OpenAI has been reported to be aggressively purchasing Macs, buying tens of thousands at a time.
I accidentally bought out all the stock and now need to find a way to get more.
No MacBook laptops—only buy Mac Mini and Mac Studio without screens or keyboards.
So the question arises: What kind of AI workload cannot be handled by NVIDIA GPUs or Google TPUs, and requires a Mac instead?
Tens of thousands of Macs for reinforcement learning—A\ did it too.
According to Information, OpenAI has purchased tens of thousands of Mac minis and Mac Studios exclusively for reinforcement learning.
Not only OpenAI, but Anthropic is also renting Mac minis through Amazon Web Services (AWS) to perform similar tasks.
These Macs were used to train the "computer-use agent," an AI system capable of autonomously operating a computer to perform multi-step tasks such as editing test code, automatically organizing emails, and summarizing documents.
This surge is directly reflected in Apple's earnings report.
In the most recent quarter, Mac sales increased nearly 29% year-over-year to $10.3 billion, growing faster than all other Apple product lines, including iPhone and iPad, making Mac Apple's fastest-growing business.
On June 23, Apple held an event titled “Business at the Park” at its headquarters, Apple Park. This is uncommon in Apple’s history, as the company has traditionally focused on the consumer market and rarely hosts events specifically targeted at enterprise customers.
Executives from Disney and Ford attended the event, as did Anthropic co-founder Jared Kaplan, along with Apple’s outgoing CEO Tim Cook and incoming CEO John Ternus.
According to an attendee, Apple repeatedly emphasized that its hardware is ideal for handling AI tasks locally, with the Mac mini serving as the centerpiece of the event.
AI training has long been dominated by NVIDIA GPUs, but Macs have been widely adopted in the niche area of reinforcement learning thanks to unified memory.
NVIDIA GPUs have separate video memory and system memory, and data transfer between them creates a bottleneck. Apple’s M-series chips use a unified shared memory pool, allowing the CPU and GPU to access the same memory directly, providing performance advantages when handling AI workloads.
In addition, unlike the slim MacBook, the Mac mini and Mac Studio are equipped with dedicated cooling systems that prevent thermal throttling during extended periods of demanding AI tasks. This is crucial for reinforcement learning training that requires hours or even days of continuous operation.
Apple is still promoting the EXO Labs open-source software project, which enables multiple Macs to be combined into a cluster for running AI models with trillions of parameters locally.
Apple’s newly released Mac Studio also highlights its clustering capability, allowing multiple Mac Studio units to be linked together to form a more powerful system for running cutting-edge models.
The timing of this product launch is also unusual. Apple typically updates its Mac lineup in October or November each year, but this time it was brought forward to August.
NVIDIA has set its sights, and Apple is scrambling to respond.
Mac's rise in the local AI field has caught NVIDIA's attention.
According to a person familiar with discussions between NVIDIA executives and competitors, NVIDIA views Apple as its biggest rival in the local AI space.
At the end of last year, NVIDIA released the DGX Spark, an AI desktop computer designed to resemble the Mac mini, directly targeting this market.
Apple is facing the practical issue of supply unable to keep up.
The massive demand for memory chips from AI data centers has caused a historic industry-wide shortage, which has also affected Apple.
The high-end Mac mini and Mac Studio, most appealing to AI developers, have been out of stock for several months.
Former Apple AI product marketing manager Todd Dailey revealed that over the past year, due to Mac supply constraints, some enterprises have begun exploring alternatives, with NVIDIA DGX Spark being a frequently mentioned option—and it is now available in stock.
Daily left Apple in April this year and is currently an independent AI consultant. He also revealed that the Mac’s popularity in the enterprise AI market was entirely accidental, not part of Apple’s intentional strategy. Apple has no dedicated engineering team for enterprise customers and no employees focused on developer relations.
Apple last sold server products in 2011 with the discontinuation of the Xserve. The Mac-based server operating system was also discontinued in 2022.
However, some have already sensed the opportunity.
Peter Voell, a former employee of OpenAI’s computing infrastructure, founded Mount Thor, a cloud computing company based on Apple hardware, which is currently in stealth mode; the website describes its product as an “AI execution environment built on Apple hardware.”
Apple is also relying on partners like Mount Thor and webAI to expand the Mac's presence deeper into the enterprise market.
Apple has recently begun building its own servers using Mac chips. However, these servers are for internal use only, powering its Private Cloud Compute service to handle AI tasks that exceed the capabilities of iPhones or Macs.
Some enterprise customers have inquired whether Apple could lease out the use of these servers, but Apple has so far declined.
References: [1] https://www.theinformation.com/articles/apple-stumbled-ai-hardware-success-mac [2] https://www.apple.com/newsroom/2026/07/apple-reports-third-quarter-results/
This article is from the WeChat public account "Quantum Bit," authored by: Focused on Frontier Technologies