Nvidia's Vera Rubin Platform Sweeps MLPerf Training Debut, Doubling GB300 Per-GPU Performance
- Dispatched: Nov 17, 2026 · 09:15 ET (14:15 UTC)
One-sentence summary: Nvidia announced on November 16 that its Vera Rubin NVL72 platform swept every MLPerf Training v6.1 benchmark, delivering up to twice the per-GPU performance of the previous-generation GB300 NVL72; on DeepSeek-V3 it scaled from 256 to 1,024 GPUs at about 97% efficiency, cutting training time from 16 minutes to about four, with agentic workloads topping 10x the throughput per watt of GB300.
Verification status: ✓ Official primary disclosure (Nvidia announcement; figures are the company's published MLPerf results)
Core Facts & Consensus Delta:
Nvidia's announcement was published at 20:13 ET on 2026-11-16 / 01:13 UTC on 2026-11-17: Vera Rubin NVL72 took first place in every MLPerf Training v6.1 test, with up to 2x the per-GPU performance of the GB300 NVL72 baseline.
On the DeepSeek-V3 benchmark, the platform scaled from 256 GPUs to 1,024 at roughly 97% efficiency, cutting training time from 16 minutes to about four. Nvidia also ran 10,240 GB300 GPUs, matching Microsoft Azure's 10,240 GB200 GPUs; interconnect is NVLink scale-up links plus Quantum-X800 InfiniBand and Spectrum-X Ethernet. Twenty partner organizations submitted results, including Dell Technologies and Google Cloud.
In the first MLPerf Endpoints round, Vera Rubin NVL72 delivered more than 10x the throughput per watt of GB300 NVL72 on agentic AI workloads at identical response speeds; that result is a provisional test using the Kimi K3 model. The same day Nvidia also introduced the Rubin HPC GPU, an expanded Vera CPU platform, the Nemotron science model, and a quantum collaboration with IBM and the U.S. Department of Energy's Fermi National Accelerator Laboratory.
Supply-Chain & Market Read-Through: Rubin is the central growth engine for Nvidia's fiscal 2027, and this debut sets its performance baseline; tonight's earnings call will be judged on Rubin shipment pace and the margin path, and the benchmark data gives the market a reference point for whether the ramp is delivering.
The supply chain is already moving to volume: ASUS announced on November 17 it is ramping volume shipments of Vera Rubin NVL72 systems in the fourth quarter, and Pegatron showed a rack design built around 256 Vera CPUs at SC26. As system builders accelerate stocking, HBM and advanced-packaging demand will scale with rack shipments, touching HBM and memory producers (SK Hynix 000660.KS, Samsung Electronics 005930.KS, Micron MU) as well as foundry and packaging capacity.
On sourcing: the figures above are Nvidia's published MLPerf results, and the Endpoints Kimi K3 number is provisional; independent third-party replication awaits broader deployment, and benchmark results should not be read as end-customer purchase volumes.
Next Catalysts & Watchpoints:
- Nvidia's fiscal third-quarter 2027 results are due 2026-11-17 at 16:20 ET / 21:20 UTC, with the call at 17:00 ET / 22:00 UTC: watch the Rubin revenue-recognition pace and fourth-quarter guidance.
- Further system-builder shipment announcements during SC26, and the next MLPerf round.
Sources:
- Silicon Report (Published: 2026-11-16 20:13 ET / 2026-11-17 01:13 UTC)
- emergent.sh
- TradingView, republishing PRNewswire (ASUS shipment announcement) (Published: 2026-11-17 01:30 ET / 06:30 UTC)
Information only, not investment advice.
Read full 2026-11-17 daily market brief & cross-asset tape → All flash alerts
More flash alerts from 2026-11-17
- WSJ: SEC Is Investigating Whether Nvidia's Disclosures Misled Investors⚠ UNCONFIRMED
- Nvidia Fiscal Q3 Revenue Hits $115.8 Billion, Beats Estimates; Q4 Guidance at $130 Billion✓ VERIFIED
- Iran Drones Strike US Radar System in Bahrain Again; Bahrain Says Attack Intercepted, No Casualties⚠ UNCONFIRMED
- Nvidia Hires Heidrick & Struggles to Lead CEO Search⚠ UNCONFIRMED
- Apple Reportedly Near Four-to-Five-Year DRAM Supply Deal With Samsung⚠ UNCONFIRMED
- Jensen Huang Breaks Silence: Would Be 'Honored' to Return as Nvidia CEO If Board Deems It in Shareholders' Interest✓ VERIFIED
- St. Louis Fed's Musalem: Several Rate Hikes Likely Needed Before Spring 2027 as Real Policy Rate Must Reach 2%–2.5%✓ VERIFIED
- Alibaba Launches Flagship Model Qwen4-Max: 2.4-Trillion-Parameter Sparse-Expert Architecture, Task Cost Down 62%✓ VERIFIED
- LPG Tanker Hit Twice in Strait of Hormuz: Crew Abandons Ship, One Missing, Two Injured✓ VERIFIED
- China Vows Countermeasures If US Presses Ahead With Tighter AI Chip Curbs✓ VERIFIED
- Nvidia's 13F Filing: Intel Stake Cut 25% in Q3 as SpaceX Becomes Largest Holding✓ VERIFIED
- Nvidia Board Establishes National Security and Export Control Oversight Committee✓ VERIFIED