Veralto's Water Quarter Priced AI's Second Constraint — and the Trade Is Permitting, Not Scarcity

The raise came from both ends of the model. Adjusted operating margin reached 25.6% as volume growth and the Veralto Enterprise System did the work, and free cash flow of $304M funded a $550M buyback that covered 2.4% of shares outstanding as of Oct. 28, 2026. Management lifted full-year adjusted EPS guidance to $4.49–$4.52 from $4.35–$4.43 and now expects free cash flow to convert at more than 105% of GAAP net earnings. Guidance went up on volume and cost work, not on any disclosed data center revenue.

The comparison that matters sits inside a single company. North America Water Quality core sales rose 8.8% in the quarter while that segment's high-growth markets fell 1.1%, with China down high single digits and Latin America down modestly. China is only 6% of company sales and management said China is still not investing, while North American industrial demand — data centers plus the power, semiconductor and mining feeders — is what management names as the driver. Scarcity alone does not create orders; a rule that requires proof of water consumption does.

Xylem is the cleanest cross-check because it sells into the same projects and actually reports a data center number. Third-quarter orders rose 10% organically to $2.3 billion, Applied Water orders rose 15%, and management said data center orders in that segment were up more than 200% year over year, with ending backlog of $5.5 billion, up 13%. Yet revenue fell 3% reported and was flat organically, and full-year adjusted EPS guidance was cut to $5.40–$5.45 from $5.55–$5.70. Data centers are on track to be 2% of Xylem's revenue on a run-rate basis when 2026 closes.

The explanation Xylem gave is worth more than the number: utility customers took longer to commit, industrial capital projects ramped more slowly than planned, and more backlog is now scheduled to ship in later periods, which is why fourth-quarter organic revenue is guided down 1% to 2%. Large metering awards are slated for deployment from late 2027 into 2028. For investors, that dates the payoff: water instrumentation tied to new AI capacity is a 2027–2028 revenue event dressed as a 2026 order book.

The structure matters more than any single rule. Federal policy moved on power: the July 2025 executive order treats a data center project as a facility above 100MW of new AI load, and the Congressional Research Service counted only 21 Army Corps permits issued for data centers between 2009 and July 2026. Water sits one level down, in states and basins, and municipal systems supply 97% of data center water. The binding constraint is therefore not a megawatt or a federal permit — it is a local approval that turns water into a measured, reported operating cost.

The hyperscale and colocation operators are not waiting for a permit to tell them how to save water. Equinix consumed 4,600 megaliters in 2025, 10% more than in 2024, yet holds global WUE of 0.91 and now publishes allocated water withdrawal and WUE for every site that uses water for cooling. It has also moved beyond municipal supply: reclaimed water was 21% of withdrawn water, its Sydney site runs on reclaimed water with upgraded treatment and real-time monitoring, and since 2023 it has avoided evaporative cooling at new builds in highly water-stressed areas. Digital Realty went further, sourcing 45% of its water from non-potable sources in 2025 and improving WUE 15.7% in a year when total water use rose just 3% against 34% portfolio growth. For suppliers, that pattern shifts the money from gallons to measurement and treatment.

The risks are symmetrical and worth naming. Permits can stop a buildout as easily as they slow it, and a permit that blocks a site removes the instrumentation demand along with the megawatts. Closed-loop cooling can make the treatment annuity smaller than the buildout. And the municipal side of the water complex is already struggling: Badger Meter reported flat third-quarter sales of $235.9M with utility water down 3% excluding an acquisition, while Pentair guided full-year sales down about 3% even as its Water Solutions segment posted a 28.7% return on sales, because pool and residential demand offsets industrial strength. Own the industrial offtake, not the water narrative.

The AI buildout's water constraint is real, regulated and priced. But the price sits on monitoring, treatment and reporting, and it is paid to companies whose revenue recurs with operating water rather than with construction. Veralto is the cleanest listed expression: Water Quality core sales up 6.4%, North America up 8.8%, free cash flow of $304M, and a data center line that doubled off a base too small to disclose.

The honest sizing matters more than the story. Data centers will be 2% of Xylem's revenue exiting 2026 and an undisclosed, similarly small share of Veralto's, so the trade to own is industrial water revenue, with the data center as the option riding on it. Expect the first hard evidence of conversion in 2027–2028 rather than in the next print, and expect the municipal half of the water complex to keep disappointing until data center bills move from connection costs into rate base.