GMI Cloud just raised $668 million to expand its AI infrastructure business, but the number buried in the announcement matters more than the funding total itself. The company’s contracted annual recurring revenue now exceeds $600 million, more than nine times where it stood at the end of 2025. That’s the kind of growth rate that makes a $668 million raise look less like a bet and more like a company keeping pace with demand it’s already struggling to meet.
The financing splits into two pieces: a $223 million Series B equity round led by AI and robotics investor ARCHIV, with Nvidia also participating, and a separate $445 million credit facility led by CTBC. A wider group joined in too, including DSC Investment, Trend Micro, KB Investment, Kyobo Life and KT Corporation, a mix that spans cybersecurity, insurance and telecom money alongside the more typical AI-focused investors.
The capital is earmarked for expanding GPU capacity across the US, Taiwan and Southeast Asia, building out the company’s inference platform further, and hiring to support all of it. This builds directly on GMI Cloud’s existing Taiwan AI Factory project and its previously announced sovereign AI initiative in Japan, suggesting the company is deliberately building regional infrastructure rather than centralizing everything in one location.
The usage numbers back up the revenue growth. GMI Cloud says live annual recurring revenue from production services, as opposed to contracted deals still ramping up, has grown more than 4.5 times over the same period, and its inference platform now processes roughly four trillion tokens weekly. That’s a meaningful jump from forecasted demand to actual, paying usage, which tends to be the harder metric for infrastructure companies to grow quickly.
Customers already on the platform include Fireworks, Higgsfield, Nous Research, OpenRouter and Reflection, several of which operate their own AI products that depend on reliable inference infrastructure underneath them. GMI Cloud is positioning its multi-region setup as a direct answer to something those customers increasingly need: the ability to deploy workloads close to users while navigating different data residency and compliance requirements across the US and Asia, rather than forcing every customer onto infrastructure built for just one region.
