Apple's generative AI investment could exceed $5 billion in two years

Apple’s investment in generative AI could top $5 billion over the next two years, according to a new report.

The company’s spending plans are primarily focused on purchasing servers, the report says. Apple plans to spend around $1 billion per year on AI, but one analyst thinks that figure could be much higher.

TF Securities analyst Ming-Chi Kuo reckons Apple will buy between 2,000 and 3,000 AI servers this year and up to 20,000 next year. That would represent 1.3% and 5% of global AI server shipments in those years, he reckons.

The bulk of those servers are Nvidia HGX H100 8-GPU servers, which cost around $250,000 a piece. Kuo estimates that would lead to spending of around $620 million this year and $4.75 billion next year. That’d put Apple’s total spending on servers over the next two years at around $5.37 billion.

The lower spending this year is due to a shortage of Nvidia AI chips and delayed orders, Kuo says.

While Apple’s server purchases are substantial, Kuo says they’re still behind the likes of Meta and Microsoft. The former plans to buy 40,000 servers in 2024, while the latter plans to purchase between 80,000 and 100,000 units.

Kuo believes Apple prefers to train its AI models on its own servers for enhanced security and privacy, rather than relying on cloud services.

He warns that if Apple does keep its AI spending to $1 billion per year, it may struggle to compete effectively in the generative AI space.

Apple could also develop its own AI server chips to save money, but Kuo says there’s currently no evidence that’s happening.

Apple is reportedly working on an upgraded version of Siri and aims to integrate AI across its applications.