Welcome to Backchannel. This is the weekend briefing for subscribers of The Closer, where we decode power plays in dealmaking, business, and influence. This week, we’re diving deep into an overlooked yet critical resource reshaping the future of AI and global tech infrastructure: water.

Every week, we publish a Jumpstart edition on Monday, three days of Negotiation Nuggets, and our bumper Backchannel edition on Fridays. It’s part of our broader mission: to build the most unflinching, high-signal analysis of global dealflow and strategic moves.

If you’re already a paid subscriber: thank you. You’re helping raise the bar for serious, independent business journalism. If you’re not yet in, now’s the time to join us.

And as always, please take a moment to let me know what we can do better or what you’d like to see more of—I deeply appreciate your feedback.

– Bradley

Modern computing infrastructure—particularly the data centers driving cloud services and artificial intelligence—quietly depends on one of the planet’s most precious resources: water. From cooling massive server farms to sustaining the electricity that powers them, water acts as an essential yet frequently overlooked backbone of the digital economy. As AI-driven computational demands soar, so too does the industry’s reliance on increasingly scarce water resources, setting the stage for critical conflicts where technological advancement intersects with environmental constraints and financial speculation.

Increasingly, these conflicts are being shaped not only by physical scarcity but also by the shadowy, high-stakes market for water rights. Private equity funds, institutional investors, and major endowments have started strategically acquiring farmland, aquifers, and senior irrigation rights, betting heavily that the growing water needs of data centers will drive up the value of secure water supplies. Entities such as Harvard University’s endowment and private investment firm Greenstone are already deeply embedded in controversial water rights acquisitions, leading to high-profile disputes and litigation, such as Harvard’s groundwater battles in California and Greenstone’s profitable transfers of Colorado River water rights in Arizona.

Below, we explore how water and computing infrastructure are deeply intertwined, highlighting specific flashpoints emerging from this crucial intersection. Companies like Oracle and OpenAI are responding by pursuing ambitious and innovative solutions—including landmark projects like Stargate in the U.S. and UAE—aimed at mitigating their dependence on volatile water rights markets. Grounded in real-world examples and global business trends, we unpack near-term risks, outline strategic shifts toward recycled water and zero-evaporation cooling technologies, and assess how the battle for water rights is reshaping the geography, economics, and future trajectory of the AI industry.

Water: The Hidden Lifeblood of Modern Computing

Running massive data centers requires expelling huge amounts of heat, and water is a critical coolant in most large-scale facilities. Instead of traditional air conditioning alone, many data centers use water-based cooling systems for efficiency. A typical 100 MW data center in the U.S. uses about 2 million liters of water per day for cooling – enough water for a town of tens of thousands. On a global scale, data centers currently consume over 560 billion liters of water each year, a number projected to double to ~1.2 trillion liters by 2030 as demand grows.

Why so much water? Large liquid-cooled servers circulate water (often potable water free of minerals) through heat exchangers to absorb heat from CPUs and GPUs, then evaporative cooling towers release that heat – but in the process, water is lost as vapor . Most hyperscale data centers still rely on such open-loop evaporative cooling, where water is “consumed” (evaporated) to carry away heat. Closed-loop systems exist (circulating water in sealed pipes akin to a car radiator) that use far less water – but these are more expensive and thus less common. In short, every time you stream a video, run a cloud AI model, or even just store files online, there’s likely water evaporating behind the scenes to keep those servers cool.

Importantly, water use in data centers varies widely by location and design. Cooler climates can rely more on outside air cooling, using water only on hot days, whereas facilities in hot, arid regions may need water year-round. For example, Microsoft’s Iowa data centers take advantage of the cool Midwest climate – they use outside air for much of the year and only start water-based cooling above ~85°F (29°C). Even so, on a hot month (July 2022) Microsoft’s Iowa facilities still pumped 11.5 million gallons (43.5 million liters) from local water works for cooling, about 6% of the area’s total water use. In contrast, a data center in a desert climate might draw water far more regularly.

The AI Boom’s Growing Thirst

The recent explosion in artificial intelligence workloads – from training giant models like GPT-4 to powering millions of AI-driven queries – has dramatically amplified the tech industry’s water usage. AI accelerates demand for dense compute, which in turn generates more heat and requires more cooling. Industry experts are now warning that AI’s “thirst” for water may soon rival or surpass other famously water-intensive industries like cattle farming and textile production . This is not hyperbole – the numbers bear it out.

