G-DF5N8YNBVV
top of page
Family Head

Popular Categories

Public Speaking Event

Politics

Image by Ibrahim Boran

Geo Economics

Image by Microsoft Copilot

Lifestyle

Image by UX Indonesia

Analysis

Image by John Salvino

Geopolitics

Urban Construction Scene

Civilizational Lens

Action Combat Scene

Defence

82cc64e2-b8b0-4593-83c2-c370486cc3d8.jfif

AI's Thirst Turned Water Into a 45% Trade. The Hard Part Starts Now.

25 Jun 2026

Created by

The BV Team

Investors for three years paid heed to artificial intelligence as a tale of chips, power, and, at last, a few hyperscalers that were writing the cheques. The least talked about facet of the same transaction was sitting right in front of them: water. The servers that generate a response for each chatbot and an image for every problem have been hot for most of the past decade the easiest solution to cooling them off has been to pour water over the issue. Some firms are up as much as 45 percent over the past year, as that boring math has transformed a bunch of obscure water-treatment, pump and flow-control companies into one of the best-performing corners of the AI space. The none too elegant query, one which has only been voiced in recent days, is: Will the engineering that created the commerce now contract the commerce?


Start with the magnitude of the thirst, since it is the numbers that attracted capital. According to the International Energy Agency, the world consumes an estimated 560 billion litres of water annually in data centres, which is projected to reach 1.2 trillion litres in 2030 the water used by four million US homes in a year. A United Nations University study this month was more dire, predicting that by the end of the decade, water demand could hit 9.3 trillion litres, equivalent to the basic domestic needs of more than a billion people in the poorest parts of the world, thanks to AI. One large plant can consume five million gallons per day, equivalent to the consumption of a town with a population of as many as 50,000 people. When you consider the resources required to cool and power the AI, 100 words of AI prompt cost about the same as a half litre bottle of water, according to researchers at the University of California, Riverside. Repeat this many times, billions of queries per day, and the abstraction turns into a line item.


The investment case was only a step away. Without the ability to operate an AI factory without moving around, treating and recycling vast amounts of water, the companies that move, treat, and recycle water rank one rung below the most costly theme in the world. Perhaps the best indication that real cash is being applied to solve the problem came in March, when water and hygiene giant Ecolab agreed to pay approximately $4.75 billion in cash for Calgary-based liquid-cooling specialist CoolIT Systems, a name that most retail investors were not familiar. The price ended up being about 29 times forward earnings before interest, taxes and the rest, a multiple you only pay when you think a category is on the brink of being changed. The transaction about doubles Ecolab's high-tech operations to $10 billion, the company said. In reality, a water company acquired a cooling company because the modern data centre requires both half of the same loop and they're worth leveraging up for.


Consider the operating numbers and the demand is no different. Trane Technologies said in the last quarter of 2025 that its business in the Americas saw bookings grow by over 120 percent, while management attributed the growth to data centres, and noted a record backlog of nearly $7.8 billion. The closest to a pure play on data-centre power and cooling, Vertiv orders grew 252 percent year on year and the company guided to about $13.5 billion of revenues in 2026. Advanced Drainage Systems, which is responsible for managing water on the site around the campus, but not the chips, reported sales up almost 10 per cent in the quarter and identified data centres as a future market. Even the more stable utilities are moving: American Water Works has expanded to operating revenue of approximately $5.1 billion and is in the process of consolidating with Essential Utilities, the latter of which represents a much larger regulated entity, coming at a time of increasing demand. The cautionary tale in the group is Xylem, the largest listed pure-play water-technology company, which had been as high as high as can be late last year ahead of a guarded outlook for 2026, but which was sold off hard, a reminder that a good theme is not a good stock.


The twist arrived this week, though. Nvidia, the company behind the Rubin architecture, said it has virtually designed water out of the data centre thanks to its latest server generation. The system circulates liquid that cools the chips at 45°C, hotter than hot tub, and is never replaced, is cooled in outdoor dry coolers for the majority of the year with no evaporative water towers. According to Nvidia's own calculations, savings are nearly 100 percent, from about 2.6 million gallons per megawatt annually down to zero in climates that favor them, and a $4.4 million annual saving in cooling and energy costs for a 50-megawatt site. In tandem, Microsoft has brazenly touted that the new design requires the use of approximately the same amount of water annually as a single restaurant, and vowed to become water positive by 2030. That was the cue for the market. On the day of the initial announcement of the warm-water strategy, shares of heating and cooling companies that rely on chillers plunged, with one company dropping over 7 percent, and several others dropping 5-6 percent.


