OneQode has announced a collaboration with AMD to deploy AMD Instinct GPUs and to introduce the AMD Helios rack-scale solution for its global AI infrastructure rollout.
This phased deployment will start with AMD Instinct MI355X GPUs and will later incorporate the AMD Helios platform as the foundation for OneQode’s planned expansion. The rollout will utilise the open AMD ROCm software stack to provide a standards-based, vendor-neutral platform for AI training and inference.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
OneQode founder and CEO Matthew Shearing said: “AMD is shipping some of the most compelling AI hardware in the market.
“The challenge now is putting that silicon in the right places, close to the customers, sovereign workloads, and AI-first organisations that need it. That’s what OneQode does.”
The initial phase of the deployment involves the use of AMD Instinct MI355X GPUs, which are based on 4th Gen AMD CDNA architecture.
Each MI355X GPU includes 288GB of HBM3E memory and delivers 8TB/s of bandwidth, with expanded datatype support for MXFP6 and MXFP4. AMD says this configuration enables high-density, large-scale infrastructure suitable for AI and HPC applications.
Moving forward, OneQode intends to deploy the AMD Helios rack-scale solution.
The Helios AI rack is based on an open design submitted by Meta at OCP 2025 and uses the next-generation AMD Instinct MI450 Series GPUs.
The MI450 Series offers up to 432GB of HBM4 memory and 19.6TB/s memory bandwidth per GPU.
A full rack can include up to 72 MI450 GPUs, delivering a reported 1.4 exaFLOPS FP8 and 2.9 exaFLOPS FP4 performance, with the aggregate memory bandwith reaching 1.4PB/s and a total memory pool of 31TB HBM4.
AMD AI business development corporate vice president Negin Oliver said: “Demanding AI workloads require high-performance compute, scalable infrastructure and an open software ecosystem.
“AMD Instinct GPUs and the unifying AMD ROCm open software stack are designed to help customers accelerate large-scale AI training and inference with the performance, efficiency and flexibility they need.”
OneQode previously entered a 110MW AI infrastructure agreement with Bitzero in Norway. The company is aiming to scale its AI capacity across Europe and Asia-Pacific, building on a cloud and telecoms network established over seven years across five continents.
The infrastructure is designed to support AI workloads such as frontier model training, enterprise deployments, and sovereign AI scenarios for governments, research bodies, and AI-first organisations.
Last month, OneQode, Hitachi Vantara, and Cylix Applied Intelligence revealed a Sovereign AI Factory initiative. This alliance targets deployment of sovereign AI infrastructure in Australia, Japan, Malaysia, and Singapore in its initial phase with future expansion intended for the US.
In this collaboration, OneQode supplies the infrastructure and connectivity. Hitachi Vantara provides AI platform integration with Hitachi iQ, and Cylix Applied Intelligence manages the AI operational layer from readiness assessments to managed production AI services.
