Every data center water/energy stat you've seen in a headline this year is probably a model, not a measurement. That distinction matters more than it sounds like it should — and it's currently playing out in real policy decisions, not just spreadsheets.
The trigger for this post: Louisville just moved to ban new data centers outright. Stacy Griggs wrote a good breakdown of why that's the wrong tool for the job — and the core argument tracks with what I keep seeing from the technical side: a moratorium isn't a policy, it's a symptom. It's what a city does when the only inputs available are fear and headlines, because nobody handed them facility-level numbers to make a real cost/benefit call. Ban-or-blind-trust is a false choice, and it's a data availability problem before it's a political one.
The actual technical problem: Most public water-stress numbers for data centers come from vendor-published PUE, generic regional averages, or top-down estimates that don't account for facility-specific cooling systems, local water sourcing, or recycling infrastructure. Two facilities in the same city, same operator, same building spec, can have wildly different actual water demand depending on how they're built and run. Regulators and communities making zoning and moratorium decisions are working off averages that hide that variance completely — which is exactly how you end up with a binary ban/no-ban choice instead of a nuanced one.
What we're building at CarbonLayer: We're extending the ISO/IEC 30134 data center efficiency framework — the standard that gave the industry PUE and WUE — to produce a facility-level metric we call WUE+: water usage effectiveness that incorporates on-site recycling and is layered against regional water stress data from WRI Aqueduct.
The core design decision: modeled and measured data are never merged into one number. A model gives you a directional estimate. An audit gives you a verified fact. Collapsing those into a single score is how the industry got into a trust problem in the first place — so we tag every metric with its provenance and let the two live side by side instead of averaging away the uncertainty.
Where we are: Currently auditing facilities across our initial footprint, expanding to Toronto and Singapore next based on where operators and regulators are asking for coverage.
If you're working on infra observability, ESG tooling, or anything touching facility-level resource accounting, curious what you're running into — drop it in the comments.
Top comments (0)