All week, it's been law firms reading from the same two pages. Today, an actual state-regulator body walked into the docket — and drew a line. This is The Data Center Daily. Today: NARUC stakes out a position in RM26-4-000, the water coming out of these campuses that nobody's tracking, and Synergy's hyperscale count. And hanging over all of it — does any single regulator actually see both the grid and the watershed at the same time? Let's get into it. Tap follow so the next episode finds you. Here's National Association of Regulatory Utility Commissioners:
NARUC appreciates the ANOPR disclaiming any intention to assert jurisdiction over distribution interconnections, generation facilities, and retail sales. It is imperative that FERC, in any final rulemaking, make clear that it is affirmatively not asserting jurisdiction over these facilities, or end-use sales which fall squarely within the exclusive jurisdiction of state energy regulatory authorities.
NARUC's initial comments hit the docket in RM26-4-000 — the first time a body that actually speaks for state PUCs has taken a position on FERC's large-load rulemaking, instead of a law firm speaking for a client. And their pitch is mostly defensive. They're thanking FERC for disclaiming jurisdiction over distribution, generation, and retail sales — then asking the Commission to say it again, affirmatively, in the final rule. Right, but read what they're actually fencing off. Retail rate-setting. Cost allocation to retail load. The exact question of who eats the bill when a hyperscaler plugs in. So is NARUC defending ratepayer transparency, or defending state PUCs from ever having to answer for large-load cost allocation? Different fights, and this filing sounds a lot more like the second one. And the stakes are higher than in a normal proceeding — this whole thing traces back to a Secretary of Energy ANOPR under section 403 of the DOE Organization Act. That's the rarely pulled lever that lets the Secretary propose rules to FERC directly. NARUC is drawing the line because the federal side started this one. What I don't see anywhere in here is the gross-load-versus-net-import gap. NARUC talks about the jurisdiction wall, fine — but does it touch how those 42,625 queue requests get studied? Or does it just plant a state flag and walk? We spend a lot of time on megawatts and gallons going in. But on the water side, what actually leaves an AI-scale campus — heat, blowdown, cooling chemicals — and are the permits even built to catch it? Mostly, it's murky — and the permit architecture wasn't built for this load profile. Virginia Mercury's reporting from earlier this month is pretty telling: in Virginia, most data centers are routed into municipal wastewater systems, the same pipes as residential discharge, with limited chemical tracking on what's actually in that water. The main vector is cooling tower blowdown. As cycles of concentration build up, operators bleed off water with corrosion inhibitors, biocides, and scale-control chemistry. In most jurisdictions, that stream hits the municipal system without facility-level constituent monitoring. Per Nixon Peabody's permitting analysis, environmental review of cooling chemicals and spill control is one of the three main permitting pinch points for AI facilities right now, alongside air permits and water conservation — and it's the one developers most often underestimate until late-stage redesign forces the issue. The federal floor, per EPA, is the Clean Water Act and Safe Drinking Water Act framework, but EPA also says implementation is almost entirely state, tribal, and local — so you get a patchwork. Ohio EPA is in a public comment period right now on a draft permit for increased data center wastewater discharge in Washington County. That tells you these decisions are landing at the county and municipal utility level, not getting coordinated regionally. Meanwhile, Data Center Knowledge reported that a proposed Virginia campus requested up to two million gallons per day of water capacity, which makes the lack of granular discharge monitoring even harder to defend. So if the discharge is hitting municipal wastewater treatment, does the plant actually know what the data center is sending it — or is it mostly just seeing volume? Mostly volume. That's the gap. The Berkeley-CLEE California analysis flags it as a structural problem: existing water-use reporting frameworks weren't scoped for the chemical complexity of liquid-cooled AI infrastructure at hyperscale. Next, watch whether state PUCs or environmental agencies start requiring constituent-level discharge monitoring as a condition of interconnection or operating permits. California and Virginia are the places most likely to force it, given the concentration of capacity and the political scrutiny already in motion. From Synergy Research Group:
Over the last three years, quarterly hyperscale operator capex has ballooned, growing by almost 180% and reaching $142 billion in Q3 of this year. This has enabled a comparable 170% increase in the amount of operational capacity added each quarter.
Synergy's Q3 print: 1,297 hyperscale facilities worldwide, nearly tripling since early 2018 — and capacity up more than fourfold over the same stretch. The stat I'd pull out is the gap between those two: the average facility is getting a lot bigger. Quarterly capex hit $142 billion in Q3, up almost 180% since ChatGPT launched. So the per-request load sitting in that interconnection queue keeps going up, even if the request count stays flat. And that's the denominator nobody's pricing. If average campus size is outrunning interconnection study methodology — the gross-versus-net-import thing we keep hitting — then the mis-sizing error compounds with every one of those 1,297 sites. Put that next to the NARUC filing we just covered. Fourfold capacity growth, 55% of it in the U.S. — does any single regulator actually have eyes on what that does to the grid and the watershed at the same time? FERC sees the interstate piece, the states see distribution, EPA sees a discharge permit nobody sized for this. Synergy's measuring 21 operators' footprints, so we're talking operational capacity — energized capacity, not announcements. Which is rarer than it sounds on this beat. If you follow the infrastructure behind AI, try Anthropic Pentagon Watch — a daily briefing on Anthropic’s fight with the DoD over Claude, military AI use, autonomous weapons, and procurement blacklisting. Find it wherever you listen to podcasts.
Links to everything we covered today are in the show notes. If one of those stories deserves a closer look, start there.
That's The Data Center Daily for today. This is a Lantern Podcast.