BRIGHT EVIDENCE PACK / Deployed
A hybrid cooling system cut direct water use at an operating national-laboratory data center.
DOE reports that NREL’s thermosyphon hybrid system avoided 2.10 million gallons over two years and cut data-center water use in half by using dry heat rejection when weather allowed while retaining a cooling tower for peak conditions.
Canonical Bright record · JSON evidence pack · Key-facts embed
Dates and assessment
- Source published
- 2019-05-01
- Bright published
- 2026-09-19
- Substantive update
- None recorded
- Evidence state
- Deployed
- Independent verification
- Not established by this source review
- Last source review
- 2026-09-19
The claim in context
The human problem
Evaporative cooling can reject heat efficiently but requires a continuing water supply, especially during hot conditions when local systems may already be stressed.
The prior constraint
NREL’s warm-water liquid-cooled HPC system could reuse some heat, but the remaining heat still relied heavily on evaporative towers.
AI’s actual role
The HPC facility supports advanced computing; the documented result is conventional thermal engineering and measured operating water reduction, not an AI-model outcome.
The documented result
A May 2019 DOE case study reports that the roof-mounted thermosyphon cooler avoided 1.16 million gallons in its first operating year and 2.10 million gallons over two years, cutting data-center water use in half. Controls route heat to campus reuse first, dry rejection when conditions permit, and the existing open cooling tower for remaining and peak loads.
Why it may matter
It is an operating example of a hybrid strategy that reduces direct cooling water without pretending the wet system disappeared or one design fits every climate.
Limitations
- The result comes from a 10,000-square-foot national-laboratory HPC facility, not a hyperscale commercial campus.
- The system retained full cooling-tower capacity for the hottest hours and did not eliminate direct water use.
- DOE reported the case; Bright did not audit meters, loads, weather normalization, or later performance.
Original evidence
Attribution
Credit Bright AI Future and link the canonical Bright record.
- Link to the canonical Bright record.
- Keep material limitations with the claim they qualify.
- Link to the original evidence when repeating a substantive claim.
- Do not describe a source check or organization-reported result as independent verification.
Linked source material, quotations, trademarks and media remain subject to their owners’ terms. No reuse right is granted for third-party media.