Consider OpenAI’s ChatGPT and GPT-4: these models were primarily trained in Microsoft’s data centers, and the training of GPT-4 (completed in 2022) consumed a massive amount of water drawn from the local Iowa watershed to cool the supercomputers. Microsoft disclosed that its global data center water consumption jumped 34% year-over-year from 2021 to 2022, reaching ~1.7 billion gallons (over 6.4 billion liters). Google saw a similar trend, with a 20% jump in water use in the same period (2021–2022), which the company also attributed largely to AI workloads. Notably, Google’s water use wasn’t uniform: in relatively wet and cool Oregon, their data center water draw held steady, whereas at a campus near Las Vegas (an arid region), water consumption doubled in a year – again highlighting how AI growth is straining water supplies most in dry areas.

Even answering your ChatGPT queries has a water cost. Researchers estimate that ChatGPT “drinks” about 500 milliliters of water (roughly a 16-ounce bottle) for every 5 to 50 prompts you feed it. This estimate includes both direct cooling water and the indirect water used by power plants to generate the electricity for the computation. While 500 mL per few dozen questions may not sound like much, the scale of usage adds up fast – millions of users querying AI bots daily translates to many thousands of gallons evaporated. In aggregate, AI inference and training are becoming significant contributors to data centers’ water footprint.

Crucially, these water demands are immediate – unlike some other environmental impacts, water withdrawals happen in real time and often in the communities hosting the infrastructure. In West Des Moines, Iowa, where Microsoft operates a cluster of AI data centers, the local water utility has already put Microsoft on notice: future expansions will only be approved if new technologies are implemented to “significantly reduce peak water usage” from current levels. This came after those data centers alone accounted for 6% of the region’s water use, prompting concerns about sustaining municipal supply for residents. Microsoft has responded by working with the utility on conservation measures and touting plans to be “water positive” (returning more water than it consumes) by 2030. OpenAI, for its part, has acknowledged that “training large models can be energy and water-intensive” and says it is striving for greater efficiency in its operations.

The story is similar across the tech world. Almost every major cloud/AI player is reporting record water usage growth, but also making new promises to curb it. Google, for instance, has pledged to replenish 120% of the water it consumes by 2030 (adding back more water to communities than it withdraws). Amazon’s AWS, which operates hundreds of data centers, announced it will be water positive by 2030 and has begun expanding the use of recycled wastewater for cooling (so that treated sewage, not drinking water, cools its servers). Microsoft and Meta have likewise committed to net positive water or replenishment goals by 2030. Achieving these goals will require major investments in efficiency and reuse – highlighting how water has become a strategic sustainability focus alongside carbon in the era of AI.

Flashpoints: Where Compute Meets Water Scarcity

As the compute industry’s water appetite grows, conflicts are emerging in water-stressed regions. The “frontlines” of this battle are places where new or expanded data centers collide with dwindling water supplies. Here are some prominent flashpoints around the world:

