Armonk, NY — IBM and AMD have announced a strategic partnership to deliver AMD Instinct MI300X accelerators as a service on IBM Cloud, aimed at boosting performance for generative AI workloads and high-performance computing (HPC) applications. Expected to be available in the first half of 2025, this collaboration will enhance the scalability and efficiency of AI models and applications for enterprise clients, integrating AMD’s accelerators into IBM’s watsonx AI and data platform, as well as Red Hat® Enterprise Linux® for AI inferencing support.

Philip Guido, executive vice president and chief commercial officer at AMD, emphasized the need for high-performance accelerators to support the increasing demands of AI models and large datasets. “As enterprises expand their AI initiatives, the need for flexible, scalable solutions that can handle compute-intensive tasks without compromising on cost or performance is critical,” he said. “The combination of AMD Instinct accelerators and AMD ROCm software, together with IBM’s watsonx AI and Red Hat platforms, will provide a comprehensive ecosystem for running large-scale AI workloads.”

Alan Peacock, General Manager of IBM Cloud, echoed the importance of delivering AI solutions to meet enterprise needs: “Both IBM Cloud and AMD share a common vision of empowering enterprises to harness AI. By leveraging AMD’s advanced accelerators on IBM Cloud, we give clients the tools to scale AI applications while optimizing both cost and performance.”

This collaboration aims to help enterprises across various industries, including highly regulated sectors, utilize IBM Cloud’s security and compliance capabilities alongside the enhanced performance of the MI300X accelerators. The MI300X accelerators, featuring 192GB of high-bandwidth memory (HBM3), will support large model inferencing and fine-tuning, enabling enterprises to run larger AI models with fewer GPUs and potentially reduce inferencing costs.

Additionally, IBM plans to enable full integration of the MI300X accelerators within the IBM watsonx AI and data platform, providing enhanced infrastructure for scaling AI workloads across hybrid cloud environments. Red Hat Enterprise Linux AI and Red Hat OpenShift AI platforms will also be optimized to run large language models (LLMs) using InstructLab on MI300X accelerators.

IBM and AMD anticipate the general availability of the MI300X accelerators on IBM Cloud in the first half of 2025, promising enterprises a new level of flexibility and performance for their AI and HPC applications.

For further information on IBM’s GPU and accelerator offerings, visit IBM Cloud GPU.