Tech titans IBM and AMD have entered into a collaboration with Zyphra, an open-source AI research and product company based in San Francisco, to provide computing infrastructure for developing advanced AI systems.
Under a multi-year agreement, IBM will supply Zyphra with a cluster of AMD Instinct MI300X GPUs hosted on IBM Cloud. The infrastructure will be used to train multimodal foundation models that combine language, vision, and audio capabilities.
The agreement marks IBM Cloud’s first large-scale deployment of AMD Instinct MI300X GPUs, alongside AMD Pensando Pollara 400 AI NICs and Ortano DPUs. Zyphra began using the cluster in early September, with additional capacity planned for 2026.
Zyphra, which recently completed a Series A funding round valuing the company at $1 billion, focuses on open-source AI research in neural network architectures, long-term memory, and continual learning.
The company plans to use the new infrastructure to develop “Maia,” a general-purpose AI system intended to enhance productivity for enterprise users.
Krithik Puthalath, Zyphra’s CEO and chairman, said the partnership enables the company to advance its work on large-scale model training using AMD technology on IBM Cloud.
Alan Peacock, general manager of IBM Cloud, said the collaboration aims to provide scalable and cost-efficient infrastructure for AI model training, while AMD’s executive vice president Philip Guido noted that the partnership combines IBM’s enterprise cloud capabilities with AMD’s high-performance computing expertise.
IBM and AMD have been working together since last year to make AMD Instinct MI300X accelerators available as a service on IBM Cloud. Both companies also plan to explore new computing architectures, including quantum-centric supercomputing, combining IBM’s quantum systems with AMD’s AI acceleration technology.