• American Southwest (Arizona, Texas, California): These areas offer cheap land, favorable taxes, and abundant solar energy – making them magnets for new data centers – but they are also some of the most water-scarce regions. A recent analysis found that two-thirds of new data centers built since 2022 are in locales with high water stress, notably California, Texas, and Arizona. In Texas, which is experiencing a data center boom, officials are increasingly anxious about water. Governor Greg Abbott noted that ensuring a stable water supply during droughts is “critical” for the state’s path forward in attracting AI data centers. Industry experts warn that “it’s water cooling that is needed to get these [AI data centers] to function effectively – and there are issues.” One stark example: In Dallas, plans for the huge Stargate AI center (discussed later) raised questions about whether the region’s grid and water systems could handle the load. State senators likened the unrestrained growth to a “Wild West” scenario and considered requiring data center operators to supply their own power and water solutions so as not to crash the public utilities. Meanwhile, Loudoun County, Virginia – nicknamed “Data Center Alley” – has seen the freshwater use of data centers jump 250% since 2019, contributing to “drought-like conditions” in some areas and prompting public calls for oversight. This is notable as Loudoun is home to the largest concentration of data centers in the U.S., illustrating that even historically water-rich regions feel the pinch when local clusters grow dense.
• Latin America (Uruguay and Chile): In 2023, Uruguay became a dramatic flashpoint. The country was in the grip of its worst drought in 74 years, reservoirs were dry to the point that the capital’s taps began running with brackish water. In the middle of this crisis, officials welcomed a new 72-acre Google data center project. It soon emerged that Google’s facility would require 2 million gallons of water per day for cooling – equivalent to the daily water use of 55,000 people in Uruguay. This news sparked fury: protests erupted across Uruguay against the data center plans. Citizens questioned how, in a water emergency that had people drinking salty water, a tech giant could claim such a huge share of precious freshwater. The public pressure was so intense that Uruguay’s government delayed and re-evaluated the project. The situation underscores the PR and political risk for companies when water needs clash with public well-being. Likewise, in Chile, residents of water-stressed suburbs of Santiago learned of plans for new Google data centers and immediately voiced strong opposition. The backlash grew to the national level – the President of Chile faced condemnation for supporting those projects amidst local water shortages. Ultimately Chilean courts intervened: environmental concerns led to the revocation of Google’s data center permits in those areas, pending rigorous water impact assessments. These Latin American examples show that in regions where water scarcity is dire, data centers can become a lightning rod for public anger, forcing governments to hit pause on tech expansion.
• Asia (India): The fast-growing digital economies of Asia are not immune. In India, for example, the city of Bengaluru (Bangalore) is emerging as a data center hub – and it also nearly ran out of water a few years ago. Bengaluru’s existing data centers (~115 MW capacity) already consume an estimated 8 million liters of water per day. The city endured its worst water crisis in centuries recently, and local activists have started to point out the tech sector’s role. There have even been election boycotts in water-deprived neighborhoods of Bengaluru, with residents demanding action on water issues. As big cloud providers plan new campuses there, the tension between “Digital India” ambitions and water security is mounting.
• Europe (Netherlands and Ireland): Northern Europe hosts many data centers due to its cool climate, but water fights have still broken out. In the Netherlands, a high-profile battle occurred over Meta’s proposed giant data center in Zeewolde. Locals and environmentalists objected that it would guzzle groundwater and provide little local benefit. After massive local opposition, the Dutch government revoked the zoning permission for Meta , effectively halting the project. Separately, Dutch media revealed that a Microsoft data center in North Holland had been using four times more water than initially disclosed, even though it was supposed to use water only sparingly on hot days . This came out during a 2022 heatwave when the region was under water-use advisories, sparking outrage. The revelation that Microsoft’s facility had drawn 84 million liters in one year (versus the 12–20 million liters promised) fed public distrust . It has since led to stricter permit scrutiny and demands for transparency in the Netherlands. Similarly, Ireland – another data center hotspot – has at times ordered moratoria on new data centers in Dublin due to stress on power and water; community groups have become increasingly vigilant about any new project’s water plans.

These flashpoints share a common theme: local communities are starting to question whether the benefits of nearby data centers are worth the heavy draw on shared water resources. Data centers often bring jobs and investment, but relatively few jobs once operational, and they always bring a significant resource footprint. When that footprint threatens drinking water supplies or agriculture (as seen in drought-stricken areas), expect more pushback and political intervention. In effect, water is becoming a limiting factor for compute expansion in several parts of the world. As Luke Metzger of Environment Texas put it, referring to the proliferation of AI server farms: “Given the enormous strain they have on our electric grid, our water supplies, our environment, do we need all of these?” That question captures the growing skepticism in some quarters.

Oracle, OpenAI, and “Stargate”: AI Ambitions in Water-Stressed Regions

While conflicts brew, tech companies and governments are also pushing bold new initiatives that by necessity must tackle the water issue head-on. A prime example is the Stargate project – a series of cutting-edge AI supercomputing centers launched by a partnership of OpenAI, Oracle, and others. The first Stargate facility in the U.S. is rising in Abilene, Texas, and a second is planned as Stargate UAE in Abu Dhabi. These mega-projects illustrate both the massive scale of AI compute on the horizon (each cluster is 1 GW+ of compute capacity) and the challenges of siting such infrastructure in water-challenged areas.

