The topic that attracted a lot of attention from analysts during Apple's recent quarterly earnings call was Apple Intelligence, a product that has yet to be officially released.

During the Q&A session, CEO Tim Cook and CFO Luca Maestri largely avoided questions about how quickly Apple Intelligence is being rolled out, whether the company is seeing increased sales from the service, and how OpenAI’s ChatGPT integration into its software is working.

One question the Apple chief was willing to address, though, was the cost of AI servers. Investors are always curious about how far companies are building AI infrastructure and what lies ahead.

ez7z3an8.png
Apple Intelligence is said to trigger the new iPhone upgrade cycle. Photo: Laptopmag

Cook acknowledged that costs are on the rise. The amount spent on AI and Apple Intelligence has been increasing every year.

Apple spent $2.15 billion on property, plant, and equipment in the second quarter, up 8% from the previous quarter and about 3% from a year ago. Some of the capital expenditures were not for AI but for other activities.

Still, Apple’s capital expenditure increase pales in comparison to other tech giants like Microsoft, Google and Meta, which are spending huge sums to build and equip data centers with Nvidia chips.

For example, Microsoft reported capital expenditures of $13.87 billion in the same period, up 55% from 2023. Alphabet’s – Google’s parent company – expenses increased 91% to $13.19 billion, while Meta’s increased 31% to $8.3 billion.

Meta CEO Mark Zuckerberg explained this: The risk of missing out on the generative AI boom is greater than spending too much on graphics processors or servers. Zuckerberg also wants to make sure Apple doesn't have complete control over the next wave of technology if it's AI.

Apple's different "playstyle"

Unlike Amazon, Google, and Microsoft, Apple doesn’t have a cloud business, leasing its infrastructure to other companies. Meta does too, but it invests in training its own large language model and uses AI to power its content recommendation engine.

Last week, Apple revealed in a technical report that it was using Google TPUs—which are cheaper than Nvidia chips—to train its Apple Intelligence models. The company also just released a beta version of Apple Intelligence to developers.

Speaking of infrastructure, Apple has the advantage of designing its own chips for phones and servers, so the iPhone maker doesn't have to spend billions of dollars on third-party chips.

Apple’s “complex” approach to data centers has pushed some of the capital costs onto partners and turned them into operating costs for Apple. One of those partners is OpenAI, the developer of the “phenomenon” ChatGPT, which will be integrated into iOS this year.

OpenAI leases Nvidia GPUs from Microsoft. Apple also leases cloud space from providers like Amazon, Google, and Microsoft. The company declined to discuss the details of the OpenAI deal, but Cook left open the possibility of making money from the deal.

(According to CNBC, Apple)