OpenAI buys Mac minis for reinforcement learning and computer-use agents
OpenAI is reported to have bought tens of thousands of Mac minis and Mac Studios in recent months for reinforcement learning and for training computer-use agents. The purchases are separate from the cloud graphics-processor clusters the company leases for large-scale model training. Apple's unified memory lets the central processor, graphics processor and neural engine draw on a single shared memory pool, a fit for long-running workloads. Rival labs have bought the same machines; Anthropic is described as renting them through AWS.
Artificial Intelligence··Morning
A purchase for reinforcement learning and computer-use agents
OpenAI is reported to have bought tens of thousands of Mac minis and Mac Studios in recent months for reinforcement learning and training computer-use agents. The same purchase is for agents that run multi-step work on their own. A report sourced to The Information says OpenAI and rival AI labs have bought the machines. The purchases are separate from the cloud graphics-processor clusters the company leases for large-scale model training.[2], [1]
Unified memory, strong chips and cooling
Apple's unified memory design lets the central processor, graphics processor and neural engine draw on a single shared memory pool, which is seen as a fit for this kind of workload. The reported reason is that strong chips, shared memory and solid cooling suit long-running AI workloads. The hardware choice sits apart from the graphics-processor clusters used for large-scale model training; the target is reinforcement learning and agents that use a computer step by step.[2], [1]
Rival labs, a supply squeeze and Mac revenue
Rival labs have bought the same machines, and Anthropic is described as renting them through AWS. Peter Voell, formerly of OpenAI's computing infrastructure team, has built an Apple-based cloud service called Mount Thor. Apple's Mac revenue rose nearly 29 per cent in the June quarter to 10.4 billion dollars. The purchase volume is reported to have contributed to the supply squeeze that pushed Apple to refresh both product lines earlier than expected. The most powerful models have been sold out for months because of a memory chip shortage.[1], [2]
