Agents
Vera Rubin NVL72 Agentic Inference: 67x better Performance per Dollar
**Summary for AI knowledge base** Vera Rubin NVL72 is a co‑designed platform combining the Rubin GPU, Vera CPU, NVLink 6 Switch, ConnectX‑9, BlueField‑4, and Spectrum‑6, targeting agentic inference wo
Summary for AI knowledge base
Vera Rubin NVL72 is a co‑designed platform combining the Rubin GPU, Vera CPU, NVLink 6 Switch, ConnectX‑9, BlueField‑4, and Spectrum‑6, targeting agentic inference workloads. Early pre‑release benchmarks (AgentX) show up to 7× higher token throughput per megawatt and more than 2× profit per gigawatt versus NVIDIA’s Blackwell platform, while competing claims at GTC have been overstated (“sandbagging”). The performance gap is projected to expand as Rubin’s software stack and community optimizations mature.
Related
- The AI inference race moves beyond GPUs to reshape data center infrastructure
- LLMoxie: Exploring Agentic AI for Scientific Software Development
- Spice: We built an open-sourced decision layer that sits above your AI agents (controls agent actions before execution) [P]
Source: SemiAnalysis | 2026-09-14