AI share of total IT capacity above which replayed H100 load swings on a generator prevented UPS bypass synchronization, in Vertiv's lab test with ST Telemedia and Baudouin
~75% of total IT capacityobserved
| 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 | One vendor-run laboratory test: H100 load profiles replayed by the Vertiv AI Load Simulator onto a Vertiv PowerUPS 9000 fed by a Baudouin generator (model not named). The same test found 20% input power smoothing cut generator voltage and frequency variation from about ±10% to about ±0.5%. The threshold depends on the genset's rating, governor and configuration. |
| Caveat | Vendor test, not an independent field measurement; the genset model and rating are not published, so treat 75% as an example of the failure, not a design limit. |
| As of | 2026-09 |
| Source | Vertiv blog, Powering AI at Scale: Pioneering Load Testing Insights with ST Telemedia and Baudouin (2026-09-03) · Results: 'When AI workloads exceeded approximately 75% of total IT capacity', 'frequency variations became severe enough to prevent bypass synchronization entirely'; 'Even at just 20% smoothing' reduced 'generator voltage and frequency variations from approximately ±10% to ±0.5%'; laboratory-controlled environment, four scenarios, H100 load profile, Vertiv PowerUPS 9000 |
| Review | checking…review by 2027-09-01 · standard cadence |
| Recorded changes | last 2026-10-19 |
| Claim id | vertiv-genset-h100-bypass-sync-threshold-2026 |
Where the guide uses it
Not quoted in a chapter yet; it is kept in the curated register.
← Full numbers register — every date-stamped figure in the guide, with revision history.