d-Matrix made Squadron inference clusters generally available to select customers, supporting 1,000-plus connected Corsair XPU accelerators across multiple racks.

d-Matrix announced Squadron, a fleet of multi-rack systems for AI inference at data-center scale. The company says the systems are now generally available to select customers, retaining an important qualification on the availability statement.

The announced scaling capability is more than 1,000 connected d-Matrix Corsair XPUs. That count refers to accelerators connected across a multi-rack system—not 1,000 racks, servers, or separately specified GPUs.

The announcement describes the systems as providing megawatt-scale inference compute for ultra-low-latency AI and emphasizes speed and energy efficiency. The excerpt supplies no exact system power rating, numerical latency results, throughput measurements, or independently verified efficiency comparison.