German register exposes utilisation and data gaps
·

German register exposes utilisation and data gaps

Germany’s first official data centre register analysis finds many listed facilities already satisfy future efficiency requirements, but low utilisation, incomplete reporting, and questionable data quality complicate conclusions about the country’s…

German register exposes utilisation and data gaps
Summary
  • Only 319 facilities are currently published in the register, and the Umweltbundesamt says substantial parts of the market appear absent.
  • Around 70% of usable records already reach PUE 1.5 or better, while average electrical-capacity utilisation is roughly 35%.
  • UBA identified reporting anomalies that limit confidence in water, heat, power, and efficiency comparisons.

Germany’s Umweltbundesamt has published its first analysis of the country’s statutory data centre register, finding that many reported facilities already meet future efficiency requirements while warning that the public dataset remains incomplete and contains significant quality problems.

The register currently publishes records for 319 data centres. Operators of facilities with at least 300kW of non-redundant electrical connection capacity are subject to reporting requirements under Germany’s Energy Efficiency Act, although operators do not have to use the public register itself as their publication route.

UBA says that distinction has left the visible dataset substantially short of a complete map of German capacity. Its comparison with other information sources suggests that many sites belonging to large international operators are missing from the public register.

The agency also found inconsistencies inside some submissions. In 12% of records, reported IT electrical capacity was equal to or greater than the connection capacity stated for the complete data centre — a relationship that raises questions over whether building systems such as cooling were properly included. Four per cent reported more recovered heat than total electrical energy consumption.

Those are not minor bookkeeping issues when performance regulation relies on ratios calculated from the same inputs. An incorrect total-energy boundary can make a PUE figure look materially better than the physical facility actually performs.

Efficiency looks stronger than utilisation

After excluding obviously problematic records for its efficiency assessment, UBA found that roughly 70% of listed data centres already achieve PUE of 1.5 or better. That is the level Germany will require from July 2027.

Around 40% already report PUE of 1.3 or better, the threshold applying from July 2030. About 7%, however, report PUE of 2.0 or worse, leaving a significant efficiency gap for that part of the published estate.

The renewable-electricity picture is similarly advanced on paper. UBA says 80% of facilities in the public dataset already account for their electricity consumption using 100% renewable energy, ahead of the statutory requirement applying from January 2027.

Capacity utilisation presents a different picture. UBA calculates that the listed facilities are using only around 35% of their available data centre capacity when measured against electrical connection capacity.

That figure needs care because a data centre is designed with operational headroom, redundancy, customer growth allowances, and peak-load requirements. A facility running at 100% of every design parameter would not necessarily represent good engineering.

Even so, UBA argues that average utilisation at the reported level cannot be explained simply by reserving capacity for peaks. It raises a more difficult infrastructure question: how much grid capacity is attached to facilities that are not currently using anything close to their electrical entitlement?

Grid capacity meets data quality

That issue has immediate relevance as Germany and other European markets face pressure to connect additional high-load sites. A megawatt reserved for a lightly utilised data centre is still a megawatt that network planners have to account for, even where the IT load has not yet materialised.

The register could eventually become a useful tool for comparing claimed new demand against actual use of the installed estate. It is not there yet. The missing sites and questionable submissions identified by UBA make the current dataset unsuitable for treating 319 records as a complete national inventory.

The policy challenge is therefore twofold. Germany has evidence that many reporting operators can already reach forthcoming efficiency thresholds, but it also has evidence that the reporting mechanism needs better coverage and validation.

As the EU builds its own rating system and explores minimum data centre performance standards, Germany’s early experience provides a warning: regulatory metrics become much less useful if facility boundaries, energy inputs, and operational data are not collected consistently.


Stay updated with the latest insights and trends in the data centre industry by subscribing to our newsletter.

← Back

Thank you for your response. ✨