Taiwan-based neocloud GMI Cloud Inc. wants to capitalize on what it says is an “unprecedented” demand for artificial intelligence computing power after raising a massive $663 million in funding from two sources.
The startup said today it has just closed on a $223 million Series B round of funding led by the robotics and AI-focused investment firm ARCHIV, with participation from Nvidia Corp., Redwood, DSC Investment, KT Corp., Kyobo Life, KB Investment and Trend Micro Inc. At the same time, it has secured a $440 million credit facility from ChinaTrust Commercial Bank, one of the largest financial institutions in Taiwan.
GMI Cloud is one of a host of specialized cloud infrastructure providers that have emerged in recent years to cater to the growing demand for access to graphics processing units and other resources needed to power AI workloads. The best-known of these neoclouds is CoreWeave Inc., which went public last year and has quickly become one of the darlings of Wall Street amid the AI boom. In Europe, there’s Nebius Group N.V., which rebranded from Yandex N.V. after selling off its Russian search engine firm Yandex.
Like its rivals, GMI Cloud offers an AI infrastructure-as-a-service platform that’s specially designed for resource-intensive AI training and inference workloads. It provides on-demand access to some of the most powerful GPUs available, including Nvidia’s H100 and H200 chips and its even newer Vera Rubin GPUs.
Its platform is powered by a proprietary Kubernetes-based infrastructure management system it calls the Cluster Engine, which automates resource allocation and instance provisioning so that developers can focus on their applications only. It’s a powerful advantage, because it means developers can spin up GPU clusters based on preconfigured environments in a matter of seconds. On traditional cloud infrastructure platforms, developers would have to manually configure these environments themselves.
GMI Cloud’s strong presence in the Asia-Pacific region is another key advantage, especially compared with CoreWeave and Nebius, whose data centers are mostly centered on the U.S. and Europe. The startup has dozens of data centers dotted around Taiwan, Thailand and Malaysia, which puts it in a strong position to cater to enterprises that want to run AI workloads locally in the region. However, it’s also building up a presence in the U.S., as well as a sovereign AI initiative in Japan that was launched earlier in the year.
Founder and Chief Executive Alex Yeh said demand for AI compute infrastructure has gone global. U.S.-based hyperscalers and AI startups are desperate to find more GPU capacity in Asia, where they have significant customer bases. In addition, there are thousands of APAC-based enterprises spearheading AI projects too, and all of these need a local source of compute.
“Our customers are scaling faster than ever and they need infrastructure that keeps pace,” Yeh said. “AI is driving a new renaissance and reliable compute is its foundation. Our goal is to build that foundation across continents, with an ecosystem of products on top of it.”
Being based in Taiwan gives GMI Cloud another major advantage, because the country also happens to be the world’s biggest factory for the AI server systems that populate data centers around the globe. Its close proximity to so many suppliers means the startup has an extremely secure supply chain and the ability to quickly procure all of the systems it needs to get its new data centers up and running, no matter where in the world they’re located.
“In AI infrastructure, a delivery date is a promise,” Yeh explained. “Customers plan launches, hiring and revenue around those dates. Our place in Taiwan’s supply chain is how we keep that promise, cluster after cluster globally.”
This reliability has already paid off for GMI Cloud in a big way. The startup said the amount of annual recurring revenue it has under contract is set to grow 10-fold at the end of the year from what it was in December 2025. As for ARR that’s already being realized, this is up five-fold in the last year, with its inference platform processing an average of more than 4 trillion tokens per week. It boasts a number of major enterprise customers, including Higgsfield Inc., OpenRouter, Nous Research Inc., Fireworks AI Inc. and Trend Micro.