One of these costs has an owner

Four hundred and seventeen to three.

That was the House vote on 16 September on the Ratepayer Protection Act, H.R. 9340. In a legislature that struggles to agree on the date, 417 members agreed that data centers and other very large electricity users should pay for the grid they make necessary, rather than spreading the cost across everyone else’s meter.

The vote is worth reading twice, and not for the reason it was reported.

What 417 to 3 actually bought

The bill does not require a single state to charge a single data center a single dollar. It amends the Public Utility Regulatory Policies Act of 1978 to require states to consider adopting a standard: that loads of 100 MW and above cover the full, incremental cost of the generation, transmission and distribution upgrades they trigger, provide financial assurances before those upgrades are built, and remain on the hook for recovery if they leave the contract early. States have one year to begin considering it and two years to decide. They may decide against it.

Its sponsors, Gabe Evans and Kathy Castor, were explicit about this. The bill, in Evans’s words, gives states the flexibility to determine what works best. Analysts do not expect the Senate to move it before November.

So the most lopsided vote of the season produced an obligation to think about assigning a cost. That is not a criticism of the bill. It is a fairly precise measurement of how far a cost has to travel before it becomes somebody’s line item, and it is the part of commercialization that technical teams consistently underestimate.

The water version

Two weeks earlier, on 3 September, the House Energy and Commerce Environment Subcommittee held a hearing with a title that could have been written for this argument: Reliable Water in the Digital Age: Examining Legislation to Strengthen Drinking Water Systems and Protect Ratepayers.

One of the drafts under discussion, the Water Cost Accountability Act, would stop public water systems from recovering the cost of a data center connection or the expansion it requires from their other customers, and would place that cost with the operator instead. It would also require EPA reporting on the water sources used by facilities above 200,000 gallons a day. Chairman Brett Guthrie has described it as an important step toward safeguarding the country’s drinking water.

It is a draft. It has not had a floor vote. The electricity version, which has, only obliges states to consider the question.

The underlying volumes are not in dispute. Water consumption by US data centers went from roughly 21 billion liters in 2014 to about 66 billion liters in 2023, with hyperscale facilities alone projected at 60 to 124 billion liters by 2028. Nobody at that hearing argued the water is not being used.

A sixteen percent response rate

What is in dispute is who knows anything precise about it.

In June, the Texas Water Development Board told the state House Natural Resources Committee that its planning work on data centers rested on a 16 percent response rate from operators, with the remainder estimated through what the board described as forensic accounting. A representative asked, on the record, whether the state water plan was really being built on a 16 percent response rate. The answer was yes.

Texas is not an outlier for being careless. It is an outlier for having counted at all.

Why this decides whether anything sells

In my book, Commercializing Clean Technology, I describe buyer risk asymmetry: the buyer’s cost of doing nothing is the number that decides the deal, and it appears nowhere on the supplier’s datasheet. Cost allocation is where that number comes from.

When a cost is assigned to a named party, three things happen more or less at once. Someone acquires a budget line. Someone acquires a reason to prefer a cheaper future to an expensive one. And someone acquires a reason to answer a supplier’s email. A market is, among other things, an administrative arrangement about who carries which cost.

When a cost is acknowledged but not assigned, none of that happens. The problem is real, the public agrees it is real, conference panels are convened about it, and the person you are selling to still has a cost of inaction of exactly zero. Zero is a difficult competitor. It never asks for a discount, never requires a reference plant, and never has to justify itself to a board.

I have spent thirty years selling water and environmental technology, and this is the shape underneath most of the stalled projects I have seen. The pilot performed. The payback was credible. The person in the room simply had no obligation that your equipment discharged. Every commercial instinct says to respond by improving the evidence. The evidence was rarely the problem.

What this changes for a supplier this quarter

Three things are worth doing while the question is still open.

Find out which of your buyers is about to acquire an obligation. If H.R. 9340 or something like it survives the Senate, every state utility commission begins a proceeding, and the 100 MW threshold becomes a date in somebody’s plan. The water equivalent will follow the electricity one, because it always does. Suppliers who know which regulatory proceeding creates their buyer’s obligation are in a different conversation from suppliers who know their own recovery rate.

Write the cost of inaction into the proposal. Not as a scare paragraph. As a number, with the mechanism that produces it and the date it starts. If no such mechanism exists yet, that is worth knowing before the forecast is built rather than after.

Watch the disclosure requirements, not only the standards. A reporting obligation for facilities above 200,000 gallons a day is not a cost. It is the instrument that makes a cost calculable, and calculable costs are the ones that eventually get assigned. In Texas the absence of that instrument is why the planning numbers rest on a 16 percent sample.

Europe did the electricity half already

Ireland’s energy regulator decided in December 2025 that large energy users connecting at 1 MVA and above must bring their own generation or storage capacity, with dispatchable capacity matching site demand at 10 MVA and above, and must source most of their electricity from additional Irish renewable generation. Different legal instrument, same principle: the load carries the infrastructure it creates.

I am not aware of a jurisdiction that has done the equivalent for water. If one has, I would like to be corrected, and so would a number of people trying to build businesses on the assumption that it is coming.

Technology works. Markets do not always adopt it. Quite often the missing piece is not a proof point but an administrative decision about whose invoice a cost belongs on, taken by people who have never heard of your company.

A cost with an owner becomes a market. A cost without one stays a topic, and topics do not sign purchase orders.

---

Legislative provisions are described as recorded in the reporting and committee material cited below rather than from the enrolled text. H.R. 9340 had passed the House but not the Senate at the time of writing, and the Water Cost Accountability Act was a discussion draft. The argument that cost allocation, rather than technical evidence, is the binding constraint in stalled clean-technology sales is the author’s analysis and is not a claim made by any source cited.