GMI Cloud is a Taiwan-based neocloud that provides on-demand access to GPU infrastructure designed for AI training and inference workloads, including Nvidia H100, H200, and Vera Rubin chips. Its platform uses a proprietary Kubernetes-based system called Cluster Engine, which automates resource allocation so developers can spin up GPU clusters in seconds. The company operates dozens of data centers across Taiwan, Thailand, and Malaysia, and is expanding into the U.S. and Japan.
Led by ARCHIV, with participation from Nvidia, Redwood, DSC Investment, KT Corp., Kyobo Life, KB Investment, and Trend Micro; plus a $440 million credit facility from ChinaTrust Commercial Bank.
The raise combines a $223 million Series B equity round with a $440 million credit facility, reflecting the capital-intensive nature of building GPU data center capacity at scale across multiple continents. GMI Cloud's position in Taiwan's AI server supply chain gives it a procurement advantage that U.S.- and Europe-focused rivals like CoreWeave and Nebius lack, which matters as hyperscalers seek GPU capacity in Asia. The company reports its realized annual recurring revenue has grown five-fold in the past year, with its inference platform processing more than 4 trillion tokens per week, signaling that regional neocloud demand is converting into substantial contracted revenue.