Scaleway Uses NVIDIA GH200 Grace Hopper Superchip

Paris, February 1st, 2024 — Scaleway, a leading European player in high-performance computing and renowned for its groundbreaking Nabu 2023 supercomputer based on the NVIDIA DGX SuperPOD architecture, unveils the newest addition to its cloud ecosystem – the NVIDIA GH200 Grace Hopper Superchip. This enhancement to the Scaleway cloud ecosystem empowers users with next-generation accelerated computing capabilities, delivering superior performance for AI and HPC workloads.

In the wake of the successful ai-Pulse conference, Scaleway reaffirms its commitment to providing customers with world-class hardware and software solutions for AI training, fine-tuning, and inference. The integration of the NVIDIA GH200 Grace Hopper Superchip will position Scaleway as a pioneer, helping ensure that AI practitioners have access to the most advanced technology available in the European market today.

Combining the new NVIDIA Grace CPU, based on Arm Neoverse V2 cores, and the proven NVIDIA Hopper architecture, the NVIDIA Grace Hopper Superchip offers up to 576GB of direct-access memory through an extremely powerful NVIDIA NVLink-C2C to connect the CPU and GPU.

The NVIDIA Grace Hopper Superchip offers powerful performance for both training and inference of AI and HPC workloads. Scaleway customers will be able to use standalone Grace Hopper Superchip instances or scale up to meet their workload needs.. The NVIDIA GH200 NVL32 is a groundbreaking AI supercomputer with a terabyte-class memory. This system unites 32 NVIDIA Grace HopperTM Superchips using a 14.4TB/s NVLink Switch System, creating a powerful single GPU. It’s perfect for handling huge AI models, including recommender systems, generative AI, and graph analytics. It provides fast access to 20TB of high-speed memory and upward to 128 petaFLOPS of FP8 performance in a compact rack, scalable for advanced foundation models. Its smaller size and flexible design offer unparalleled scalability for cutting-edge AI applications.

"Being the pioneer in using NVIDIA Grace Hopper is a testament to Scaleway’s unwavering dedication to next-generation solutions. This groundbreaking addition not only underscores our commitment to offering unparalleled hardware for AI and HPC, ensuring our customers not only possess the essential tools to lead in HPC, but also establishes Scaleway as the forefront innovator, championing a resilient European alternative in this dynamic landscape” emphasizes Damien Lucas, CEO of Scaleway.

Scaleway instances with NVIDIA Grace Hopper will enable new workloads to be seamlessly deployed in the cloud. For businesses and researchers at the forefront of AI technology, Scaleway’s deployment of NVIDIA Grace Hopper represents a significant leap forward, opening up access to state-of-the-art computing resources for pushing the boundaries of innovation.
Customers can pre-register their interest here.

About Scaleway

Founded in 1999, Scaleway is a Cloud Provider that supports the development, training, deployment, and scaling of infrastructure in a complete Cloud ecosystem. Allowing European organizations to train, test, develop, and host in a resilient environment with access to more than +50 cloud services. Located in Paris, Amsterdam and Warsaw, Scaleway’s Cloud portfolio is used by 25,000+ businesses, who choose Scaleway for its multi-AZ redundancy, and native secured user management and security services.

With fully managed services from Dedicated servers, Bare Metal, Computing, Containers and Serverless, Scaleway provides flexibility in Cloud infrastructures. Offering customers the ability to choose where their customer’s data resides, to choose what architecture works best for their business, and to choose a more responsible way to scale.
Scaleway is the Cloud provider that enables AI innovation in Europe.

Press Contact :

Tiphaine Szczypawka, Communications Manager,

Published on