AMD CEO Lisa Su Backs Open-Source AI and Announces Helios Server Production
AMD CEO Lisa Su has voiced strong support for open-source AI models, emphasizing that they provide transparency and control. Additionally, she announced that AMD's second-generation AI server, the "Helios" rack-scale system, is in full production and scheduled to ship by the end of Q3. AMD's endorsement aligns with Nvidia's stance, reinforcing industry support for open-source AI amid regulatory debates. Furthermore, the launch of the Helios rack-scale system positions AMD as a direct competitor to Nvidia's dominant rack-level AI infrastructure, with Microsoft already signed on as a major customer. The Helios rack-scale system integrates AMD's latest Instinct MI455X AI accelerators, EPYC Venice CPUs, and Pensando networking, all manufactured by TSMC. This is AMD's first fully integrated rack-scale solution designed to train and run large-scale frontier AI models.
## BACKGROUND
Rack-scale architecture treats an entire rack of servers, GPUs, and networking components as a single cohesive system, which is crucial for hyperscalers running massive AI workloads. As the debate over open-source AI versus closed proprietary models intensifies, hardware giants like Nvidia and AMD are advocating for openness to foster ecosystem growth and maintain hardware demand.