GMI Cloud has raised roughly ₿7,706 ($668 million) to expand the physical infrastructure behind its AI cloud, pairing a large equity round with a credit facility as demand for GPU capacity keeps pushing neocloud operators toward increasingly capital-intensive builds.
Using Bitcoin at approximately $86,682 per BTC on October 2 for conversion, the financing breaks down to about ₿2,573 ($223 million) of Series B equity and ₿5,134 ($445 million) in credit. In its financing announcement, GMI said ARCHIV led the equity round, NVIDIA participated, and CTBC led the credit facility.
The money is going into GPU capacity, not just software
The expansion targets the United States, Taiwan and Southeast Asia. That makes the round another example of AI infrastructure behaving more like industrial infrastructure: GPUs must be purchased and installed, data halls need power and cooling, networks need enough bandwidth to feed accelerators, and storage has to keep large training and inference workloads supplied with data.
GMI Cloud’s own announcement on X says the capital will expand GPU capacity and scale its inference platform while contracted annual recurring revenue has grown more than ninefold since the end of 2025.
A ₿6,922 ($600 million) contracted-revenue base changes the story
The financing is notable because GMI says contracted ARR has surpassed roughly ₿6,922 ($600 million). Data Center Dynamics reported that the company plans to use the capital for U.S., Taiwan and broader APAC capacity expansion, inference services and hiring.
That demand profile resembles the infrastructure cycle BitcoinVersus.Tech has been tracking across the sector. QumulusAI’s activation of 616 Blackwell GPUs for Runpod showed how quickly new accelerator capacity can become a cloud product, while Crusoe’s compute agreement with Perplexity demonstrated how AI companies are increasingly securing infrastructure through specialized GPU-cloud operators.
Inference is becoming an infrastructure workload of its own
Training giant models gets the headlines, but serving models continuously can become an equally demanding infrastructure problem. GMI says it is processing roughly one trillion inference tokens per day, up about twentyfold in six months. Sustaining that traffic requires more than accelerators: scheduling, networking, storage, observability and thermal stability all affect how much useful work a GPU fleet can deliver.
AI clouds are starting to look like financed infrastructure projects
The ₿5,134 ($445 million) credit component is particularly revealing. GPU clouds are no longer scaling only through venture equity. They are increasingly mixing equity, debt, customer commitments and hardware economics to finance facilities whose useful output is compute rather than office space or traditional industrial production.
BitcoinVersus.Tech recently examined NVIDIA’s effort to make GPU-backed AI loans easier to finance. GMI’s round fits the same larger transition: accelerators, power, cooling and contracted compute revenue are being assembled into an investable infrastructure stack.
For builders, the competitive question is moving beyond who can obtain the newest GPU. The harder problem is who can finance, energize, cool, network and keep those GPUs utilized across continents. GMI Cloud’s new capital is a bet that the next AI bottleneck will be solved as much by infrastructure execution as by model architecture.
BitcoinVersus.Tech
Advertisement
Editor’s Note
We volunteer daily to ensure the credibility of the information on this platform is Verifiably True. If you would like to support to help further secure the integrity of our research initiatives, please donate here: 3C9o19EH5HSiwEPyCTmEKzxhNCbo2X6TTb
BitcoinVersus.tech is not a financial advisor. This media platform reports on financial subjects purely for informational purposes.
Leave a comment