Apple’s latest Mac Studio models now offer a way to link multiple units into a unified cluster over Thunderbolt 5 and RDMA (Remote Direct Memory Access), enabling users to pool memory and significantly ramp up AI workloads. By connecting four Mac Studios together, Apple says inference workloads run up to three times faster than on a single Mac Studio, especially for demanding, “frontier-class” open-weight AI models. Mac Studio pairs—or groups—can share memory so that large models which don’t fit on one machine can be spread across several.
Technical Breakdown: Thunderbolt 5 Meets Shared Memory
The foundation of this clustering capability comes from the new Mac Studio’s adoption of Thunderbolt 5, which provides up to 120 Gb/s of external bandwidth. Alongside that, Apple integrated support for RDMA, a networking tech that allows memory operations to be moved directly between machines without taxing the CPU. Together, these let clustered Mac Studios use a shared memory pool to operate large AI models that previously required heavier infrastructure.
Even single Mac Studio units already offer sizable performance, thanks to Apple’s M5 Max and M5 Ultra chips. These machines also feature a PCIe Gen 6 SSD architecture, which Apple says can be up to twice as fast as its predecessor, along with the new Thunderbolt 5 ports.
When Clustering Matters
Not all AI users will need a cluster: for many tasks, a single Mac Studio will suffice. But for especially large inference workloads—such as running open-weight transformer models, or anything that stretches memory boundaries—clustering becomes a game-changer. Shared memory allows for model weights that exceed one system’s limits, and the performance bump (up to 3× faster with four units) helps justify the extra investment.
This feature joins a broader set of upgrades: Wi-Fi 7, Bluetooth 6, and newer SSDs for faster I/O; preorders are already open, with availability set for September 22 in the U.S. and many other regions.
For content creators or researchers chasing top-tier AI workloads without moving into full rack-scale clusters, this marks a shift. Mac Studios become modular AI workstations capable of scaling up within a desk setup.
The move toward clustered Mac Studios shows how Apple is targeting a more professional AI workflow, focusing on inference and memory-heavy models. For users working with bleeding-edge open-weight tools—or those who can spread their workload across multiple units—this clustering ability is a turning point. It’ll be worth watching how software frameworks and developers optimize for this shared-memory setup, and whether Apple provides tools to make clustering seamless.
