IBM和AMD正在授权一个新的AI集群, 用于启动Zyphra, 开发一个多式超级试剂, 挑战Nvidia的支配地位。 IBM and AMD are powering a new AI cluster for startup Zyphra to develop a multimodal superagent, challenging Nvidia’s dominance.

IBM和AMD发起了一个多年伙伴关系,向旧金山AI启动的Zyphrra(San Francisco AI)提供由AMD的Intinct MI300X GPUs、Pensando网络加速器和Ortano DPUs驱动的关于IBM云的大规模AI培训集群。 IBM and AMD have launched a multi-year partnership to provide Zyphra, a San Francisco-based AI startup, with a large-scale AI training cluster on IBM Cloud powered by AMD’s Instinct MI300X GPUs, Pensando network accelerators, and Ortano DPUs. 于2025年9月部署, 并定于2026年扩展, 基础设施支持Zyphra开发Maia, Deployed in September 2025 and set to expand in 2026, the infrastructure supports Zyphra’s development of Maia, a multimodal superagent designed to process language, images, and audio for enterprise productivity. 并挑战Nvidia的主导地位, 利用混合云和高性能计算实现可扩展的高效的AI创新。 The collaboration marks a major push to advance open-source AI and challenge Nvidia’s dominance, leveraging hybrid cloud and high-performance computing to enable scalable, efficient AI innovation.