Bank credit supplies two-thirds of the $663M GMI Cloud raised to expand GPU rentals

An infrastructure lead with a launch next quarter and most of her users in Thailand and Malaysia needs GPUs in that region by a date she can put in the plan. Alex Yeh, GMI Cloud's founder and chief executive, is selling to exactly that person. "In AI infrastructure, a delivery date is a promise," Yeh said. "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." ^[12]

Of the $663 million behind that promise ^[4], the $440 million ChinaTrust facility is about 66 percent ^[3][1]. The equity is the $223 million Series B ^[1]. Nvidia, whose chips are the product GMI Cloud rents out ^[6], took part in that round ^[1]. Trend Micro is on two lists: it invested in the Series B and it is a named customer ^[2][16].

Here's what the company pitches. "AI is driving a new renaissance and reliable compute is its foundation," Yeh said ^[11]. Here's what a buyer can test. SiliconANGLE describes GMI Cloud's Kubernetes-based Cluster Engine as letting developers spin up GPU clusters from preconfigured environments in a matter of seconds ^[7]. On traditional clouds, the publication says, developers configure those environments by hand ^[7].

The revenue evidence comes as ratios. Realized annual recurring revenue is up five-fold over the last year, the company said ^[14]. Contracted ARR is set to finish this year at ten times its December 2025 level ^[13]. The two figures measure different things over different windows, and neither is a retention number ^[13][14]. The company did not give either figure in dollars or publish GPU prices ^[13][14].

Yeh's case, as SiliconANGLE relays it, is that U.S. hyperscalers and AI startups want more GPU capacity in Asia, where they have large customer bases ^[17]. He adds thousands of APAC enterprises that need a local source of compute ^[17].

Two axes sort the decision. One is whether the workload has to run in Taiwan, Thailand or Malaysia. The other is whether the team can commit to a term or needs on-demand capacity. In region with a term to sign, GMI Cloud belongs on the shortlist. The delivery date Yeh calls a promise should be written into the contract, with a service credit if it slips ^[12]. The cost of that choice is the term itself: capacity booked there cannot move until the contract ends. In region and on-demand, a trial settles the Cluster Engine claim, since the time to a first working cluster can be measured on day one ^[7]. Out of region and on-demand, the round changes the least. Out of region with a term, the relevant sites are GMI Cloud's U.S. build, which is still under way, and a Japanese sovereign AI initiative launched earlier this year ^[9].