OpenAI has acquired thousands of Apple Mac devices to enhance AI training capabilities, marking a strategic shift in its operations.
OpenAI has made a significant investment in Apple technology by acquiring tens of thousands of Mac mini and Mac Studio computers to facilitate AI training, as reported by The Information.
The organization is leveraging these devices to enhance AI agents' ability to interact autonomously with desktop operating systems, focusing on tasks related to reinforcement learning and agent development.
In a contrasting approach, competitor Anthropic is renting Mac mini units from Amazon Web Services to address similar needs.
Shortly before this news broke, on August 25, Apple introduced updated models featuring the M5 Ultra and M6 chips.
The technical specifications reveal distinct roles for these computers in AI labs:
- Mac mini M6: Its compact form factor allows for the deployment of numerous independent working environments, essential for AI agents to learn interactions such as clicking and typing.
- Mac Studio M5 Ultra: With a capacity of up to 512 GB of unified memory, it can support local execution and hosting of highly demanding language models.
For example, the flagship Mac Studio M5 Ultra model, equipped with 512 GB of memory and a bandwidth of 1.2 TB/s, is capable of running the open language model GLM-5.3-Flash, which has 320 billion parameters and may require up to 300 GB of RAM for optimal performance.
In comparison, achieving similar capabilities on a traditional desktop setup would necessitate a complex arrangement of high-end RTX 5090 graphics cards, which would incur significant energy costs and bandwidth limitations.
Previously, on April 30, 2026, Apple CEO Tim Cook noted a shortage of Mac mini and Mac Studio units, attributing this to the rising demand from AI tools. The company has since repositioned its marketing to highlight these devices as ideal for "always-on local agent computing."
While Apple's technology is not intended to replace Nvidia's GPU clusters, which are the standard for training foundational models, it aims to establish a new infrastructure where trained neural networks can practice real-world tasks.
Details regarding the number of devices acquired, their specific models, or the total cost of this investment have not been disclosed by OpenAI or Anthropic.
