NVIDIA Says Vera Rubin Hits 35X Tokens Per Megawatt
At the AI Infra Summit in Santa Clara on September 15, 2026, Ian Buck, NVIDIA's vice president of hyperscale and high performance computing, made AI factory efficiency the centerpiece of his infrastructure keynote. NVIDIA says the metric for AI infrastructure is shifting from peak performance to validated agentic tokens per megawatt.
Transcript
NVIDIA says Vera Rubin makes up to thirty five times more tokens per megawatt. Power is the constraint.
Ian Buck announced it Tuesday at the AI Infra Summit. The metric is shifting to tokens per megawatt.
That number pairs Vera Rubin with NVIDIA Groq three L P X at long context.
On a Semi Analysis benchmark of real agentic coding sessions Vera Rubin showed up to thirty times higher throughput per megawatt.
Perplexity saw sandbox starts nearly twice as fast on the Vera CPU.
Almost every figure here is an up to number supplied by NVIDIA. Your workload will differ.
Inboxsmith builds AI phone and email agents for small business. Please like and subscribe for more news.
Sources
Every claim in this video comes from the top ranking coverage of this topic. The claims and where each one came from:
- "The combined platform delivers up to 35X higher token throughput per megawatt than GB200 NVL72 for 2-trillion-plus-parameter models at long context." and "In AI factories, power is the constraint." Supports the hook.(NVIDIA (blogs.nvidia.com), first party corporate blog post titled "AI Infra Summit: NVIDIA Vera Rubin and DSX Platform Advancements Showcase Energy Efficiencies of Optimizing Tokens Per Watt for AI Factories", dated September 15, 2026, by NVIDIA Writers, filed under AI Infrastructure; read from the alert-time snapshot captured 2026-09-15T17:05:23Z.)
- "Ian Buck, vice president of hyperscale and high-performance computing at NVIDIA, Tuesday spoke on AI factory efficiency at the AI Infra Summit, the Santa Clara Convention Center event that has morphed into a Coachella of infrastructure tech." and "The metric for AI infrastructure is fast shifting from peak performance to validated agentic tokens per megawatt." Supports scene 1.(NVIDIA (blogs.nvidia.com), first party corporate blog post titled "AI Infra Summit: NVIDIA Vera Rubin and DSX Platform Advancements Showcase Energy Efficiencies of Optimizing Tokens Per Watt for AI Factories", dated September 15, 2026, by NVIDIA Writers, filed under AI Infrastructure; read from the alert-time snapshot captured 2026-09-15T17:05:23Z.)
- "That's where NVIDIA Groq 3 LPX comes in, adding deterministic ultralow-latency inference to Vera Rubin and complements DSX MaxLPS. The combined platform delivers up to 35X higher token throughput per megawatt than GB200 NVL72 for 2-trillion-plus-parameter models at long context." Supports scene 2.(NVIDIA (blogs.nvidia.com), first party corporate blog post titled "AI Infra Summit: NVIDIA Vera Rubin and DSX Platform Advancements Showcase Energy Efficiencies of Optimizing Tokens Per Watt for AI Factories", dated September 15, 2026, by NVIDIA Writers, filed under AI Infrastructure; read from the alert-time snapshot captured 2026-09-15T17:05:23Z.)
- "SemiAnalysis AgentX measures inference on recorded real-world agentic coding sessions, with actual context growth, tool calls delay and sub-agent spawning preserved, rather than on single-request benchmarks." and "On the DeepSeek V4 Pro model, Vera Rubin NVL72 delivers up to 30x higher throughput per megawatt than NVIDIA GB300 NVL72." Supports scene 3.(NVIDIA (blogs.nvidia.com), first party corporate blog post titled "AI Infra Summit: NVIDIA Vera Rubin and DSX Platform Advancements Showcase Energy Efficiencies of Optimizing Tokens Per Watt for AI Factories", dated September 15, 2026, by NVIDIA Writers, filed under AI Infrastructure; read from the alert-time snapshot captured 2026-09-15T17:05:23Z.)
- "Perplexity benchmarked the Vera CPU for its new SPACE secure sandbox platform for agentic AI, showing 1.9x faster sandbox starts." Supports scene 4.(NVIDIA (blogs.nvidia.com), first party corporate blog post titled "AI Infra Summit: NVIDIA Vera Rubin and DSX Platform Advancements Showcase Energy Efficiencies of Optimizing Tokens Per Watt for AI Factories", dated September 15, 2026, by NVIDIA Writers, filed under AI Infrastructure; read from the alert-time snapshot captured 2026-09-15T17:05:23Z.)
- Every headline figure in the post is an NVIDIA published "up to" number: "The combined platform delivers up to 35X higher token throughput per megawatt than GB200 NVL72...", "Vera Rubin NVL72 delivers up to 30x higher throughput per megawatt than NVIDIA GB300 NVL72", "NVIDIA DSX MaxLPS can deliver up to 1.4x more tokens per megawatt through factory-wide power optimization". The post also states "Vera Rubin is in full production and scaling across the ecosystem, and performance will continue to improve with ongoing software optimization", so the cited gains are vendor projections rather than independently measured customer deployments. Supports scene 5.(NVIDIA (blogs.nvidia.com), first party corporate blog post titled "AI Infra Summit: NVIDIA Vera Rubin and DSX Platform Advancements Showcase Energy Efficiencies of Optimizing Tokens Per Watt for AI Factories", dated September 15, 2026, by NVIDIA Writers, filed under AI Infrastructure; read from the alert-time snapshot captured 2026-09-15T17:05:23Z.)