AMD launches Helion rack server for agentic AI workloads

AMD unveiled its AMD Helion data center rack server on Monday at the OCP Global Summit in San Jose, California. The system serves as a reference design for partners and clients, offering instructions on building systems for agentic artificial intelligence.

Helion can be fully equipped with sixth-generation EPYC CPUs for agentic AI tasks. It can also combine CPUs with Instinct MI350P GPUs to create a general-purpose AI platform, allowing customers to adapt the compute infrastructure to varied workloads.

Dan McNamara, senior vice president and general manager of compute and enterprise AI at AMD, stated that agentic AI requires flexible compute and infrastructure to handle everything from enterprise tools to code execution. The new design positions AMD to compete more directly against Nvidia's rack server designs, such as the NVL72.

In its most recent quarterly earnings report, AMD beat analyst forecasts as revenue grew 50% to $11.5 billion and earnings per share jumped 245%. Meanwhile, Nvidia posted $96.2 billion in sales for its prior quarter. AMD shares have gained roughly 182% year to date, compared with a 25% increase for Nvidia over the same stretch.