GMI Cloud is a GPU-based cloud provider that delivers high-performance and scalable infrastructure for training, deploying, and running artificial intelligence models.

GMI Cloud supports users with three key solutions. The Inference Engine provides ultra-low latency and automatically scaling AI inference services, the Cluster Engine offers GPU orchestration with real-time monitoring and secure networking, while the GPU Compute service grants instant access to dedicated NVIDIA H100/H200 GPUs with InfiniBand networking and flexible on-demand usage.

Currently, NVIDIA H200 GPUs are available, and support for the Blackwell series will be added soon. In the Cluster Engine (CE), scaling is not automatic — customers need to adjust compute power manually using the console or API. By contrast, the Inference Engine (IE) supports fully automatic scaling, allocating resources according to workload demands to ensure continuous performance and flexibility.

NVIDIA H200 GPUs are available on-demand at a list price of $3.50 per GPU-hour for bare-metal as well as $3.35 per GPU-hour for container. The pricing follows a flexible, pay-as-you-go model, allowing users to avoid long-term commitments and large upfront costs. Discounts may also be available depending on usage.

As a NVIDIA Reference Cloud Platform Provider, GMI Cloud offers a cost-efficient and high-performance solution that helps reduce training expenses and speed up model development. Dedicated GPUs are instantly available, enabling faster time-to-market, while real-time automatic scaling and customizable deployments provide users with full control and flexibility.