Conference Presentation, Fireside Chat, Interview
AI Compute: Rebuilt for the Agentic Era | Mark Papermaster, AMD | RAISE Summit 2026
- AMD targets server market share leadership within the next few months while transitioning from historical 10% productivity gains to dramatic improvements in the current cycle.
- The company's GPU release cadence is on a manual schedule, coinciding with a shift in design efficiency that reduces feature development time from months to weeks or days.
- AI workload requirements are shifting such that four times more CPUs were previously needed, but the current CPU-to-GPU ratio is evolving toward a one-to-one relationship.
- A 72-node Helios AI rack is scheduled for release this fall, featuring significant efficiency improvements expected to be delivered in every generation.
- Two nanometer Venice CPUs will be unveiled alongside Helios rack details at the Advanced AI conference taking place in two weeks.
- Sales force adoption of AI training has reached 100%, with phase two in-depth solution training planned for the near future to support new capabilities.
- Organizational strategy involves restructuring to operate with startup-like speed to match market velocities, including a strategic plan to double down on European markets.
- AMD is prioritizing open ecosystems by collaborating with a broad partner base to build sovereign systems rather than limiting adoption to four or five key customers.