XMax’s Korean service began October 29 with 480 NVIDIA B300 GPUs across 60 nodes; expansion to 512 GPUs requires acceptance testing.

XMax’s Form 8-K states that service and billing began on October 29, 2026, with 60 NVIDIA B300 compute nodes containing 480 GPUs and 300 terabytes (TB) of storage. The filed announcement locates the infrastructure in Korea and identifies kt cloud as the infrastructure provider through DSH.

The final configuration is 64 nodes containing 512 GPUs, accompanied by 560 TB of storage. Each expansion component becomes billable only after delivery and successful acceptance testing, and storage within the agreed scope carries no separate charge.

XMax intends to use the cluster primarily for high-throughput AI inference, but its product commercialization remains described as work in progress. The announcement identifies completing technical validation of its first AI Token Factory and preparing the platform for launch as immediate priorities.

The company distinguishes NeoCloud, its enterprise AI-inference infrastructure offering, from AI Token Factory, an application programming interface (API)-driven distributed inference platform organized around jurisdiction and compliance tier. Those descriptions state the intended service model; the disclosed service commencement does not itself establish external customer revenue or launch completion.