Cloud infrastructure startup GMI Cloud Inc. today announced that it has closed a $82 million early-stage funding round led by Headline Asia.

The Series A investment also included the participation of Thailand-based energy company Banpu Next and Wistron Corp., a Taiwanese electronics maker. The bulk of the round, $67 million, took the form debt financing. The remaining $15 million was provided as equity funding.

Santa Clara, California-based GMI Cloud operates an infrastructure-as-a-service platform geared towards artificial intelligence workloads. The platform provides access to Nvidia Corp.’s H100 graphics processing units. A few weeks from now, GMI Cloud will also start offering the H200, an upgraded version of the H100 with more memory.

For users, spinning up a GPU-powered cloud instance usually takes several minutes. GMI Cloud says that its platform performs the task in a few fractions of a second. Customers can link together their instances into a cluster and, if necessary, run multiple clusters side by side. 

After spinning up a GPU environment, developers have to install the AI model they plan to run along with the external software components the model requires to work. Configuring those components can require a significant amount of time. To speed up the task, GMI Cloud provides preconfigured AI software that can be deployed without extensive customization. 

The company offers access to NIMs, Nvidia-developed versions of popular AI models. NIMs are packaged into containers to ease deployment. They also include performance optimizations designed to make AI models run faster on Nvidia silicon. 

For customers with more advanced requirements, GCI Cloud offers the ability to deploy custom models on its cloud. The platform integrates with several of the code editing tools that developers most commonly use to build AI applications. It also supports open-source machine learning frameworks such as PyTorch.

Under the hood, GCI Cloud’s platform is powered by a Kubernetes-based infrastructure management platform called Cluster Engine. The software automates the task of provisioning new instances and allocating hardware resources to AI workloads. If one of the servers on which a model runs experiences technical issues, Cluster Engine can move the data inside to a different machine without incurring downtime. 

Since launching in 2022, GMI Cloud built up a user base that includes several dozen companies. The company says that its customers include healthcare, telecommunications and research organizations.

GMI Cloud relies on data centers in Taiwan, Thailand and Malaysia to host customers’ workloads. It will use the funding round announced today to build a cloud facility in Colorado. Additionally, the company reportedly plans to hire up to 20 new employees by the end of the year. 

Image: Unsplash