MYBIGGAMING
LIVE
Agothic
Hardware 15 October 2026 • 2 min read •

Scaleway Launches AMD Instinct-Powered AI Cloud Services via Giga Computing

Scaleway is expanding its European AI infrastructure by integrating AMD Instinct MI355X GPUs into cloud platforms built on Giga Computing liquid-cooled servers.
Author: Гика • PC •
Scaleway Launches AMD Instinct-Powered AI Cloud Services via Giga Computing

Cloud provider Scaleway has unveiled new AI and cloud platforms featuring AMD Instinct GPUs at the ai-PULSE event in Paris. Giga Computing (a GIGABYTE subsidiary) is providing the core compute infrastructure, delivering the high-density servers and cooling solutions required for this large-scale deployment.

Expanding AI Capacity within Europe

The collaboration aims to bolster Europe's AI compute capabilities, offering enterprises and researchers high-performance cloud services within EU data centers. By incorporating AMD Instinct MI355X GPUs, Scaleway enables customers to train, fine-tune, and serve large-scale models using the open AMD ROCm software stack. This setup supports the Cloud Sovereignty Framework, ensuring that sensitive workloads remain under EU jurisdiction.

The hardware foundation consists of GIGABYTE G4L3-ZX1-LAT4 DLC servers. Each 4U high-density unit is equipped with eight AMD Instinct MI355X GPUs, each featuring 288GB of HBM3E memory, paired with two AMD EPYC 9005 CPUs. This configuration is specifically engineered to handle the massive memory bandwidth required for high-throughput AI inference and large model training.

Integrated Ecosystem and Thermal Management

To meet the demands of modern AI workloads, Scaleway has assembled a multi-vendor ecosystem to ensure production readiness. The infrastructure includes:

  • Bull for cluster testing and deployment;
  • VAST Data for the data platform;
  • Motul for closed-loop direct liquid cooling (DLC).

The integration of direct liquid cooling is a key technical component, allowing for higher rack power density and stable thermal performance even under sustained, full-load GPU operations.

Article author

Гика

Cookies

We use cookies

The site uses technically necessary cookies for sign-in, security, storing preferences, and correct page operation. Analytics and other non-essential technologies should only be enabled after your consent. You can read more about data categories, purposes, and retention in the privacy policy.

Cookie policy
Quick actions