Data centreGermanyEIR-2257

Smouldering fire disables Lichtenberg II supercomputer at TU Darmstadt

2025Published 12 Oct 2026Responsible editor: Chief Editor

MajorElectrical

Estimated damageModelled by the register's adjusterHow? Know the real figure?

Low confidence

out of service for about •• daysShown in the record →Full estimate

From the recordRoot cause Failed part

By signing up you agree to the privacy policy.

What happened

In late March 2025 a small smouldering fire broke out in the electrical system that mainly supplied the second expansion stage of the Lichtenberg II high-performance computer (LB2A2) at the TU Darmstadt data centre in Darmstadt, Germany. The university computing centre (HRZ) says a short circuit and fire occurred on a power rail. A 2000 amp circuit breaker protecting the server room's main power rail tripped. The gaseous fire extinguishing system had to be refilled or replaced before normal operation could resume.

The fire caused a total failure of the whole HPC system and the adjacent housing area for about 2.5 weeks. The first expansion stage (LB2A1) and all GPU systems remained in operation. A provisional repair of the 2000A power rail was completed on 20 January 2026, and capacity was being restored in stages through 2026.

Contributing factors

An external expert report found that the oldest components of the electrical system were overloaded because of incorrect and insufficient specifications. The report recommended recalculating the whole heavy-current electrical system of the data centre. The university also reported delays in repairs and said cooperation and communication with the previous manufacturer of the electrical system were not optimal.

Lessons learned

Recalculate and verify the current ratings of heavy-current distribution against actual load whenever IT load is expanded.
Keep the original equipment supplier engaged and contractually responsive for fault investigation and repair.
Design power distribution so that a single busbar or breaker fault cannot take down a whole HPC system and adjacent facilities.

More lessons on investigation and fleet-wide checks, spacing, layout and fire spread, from every incident on the register.

Source

hrz.tu-darmstadt.de

Published 2026-10-12. Written from public reporting. Descriptive, not investigative. See the data accuracy disclaimer.

See the register's stories first in Google: add it as a preferred source.

Chart to reuse with this story: Data centre incidents by year, free with a credit (PNG).

The record behind this entry

The record behind EIR-2257 holds the site, the operator, the root cause, what failed, the fire service response and the downtime as fields you can filter, count and export, with its source and their verification status. The cause is not hidden: the contributing factors above give it as the sources report it. The record adds it as a field.

Site: ••••••••••••
Operator: •••••••••
Failed part: •••••••
Fire service response: ••••••••••
Downtime: ••••
Sources: ••••••••

Opens with this page:

Or , both for reading; or £39 a month, which also lets you propose edits to entries. See a complete record first, if you like.

Open the record in Lookup

Or get this one record free by email

Data centre incidents in Germany: 6 entries, with a free alert for the next one.

Browse the full register or submit an incident.