This article was contributed by Nscale, in collaboration with Joerg Roskowetz, Director, Solution Architect AI, AMD. For more information, visit https://www.nscale.com.
Nscale has launched its latest GPU cluster, powered by AMD InstinctTM MI250X accelerators in its Glomfjord data centre in Norway. Currently powering some of the world’s top supercomputers, AMD InstinctTM MI250X GPUs are designed to supercharge HPC workloads and meet the growing demands of AI training, fine-tuning, and inference.
Strategically located near the Arctic Circle in Norway, our Glomfjord data centre leverages the environment’s natural cooling to optimise energy efficiency. Powered by 100% renewable energy, the data centre features advanced adiabatic cooling, ensuring streamlined operations, scalable solutions and an eco-friendly footprint.
Nscale has access to some of the world’s lowest-cost renewable energy, making it one of the lowest-cost AI training hubs. Our vertical integration and low-cost power enable us to deliver sustainable, high-performance AI infrastructure more quickly and cost-effectively than our competitors.
Experts in moving from CUDA to ROCmTM
Transitioning your workload from CUDA to ROCm can be complex. Our team at Nscale provides the support and guidance needed to make that migration process smooth and efficient. Whether you’re looking to explore the performance benefits of MI250X or preparing for the upcoming MI325X release, our live cluster is ready for you to experience the powerful capabilities of AMD Instinct accelerators.
Built for AI Workloads
Our vertically integrated platform is designed to easily support the most demanding AI workloads. Nscale provides bare metal and virtualised GPU nodes, Kubernetes-native services (NKS), AI workload scheduling powered by SLURM, and advanced inference services.
GPU Nodes
Nscale’s bare metal and virtualised GPU nodes are engineered for those demanding high-performance computing without compromise. With one-click deployment, you can set up your infrastructure in minutes, allowing your teams to focus on innovation while we handle the complexity.
Whether it's AI training, deep learning, or data processing, our service ensures you get peak performance with minimal overhead – for both short bursts and long-term projects.
Nscale’s Kubernetes (NKS)
Nscale’s Kubernetes Service (NKS) provides a managed environment for deploying containerised applications using Nscale's GPU-powered compute. The service provides a highly available control plane and lets you quickly provision worker nodes for scaling AI workloads.
This enables you to deploy, scale, and manage cloud-native AI applications, without the complexity of managing the underlying infrastructure.
Nscale SLONK
Nscale SLONK is a solution that enables you to manage and use your own high-performance compute cluster in the cloud. It provides a batch environment, including a comprehensive portfolio of AI/ML and HPC software designed to schedule, run, and monitor a range of scientific workloads, such as AI/ML training and large-scale simulations.
Nscale’s Infrastructure Services power our AI Workload Scheduler. The computing resources are accessible through the batch job scheduler SLURM. The software environment comprises a blend of proprietary software and open-source packages.
Nscale Inference Service
The Nscale Inference Service is a platform for deploying AI and ML models. Built using Kubernetes and KServe, it leverages cutting-edge horizontal and vertical scaling techniques enabling efficient hardware utilisation and cost savings. It's designed to streamline the process of deploying models into production.
It provides support for both serverless and managed compute deployments providing unrivalled flexibility and control over cost and performance.
Expanding Our Reach
With a growing pipeline of 1GW+ greenfield sites across Europe and North America, Nscale can deliver bespoke GPU clusters at any scale. This capacity, coupled with our strategic collaborations with industry leaders such as AMD, allows us to offer world-class infrastructure that meets the needs for AI training, fine-tuning, and inference—affordably and at scale.
With Nscale, you'll benefit from a robust, scalable AI infrastructure that grows with your needs, all while reducing costs and meeting sustainability goals.
To explore how Nscale’s AI infrastructure can power your next AI project, visit our website or reach out to the team for a tailored solution.