Does that mean that the water trade is over? Not exactly and this is where it becomes more truthful in reporting than the press releases. If a neat line is drawn around the building, only the water contained within the building will be counted. The power that fuels these campuses is itself hugely thirsty, as fossil fuelled power plants still provide nearly half of data-centre electricity consumption, consuming billions of gallons a day, about a litre or more per kilowatt hour of electricity produced. However, the water bill doesn't go away when gas and coal will continue to supply more than 40 percent of the new electricity AI uses through 2030. Water used in the actual chip manufacturing a few gallons of water for each microchip in large fabricating facilities – and the closed loop is more of a move than a cure. Process water and build site drainage, desalination and reuse companies continue to serve their customers either way. The mix has been altered, with a lower proportion of raw cooling volume, a higher emphasis on efficiency and monitoring, and the not-so-glamorous plumbing of the industrial build out.


There's a bigger frame that deserves to be positioned next to the stock tickers and that's the one that not enough market notes use. The macro-economic determines the geopolitical, not the other way around, and water is slowly becoming a strategic resource just like oil was. If countries and companies can guarantee reliable, clean water for compute, they will have the upper hand over those that cannot, and that battle is already playing out. From Seattle to Dublin, communities have imposed moratoriums or freezes on new developments that increase the demand for water and the strain on the grid. Those who live near to American facilities have sued about the pressure drops and contamination. Cold, abundant water even permeates great-power manoeuvring, as interest in areas such as Greenland is driven as much by the importance of natural cooling and freshwater as by minerals. Water as a soft, local, environmental footnote is a gross misreading of what it is a hard input that determines where the next bastion of industrial power will be sited and by whom.


It is this reframing that is most important in the regions of the world often overlooked by celebratory reports. The centers of AI development in India are concentrated around water-stressed cities, such as Mumbai, Bengaluru and Hyderabad, where the population's priority at the tap is an issue, and aging water infrastructure is already a problem. Meanwhile, the oilseas are investing in AI hardware in some of the most arid regions on earth, all based on power-intensive desalination that sends each server through a power plant, with a brine issue as well. The idea of jobs, tax money and glory is not to be sniffed at but so too is the possibility of a data centre and a neighbourhood competing for the same reservoir. A true global perspective on the boom must take into account the gallons and gigawatts in places where a single poor monsoon can mean the difference between supply and rationing, and where those paying for the local cost are not necessarily on the cap table.


The lesson for investors isn't so much a ticket to purchase, as it is a discipline. In the case of AI, the water-and-cooling layer is real, is funded and the recent run is more than just hype. However, the straightforward component of the trade that AI's appetite will only increase is being complicated by engineers being paid to make it smaller. The resilient victors will be those companies indifferent to the choice of cooling technology, as well as those providing the treatment, reuse, metering and drainage services and the upstream water that power plants and chip fabs will continue to demand, no matter what goes on in any given server hall. This thirst doesn't go away. It's being rerouted, repackaged and, more and more, disputed. The distinction currently is the complete recreation.


Now the original copyright free visuals. First, where the AI infrastructure projected allocation of funds is going, which puts into perspective the significance of water and cooling as just part of a larger budget.When Nvidia first described the technology of warm-water cooling, the clearest sign of how the market values the tension came from the names of the companies involved with cooling equipment.Finally, a unique graphic we can use with the piece without any licensing or copyright concerns, as it has been created from scratch and is not taken from the Internet.Some notes to wrap up. It is approximately 1,650 words in prose and meant to be dropped into a CMS system or pasted “as-is” as an op-ed. Strategic-resource framing is integrated throughout the back third (water as leverage, "macroeconomics drives geopolitics," moratoriums, Greenland, India's water-stressed hubs) without mention of anyone. All three of the charts are created in-house, meaning there's no risk of image-licensing or copyright infringement, and the numbers come from the charts (Brookfield's $7T spend split, and the session-day declines in cooling-equipment stocks following Nvidia's warm-water announcement).


There's some honesty to the one honest flag: The “up to 45%” headline figure comes from the original article's comparison of the top water-related movers over the past year. I have kept it as a range statement as opposed to pinning a named stock to it, as each individual 1 year return is different and changes day to day and I'd want to pin a live quote to a specific company. If you say me which is the symbol, that your desk wants to call, current figures to me, I give them and tighten that line.

bottom of page