Introducing GMI Cloud: New On-Demand Instances Speed ​​Up Access to NVIDIA GPUs


GMI Cloud is a new GPU cloud platform designed for AI and ML workloads, accelerating access to NVIDIA GPUs. The new on-demand cloud computing service is built for companies looking to leverage AI and go from prototype to production, giving users instant access to on-demand GPU computing resources from GMI Cloud. can do.

Rapid increase in compute demand

The rapidly increasing demand for AI compute power requires companies to take a strategic approach. In a rapidly changing landscape, companies are being asked to pay 25-50% upfront costs and sign three-year contracts with the promise of access to GPU infrastructure in 6-12 months. With the AI ​​shift, businesses need flexible computing power.

Instant GPU, infinite AI

Through collaboration with Realtek Semiconductor (TPE:2379) and GMI Technologies (TPI:3312), and by leveraging Taiwan’s strong supply chain ecosystem, GMI Cloud enables faster deployment and higher operational efficiency. It can be achieved. Our physical presence in Taiwan allows us to reduce GPU delivery times from months to days compared to non-Taiwan GPU providers. GMI Cloud will be a competitive new entrant in the market.

Also Read: Tassat Partners with Glasstower Digital on Cross-Border Digital B2B Payments

Alex Yeh, Founder and CEO of GMI Cloud, said: “Our mission is to fulfill humanity’s AI ambitions with an efficient, turnkey GPU cloud. We’re not just building a cloud, we’re building the backbone of the AI ​​era. GMI Cloud is dedicated to transforming the way developers and data scientists leverage NVIDIA GPUs and creating ways for humanity to benefit from AI.”

Why it matters

Technology leaders are seizing the opportunities presented by AI, but businesses large and small are facing barriers to accessing compute power.

New businesses don’t have the budget or long-term prospects to pay the initial costs of deploying large-scale GPUs. Companies need the flexibility to scale up or down depending on traction. So instead of having the funds to hire top AI talent, you need the option to pay for GPUs as an operating expense. On-demand access allows you to set up infrastructure without special skills, providing an option for teams that need instant, low-cost, and scalable access to GPU compute. .

Large companies face challenges as well. Enterprise data science teams need the flexibility to experiment, prototype, and evaluate AI applications to stay ahead of competitors before the AI ​​wave passes. However, not all companies are ready to commit to the long-term contracts and unproven capital investments required for large compute reserves. The flexibility of instant GPU access allows data science teams to conduct multiple prototype projects that require processing large datasets and fine-tuning models without incurring significant investment risk.

Get started now

GMI Cloud is a GPU cloud platform from NVIDIA with a rich preloaded Kubernetes-managed software stack designed specifically for AI and ML workloads. It includes pre-built images using NVIDIA TensorRT and will soon support all NVIDIA pre-built containers, including inference servers like NVIDIA Triton. Competitively priced at $4.39 per hour for NVIDIA H100 Tensor Core GPUs, GMI Cloud locks users into large reserve packages for large enterprises that charge up to 4x more for on-demand. provides on-demand access at a lower price compared to Instance types and sizes are purpose-built to efficiently deploy, fine-tune, and infer models of various sizes, including Llama3 8b, 70b, Mixtral 8x7b, Google Gemma, and Stable Diffusion.

SOURCE: BusinessWire