• Stargate Texas (Abilene): In early 2025, OpenAI’s Sam Altman and Oracle’s Larry Ellison stood beside U.S. officials to announce a $500 billion public-private effort to build “nation-scale” AI infrastructure, kicking off with a 1.2 GW data center campus in Abilene, West Texas. The facility, spread over 900 acres, will house up to 400,000 of NVIDIA’s latest AI chips and is expected to consume power on the order of a mid-size city. To address electricity needs, the developers are even constructing a dedicated on-site natural gas power plant. But what about water? West Texas is arid, and Abilene has experienced droughts, so this was a key question. Project leads have stated that cooling will use a closed-loop liquid system with recycling, to keep water needs “lower than usual” for a facility of this size. The city government committed to support the project’s water requirements but was reassured by designs that “recycle liquid” and reduce total usage. In practice, this likely means techniques like liquid immersion cooling or chilled water loops that minimize evaporation. If indeed the site can avoid open evaporative cooling towers, it would save millions of gallons annually. However, independent estimates suggest a cluster this large could require on the order of tens of millions of liters of water per day if it used conventional cooling. Clearly, Oracle and partners are highly motivated to deploy advanced cooling here – both to honor sustainability pledges and to ensure local authorities remain on board. Texas officials, from the governor to local lawmakers, have voiced support for Stargate if it can secure its own resources without straining the public grid and water supply. In other words, Stargate Texas is a test case: can a hyper-scale AI data center be made compatible with a water-scarce environment? The developers are betting on technology and engineering to say yes, by using sophisticated cooling infrastructure that significantly shrinks water consumption.
• Stargate UAE (Abu Dhabi): Even more striking is the plan to build a twin “Stargate” in the United Arab Emirates, a country synonymous with desert landscapes and scarce freshwater. Announced in May 2025, Stargate UAE is a 1 GW AI compute cluster to be operated by OpenAI and Oracle in Abu Dhabi. It’s part of a broader UAE–US tech partnership, and will be the first such AI megacenter outside the U.S. The UAE’s motivations are clear: reduce latency for AI services in the region, anchor sovereign AI capability, and signal its ambition to be a global AI leader. But the climate reality is harsh – summer temperatures regularly soar above 45°C (113°F) and freshwater comes almost entirely from energy-intensive desalination. How to cool a gigantic data center here, without guzzling the UAE’s desalinated drinking water supply? The public announcement highlighted that the 5 GW campus housing Stargate UAE will be powered by a mix of nuclear, solar, and natural gas to minimize carbon emissions. However, little was said publicly about water, even though it’s just as critical for sustainable operation. It’s expected that the UAE will leverage its expertise in desalination and potentially use non-potable water sources for the data center. In similar projects, facilities have used treated wastewater (“purple water”) or even directly used seawater for cooling after proper treatment. Indeed, companies like Gradiant (a water technology firm) are expanding in the Middle East to enable exactly this: turning otherwise unusable water – from seawater to industrial brine – into high-purity coolant for data centers. Gradiant’s founders note they are “feedwater-agnostic” and can supply data centers with recycled water so that “we don’t tap into municipal or potable supplies.” The UAE is also exploring cutting-edge ideas, for example underwater data centers in the Gulf. A concept floated by technologists is to submerge modules in the Arabian Gulf waters, using the natural cooling of the sea – much as Microsoft’s experimental Project Natick did off Scotland – to drastically cut or even eliminate fresh water use. While an underwater Stargate is still hypothetical, the fact it’s being discussed underscores the pressure to “think outside the box” for cooling in extreme climates. Regardless, when Stargate UAE goes live (targeted 2026 for initial 200 MW), it will be watched closely as a bellwether: Can a desert nation host a world-class AI supercluster without unsustainable water use?

Liquid Assets: How Water Rights Are Shaping AI’s Expansion

If water scarcity is the hidden constraint on AI’s future, then water rights are the quiet battlefront where tech giants and Wall Street converge.

A growing class of private equity funds, institutional investors, and even university endowments are quietly buying up farmland, aquifers, and senior irrigation rights in water-stressed regions—betting that as the global thirst for compute grows, the market value of secure water supply will surge.

Harvard University’s endowment famously amassed thousands of acres of vineyards in California’s Paso Robles, acquiring extensive groundwater access amid rising drought concerns. Their moves triggered local backlash, accusations of speculation, and legal battles over water-storage permits. Similarly, Saudi-owned Fondomonte drew headlines—and lawsuits—for pumping vast amounts of groundwater in Arizona to grow alfalfa destined for Middle Eastern dairy cows. Such cases underscore the increasingly fraught intersection of global finance, agricultural rights, and resource sustainability.

But the financialization of water isn’t just farmland or agriculture. Hyperscale data center operators like Oracle, Microsoft, and AWS are now forced to reckon directly with this opaque market. In regions like Arizona and Texas, tech giants building AI-focused data centers are finding themselves not merely negotiating land deals and electricity rates but scrambling to secure long-term water allocations from rights-holders. Often these rights are controlled by funds and entities whose very existence hinges on monetizing scarcity.

Take Greenstone, a private investment firm active along the Colorado River in Arizona, which recently flipped senior water rights from farmland in Cibola, AZ, to a fast-growing municipality 175 miles away, grossing more than $10,000 per acre-foot transferred. The transfer sparked a federal lawsuit, highlighting the rising stakes and scrutiny around such transactions. This speculative activity indirectly shapes the economics of where data centers can be sited and at what cost.

