Apple could purchase as many as 2,000 and 3,000 servers this year, and up to 20,000 in 2024. Kuo expects Apple to purchase servers equipped with Nvidia’s HGX H100 8-GPU for generative AI training, upgrading to B100 next year. H100 is Nvidia’s AI supercomputing platform, and each one is priced at around $250,000, meaning Apple will likely spend at least $620 million in 2023 and $4.75 billion in 2024 on AI server purchases.
Kuo expects Apple to purchase and install its own servers to train large language models rather than virtual hosting from other cloud service providers, providing improved security and privacy. While Kuo says Apple could develop its own server chips, he has not yet seen any evidence that Apple is doing so.
While $5 billion is a healthy investment into AI, Apple will be spending less than other companies, such as Microsoft and Meta. Apple will need to spend several billion dollars in labor costs, infrastructure, and more, to have a chance of catching up with competitors.
Over the weekend, Bloomberg’s Mark Gurman said that Apple is on track to spend $1 billion per year on its AI efforts. Gurman says that Apple is working on a smarter version of Siri, as part of an effort to integrate AI into Apple’s apps.