Rack-level energy storage per GPU in the Vera Rubin NVL72 power-shelf capacitor system, about six times the GB300 figure
~400 J/GPUobserved
| Value kind | observed — Reported measurements, counts, and specifications keep the precision and scope stated by their source; an exact specification is not treated as a range. |
|---|---|
| Scope | NVIDIA's design figure for the Vera Rubin power-shelf PSU capacitor system, from its Vera Rubin POD post of 16 March 2026; the same post says it cuts peak current demand by up to 25%. This is NVIDIA's design statement, not an independent field measurement. CoreWeave announced the initial bring-up and validation of Vera Rubin NVL72 racks on 1 June 2026, and NVIDIA's quarterly report dated 26 August 2026 says production shipments began in its fiscal third quarter of 2027. Shipping racks are not an independent test of the reservoir. |
| As of | 2026-03 |
| Source | NVIDIA Technical Blog, NVIDIA Vera Rubin POD: Seven Chips, Five Rack-Scale Systems, One AI Supercomputer (2026-03-16) · Intelligent Power Smoothing: 6x more rack-level energy storage (400 J per GPU) |
| Review | checking…review by 2027-03-01 · standard cadence |
| Recorded changes | last 2026-10-13 · 4 revisions tracked |
| Claim id | vera-rubin-power-smoothing-reservoir-target |
Where the guide uses it
- 4.5 UPS & Energy Storage: From Ride-Through to Transient Absorption
- 4.12 Metering, Power Quality, Monitoring & Electrical Operations
- 5.12 Cooling-Controls Transient Dynamics & Setpoint Stability
- 13.3 Electrical Power Acceptance (L3/L4)
- 13.4 Commissioning On-Site Generation & Microgrid Controls
- 13.6 Level 5 Integrated Systems Testing (IST) & Failure-Mode Demonstration
← Full numbers register — every date-stamped figure in the guide, with revision history.