IBM and AMD Team with Zyphra to Build AI Infrastructure on IBM Cloud

Analyst(s): Ray Wang
Publication Date: October 6, 2025

IBM and AMD are partnering with Zyphra to build one of the largest generative AI training clusters to date. Powered by AMD Instinct MI300X GPUs on IBM Cloud, the collaboration aims to support Zyphra’s $1 billion open-source superintelligence lab, which is developing multimodal AI models.

What is Covered in this Article:

• IBM and AMD’s multi-year agreement with Zyphra for advanced AI infrastructure.
• Deployment of AMD Instinct MI300X GPUs and networking hardware on IBM Cloud.
• Zyphra’s plans to use the cluster for Maia, its general-purpose enterprise AI superagent.
• Context on Zyphra’s $1 billion valuation and open-source superintelligence mission.
• Strategic implications of IBM and AMD’s collaboration on the AI infrastructure market.

The News: IBM and AMD have announced a multi-year deal to build advanced AI infrastructure for Zyphra, an open-source AI research and product company based in San Francisco. Under the agreement, IBM Cloud will deliver one of the largest dedicated clusters of AMD Instinct MI300X GPUs, combined with AMD Pensando Pollara 400 AI NICs and Ortano DPUs, to power training for frontier multimodal foundation models.

Zyphra, which recently raised a Series A round at a $1 billion valuation, will use the cluster to train language, vision, and audio models for Maia, its enterprise-focused general-purpose AI agent. The first phase went live in September, with further scaling set for 2026.

IBM and AMD Team with Zyphra to Build AI Infrastructure on IBM Cloud

Analyst Take: The IBM-AMD-Zyphra deal marks a latest move in the AI infrastructure space, blending enterprise cloud experience, advanced GPU, and an open-source AI push. With Zyphra now valued at $1 billion and aiming to create a superintelligence lab, this partnership is designed to bring the scale, efficiency, and flexibility needed for frontier model training. It’s notable not just for Zyphra’s ambitions but also for IBM and AMD’s positioning in a market still dominated by NVIDIA systems.

Scaling AI Model Training on IBM Cloud

The core of the partnership is IBM Cloud’s rollout of AMD Instinct MI300X GPUs, the first large dedicated training cluster on IBM Cloud built entirely with AMD hardware. This gives Zyphra the infrastructure to speed up multimodal model training at levels usually seen only among the biggest AI players. AMD’s end-to-end stack – from compute to networking – has been scaled on IBM Cloud for the first time. The setup shows IBM Cloud’s reliability and scaling while highlighting AMD’s role in high-performance AI. The companies are allowing Zyphra to move from startup scale into enterprise-level AI development.

Backing Zyphra’s $1 billion Open-Source Superintelligence Lab

With its $1 billion Series A funding, Zyphra now has the financial base for its open-source superintelligence goals. Partnering with IBM’s enterprise infrastructure and AMD’s chip technology gives it access to resources far too costly to build alone. The cluster will train multimodal foundation models that power Maia, Zyphra’s AI superagent for enterprise knowledge workers. Its focus on fresh neural architectures, long-term memory, and continual learning shows a push to stand apart from closed-source AI. This partnership strengthens Zyphra’s position as a leading open-source option in the AI research field.

IBM and AMD’s Wider Tech Roadmap

Looking past Zyphra, IBM and AMD are aligning around the future of AI and high-performance computing. Their hybrid cloud model offers enterprise clients scalable, cost-effective AI infrastructure with multi-cloud flexibility. They have also announced plans to work together on quantum-centric supercomputing, blending IBM’s quantum systems with AMD’s accelerators. This signals a long-term alignment on next-gen computing.. As Zyphra grows its training capacity in 2026 and beyond, IBM and AMD are set to scale with it while also expanding their joint offerings to other enterprises and startups.

What to Watch:

• Scaling challenges for Zyphra’s multimodal model training as infrastructure demands grow.
• AMD’s ability to translate Zyphra’s success into a broader enterprise AI market share.
• IBM Cloud’s competitive positioning against larger providers in AI workloads.
• Zyphra’s progress in advancing Maia and sustaining its open-source superintelligence mission.
• Energy and cost implications of large GPU cluster training at enterprise scale.

See the complete press release on the collaboration between IBM, AMD, and Zyphra on the IBM website.

Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.

Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.

Other insights from Futurum:

IBM Q2 2025 Earnings Exceed Expectations with Double-Digit Profit Growth, GenAI Book Surges Past $7.5B

IBM and AMD Join Forces on Quantum-Centric Supercomputing Initiative

Generative AI Gets Competitive: IBM’s Real-Time Strategy at the US Open – Six Five On The Road