The topic of greatest interest to analysts on Apple’s quarterly earnings call on Thursday was a product that’s not even available to the general public yet – Apple Intelligence. The company’s forthcoming artificial intelligence system could spur a fresh cycle of iPhone upgrades and hardware sales. But CEO Tim Cook and CFO Luca Maestri spent a good part of the Q&A portion of the analyst call dodging questions about the pace of Apple’s rollout, whether the company is already seeing a sales boost from the service, and Apple’s deal with OpenAI to integrate ChatGPT into its software.

One question Cook was willing to partially address was about the company’s spending on AI servers. It’s an issue that’s come up throughout tech earnings season, as investors try to gauge where companies are in their AI infrastructure buildouts and how much more is coming.

“Embedded in our results this quarter is an increase year over year in the amount we’re spending for AI and Apple Intelligence,” Cook told CNBC’s Steve Kovach on Thursday.

Apple reported $2.15 billion in payments for property, plant and equipment in the June quarter, up 8% quarter-over-quarter and about 3% from a year earlier. Some of those capital investments aren’t for AI, but for other Apple operations.

The rise in Apple’s capital expenditure is tiny compared to its mega-cap peers, such as Microsoft, Google, and Meta. Those companies are spending huge sums to build and equip AI-focused data centers with Nvidia chips.

For example, in the June quarter, Microsoft reported $13.87 billion in capital expenditures, according to FactSet, which is a 55% year-over-year increase. Alphabet’s expenses jumped 91% to $13.19 billion, while Meta’s capital expenditures rose 31% to spent $8.3 billion during the quarter.

Yet Apple is Playing a Different Game

Unlike Amazon, Google and Microsoft, Apple doesn’t have a cloud business that involves renting out infrastructure to other companies. Meta isn’t in that business either, but the company is investing in training its own open-source large language model, and in using AI to power its massive recommendation engine.

Apple revealed this week in a technical paper that it rents Google TPUs in relatively small quantities to train its Apple Intelligence models.

As it builds out its infrastructure, Apple has the advantage of having designed its own chips, both for its phones and servers, so the company doesn’t have to spend billions of dollars on third-party processors.

In addition, Apple takes a “hybrid” approach to data centers that pushes some of its capital expenditures onto its partners, and turns them into operating expenses for Apple. One of those partners is OpenAI, whose ChatGPT technology will be integrated into iOS later this year. Apple also rents cloud capacity from providers including Amazon, Google, and Microsoft. For more on this, read the full CNBC report.