IBM and Intel bring Gaudi 3 AI Accelerators to IBM Cloud

In a major announcement made at Intel Vision 2025, IBM and Intel have joined forces to introduce the Intel Gaudi 3 AI accelerators on IBM Cloud. This collaboration is set to deliver a new level of performance and affordability for enterprise AI workloads, especially in the areas of generative AI, model fine-tuning, and inferencing.

Intel Gaudi 3 AI accelerators are now available in a public cloud environment for the first time, with IBM Cloud providing the hosting infrastructure. The service is currently live in Frankfurt and Washington, D.C. regions, with an expansion planned for Dallas in Q2 2025.

AI adoption is soaring, but infrastructure costs remain a concern. According to IBM’s “AI in Action 2024” report, 67% of business leaders reported a revenue increase of 25% or more due to AI. However, the challenge lies in scaling these solutions economically.

With this joint offering, IBM and Intel aim to make AI more accessible, secure, and budget-friendly. The Intel Gaudi 3 accelerators are optimized for high-throughput AI processing while being more cost-effective than some existing options in the market.

Businesses can choose how they want to integrate Gaudi 3 into their AI infrastructure:

• Standalone Servers via IBM Cloud VPC: Ideal for users who need full control over software stacks. These support Red Hat Enterprise Linux AI images and offer customizable compute, storage, and networking.
• Container Worker Nodes: IBM plans to integrate Gaudi 3 into Red Hat OpenShift AI clusters in Q2 2025, catering to developers who prefer a containerized environment.
• Bring-Your-Own watsonx.ai License: Enterprises can run IBM watsonx.ai, the company’s AI development studio, directly on Gaudi 3-powered virtual servers, giving them control over the entire AI stack.
• Deployable Architectures (DAs): For faster deployment, IBM will offer pre-configured architectural templates including support for watsonx, OpenShift, and VPC, set to launch in 2H 2025.

Beyond just performance, IBM and Intel are also focused on security and data integrity. The companies recently enabled Intel Trust Domain Extensions (TDX) on IBM Cloud VPCs to support confidential computing, ensuring that client data remains secure and isolated.

This new chapter in the IBM-Intel partnership opens doors for organizations looking to scale their AI ambitions without overspending on infrastructure.