OpenAI Buys Tens of Thousands of Macs for AI Training, Challenging Nvidia's Dominance

The purchases have had a visible impact on Apple's financial results. In the latest quarter, Mac sales rose nearly 29% year-over-year to $10.3 billion, outpacing all other Apple product lines including the iPhone and iPad. At a June 23 event at Apple Park called "Business at the Park," Apple executives highlighted the hardware's suitability for local AI tasks, with the Mac mini taking center stage. Attendees included executives from Disney and Ford, as well as Anthropic co-founder Jared Kaplan.

Macs are not typically used for large-scale AI training, but they offer advantages in certain workloads. Nvidia GPUs separate video memory from system memory, creating bottlenecks, while Apple's M-series chips use a unified memory pool accessible to both the CPU and GPU. This design can improve performance for AI tasks. Additionally, Mac mini and Mac Studio have dedicated cooling systems that prevent thermal throttling during sustained training runs that last hours or days.

Apple has also been promoting EXO Labs, an open-source project that clusters Macs to run trillion-parameter AI models locally. The newly released Mac Studio emphasizes clustering capabilities, allowing multiple units to be linked together. Apple's early product launch this year, ahead of its usual October or November refresh, is seen as part of this push.

Nvidia has taken notice. According to a person familiar with discussions with Nvidia executives, the company regards Apple as its biggest competitor in local AI. Nvidia released the DGX Spark, a desktop AI computer styled similarly to the Mac mini, to compete head-on.

Apple, meanwhile, faces supply constraints. A historic shortage of memory chips, driven by AI data center demand, has left high-end Mac mini and Mac Studio configurations out of stock for months. Todd Dailey, a former Apple marketing manager for AI enterprise products, said some companies have begun seeking alternatives, frequently mentioning Nvidia's DGX Spark, which is available in stock. Dailey left Apple in April and is now an independent AI consultant. He also noted that Apple's success in the enterprise AI market was "completely unexpected" rather than part of a deliberate plan, as Apple lacks a dedicated enterprise engineering team.

Apple hasn't sold server products since discontinuing the Xserve in 2011, and its Mac-based server operating system was discontinued in 2022. But entrepreneurs are sensing opportunity. Peter Voell, a former OpenAI compute infrastructure employee, founded Mount Thor, a stealth cloud computing company that describes its product as an "AI execution environment based on Apple hardware." Apple is also relying on partners like Mount Thor and webAI to push Macs deeper into the enterprise.

Apple has begun building its own servers using Mac chips, but these servers are currently used only internally for Apple's Private Cloud Compute service, which handles AI tasks beyond the capacity of iPhones or Macs. Some enterprise customers have asked Apple if they could buy access to these servers, but Apple has so far declined.