IBM and AMD to Deploy High-Performance AI Accelerators on IBM Cloud by 2025

IBM has unveiled a major initiative to deploy AMD's Instinct MI300X accelerators from Advanced Micro Devices (AMD) as a service on IBM Cloud by the first half of 2025. This strategy aims to enhance high-performance computing applications and generative AI for enterprise clients, delivering unprecedented speed and efficiency in corporate AI deployments.

IBM stated that this collaboration with AMD will tightly integrate the AMD Instinct MI300X accelerators with its Watson AI platform and Red Hat Enterprise Linux AI inference. This integration is set to significantly improve the scalability and cost-effectiveness of enterprise AI deployments. With 192GB of high-bandwidth memory, these accelerators will support extensive model inference and fine-tuning, potentially reducing the number of GPUs required and lowering overall costs.

It is reported that users will access these new accelerators through container services such as Kubernetes and Red Hat OpenShift, as well as IBM Cloud virtual servers for virtual private clouds. IBM clarified that these offerings are part of a broader plan to enhance AI infrastructure for enterprise clients, including those in regulated sectors.

IBM emphasized that this partnership demonstrates a shared vision with AMD to expand enterprise AI capabilities. By prioritizing scalability and operational efficiency, IBM will leverage the security and compliance tools of IBM Cloud to meet enterprise needs and drive the deployment of hybrid cloud AI solutions.

The service is expected to officially launch in early 2025, with further upgrades and optimizations anticipated as the collaboration deepens.