI-powered Siri digital assistant, which was announced last year at WWDC 2024. The company had planned to launch the new Siri features earlier this year, but it looks like it will only happen in 2026.

An Apple executive called the delays and missteps in the launch of Apple Intelligence AI features “disgusting” and “disgraceful” as the company promoted its AI capabilities in TV ads, Bloomberg reported earlier. The company has already been sued over the matter. As a result, Mike Rockwell, head of Vision Pro, was appointed as the new head of AI instead of John Giannandrea.