As water rights become a hotly traded commodity, tech companies are pivoting toward recycled-water pipelines and zero-evaporation cooling technology—not just as sustainability moves, but as strategic hedges against volatile water-rights markets. AWS, Google, and Oracle, for example, are already funding municipal reclaimed-water projects, effectively locking in supply through infrastructure investment rather than outright rights purchases.

This strategic shift carries deeper implications. If tech giants begin treating water infrastructure as critical internal assets—like server chips or power contracts—it will accelerate water’s transformation from a regional public good into a global strategic resource, controlled increasingly by private capital.

In the emerging AI-driven economy, then, water rights are more than legal abstractions—they’re liquid assets with the power to shape geography, economics, and even geopolitics. As these battles intensify, one thing is clear: whoever controls the water beneath AI’s expansion controls the future itself.

For further reading:

• Associated Press – “AI behind ChatGPT was built in Iowa — with a lot of water” 
• Data Center Dynamics – “Is water the new power?” (Z. Skidmore, June 2025) 
• Winsome Marketing – “Oracle’s $40B Nvidia Deal… Environmental Questions” 
• G42 Press Release – “Global Tech Alliance Launches Stargate UAE” (May 22, 2025) 
• OpenAI Blog – “Introducing Stargate UAE” (May 22, 2025) 
• TechRepublic – “Stargate’s First AI Data Center in Texas: 10 Things to Know” 
• Dallas Observer – “Texas Runs Short on Electricity and Water for AI Centers” (Emma Ruby, 2024) 
• Lawfare – “AI Data Centers Threaten Global Water Security” (Nov 2023) 
• DCD – “AWS using reclaimed wastewater for cooling at 20 locations” (Nov 2023) 
• Techzine – “Microsoft data center guzzles water during heatwave” (Aug 2022)

Top Reads This Week

Amazon buys Bee, the $49 AI bracelet that records your life

Amazon’s Ambient-Intelligence unit scooped up the low-cost wearable startup, adding a life-logging device with an always-on mic and on-device vector database—slotting it neatly into Alexa’s expanding AI ecosystem.

Why it matters: A clear signal Amazon aims to dominate entry-level AI hardware as competitors (Humane, Mozilla, Apple) focus on premium tiers. The privacy debate is already heating up.

Waystar acquires Iodine Software for $1.25 billion

The healthcare-payments giant picked up Iodine’s clinical documentation improvement (CDI) AI, automating medical chart coding and reimbursement. Advent International rolls its equity stake fully into the merged entity, with an 18-month lock-up.

Why it matters: The deal sets a billion-dollar benchmark in AI-driven healthcare revenue-cycle management and puts a clear valuation floor under clinical-documentation AI startups.

Canal+ secures approval for $2.9 billion MultiChoice takeover

South Africa’s Competition Tribunal approved the French broadcaster’s all-cash bid, imposing local-content and employment safeguards.

Why it matters: Creates a 31-million-subscriber pan-African pay-TV powerhouse, positioning Canal+ as a key gatekeeper for lucrative French sports content in Africa’s fast-growing OTT market.

Shein files Hong Kong IPO prospectus to pressure stalled London listing

Fast-fashion giant Shein filed a prospectus in Hong Kong after UK regulators (FCA) failed to sway Beijing, forcing a dual-track race between HKEX and the London Stock Exchange for a potentially £40+ billion float.

Why it matters: A classic venue-arbitrage tactic increasing pressure on London regulators to ease disclosure rules and maintain competitive IPO conditions.

IPO window opens cautiously: McGraw Hill ($3.25B) & Accelerant ($6.4B)

Educational publisher McGraw Hill raised $415 million; insurance marketplace Accelerant raised $900 million. Both IPOs traded flat, opening at their issue prices, reflecting cautious but active investor appetite.

Why it matters: Confirms that IPO markets remain viable for solid, realistically priced offerings—a critical signal for private equity firms eager to exit mature portfolio companies.

Blackstone signals strongest deal pipeline since 2021

Blackstone’s Q2 earnings exceeded expectations, highlighting accelerating M&A activity and a clear line-of-sight on substantial carve-outs exceeding $50 billion.

Why it matters: Blackstone’s optimism is a leading indicator for broader middle-market M&A activity, suggesting substantial LP dry powder is beginning to find attractive deployment opportunities.

🟠 Have a favorite example of scarcity positioning? Hit reply—top story gets a Closer mug and we'll share it in next week's edition and social media.

If you’ve read this far, we want to hear from you! We hope The Backchannel is becoming an essential part of your week – an insider guide to the deals that are changing our world. Ideas, feedback, or just want to chat? Please get in touch. Until next time, keep making moves and thinking outside the box. Deals change the world – and so can you.