How Much Water Does AI Use? The Hidden Cost of Green Computing

How Much Water Does AI Use? The Hidden Cost of Green Computing

As I observe the world's frantic sprint to power its insatiable demand for Artificial Intelligence with green energy, I’ve uncovered an invisible crisis brewing beneath the surface: the surprising and escalating thirst for ultra-pure water. This isn't just about cooling data centers, a direct water footprint that often grabs headlines. What I've found is a far more massive, yet largely overlooked, indirect water consumption embedded deep within the green energy supply chain, particularly for producing green hydrogen and ammonia. This hidden dependency, I believe, could significantly derail ambitious decarbonization targets by 2030, impacting everything from the operational efficiency of data centers to global food security. It’s not solely the energy grid that presents a bottleneck; it’s the fundamental resource for the cleanest fuels.

The Paradox of Purity: Water for Green Hydrogen’s Thirst

My research shows that green hydrogen, produced via electrolysis powered by renewables, is indeed lauded as a cornerstone of the clean energy transition. It’s critical for decarbonizing heavy industry, transportation, and, yes, even providing stable power for AI’s burgeoning data centers. However, this seemingly clean process demands water of astonishing purity. Electrolyzers, especially the efficient Proton Exchange Membrane (PEM) units, require ultrapure, reagent-grade water, with conductivity often targeted at less than 0.1 µS/cm. Some specifications I've seen even push this to as low as 0.055 μS/cm. This isn't just filtered tap water; it’s water akin to what’s used in semiconductor manufacturing or pharmaceuticals, demanding rigorous treatment to remove impurities like dissolved salts, organic substances, and particles.

The stoichiometric minimum for producing 1 kilogram of hydrogen is around 9 liters of water. Yet, in real-world applications, purification losses, cooling, and other auxiliary processes dramatically push the actual consumption much higher. I’ve seen analyses suggesting actual consumption ranges from 15-25 liters per kilogram of hydrogen. When considering desalination for input, a kilogram of hydrogen can require 35 kilograms of desalinated water due to cooling needs. What I've discovered from CSIRO’s work in Australia, for instance, is that if the feed water is less pure, such as saline, coastal, or brackish, the water needs can rise dramatically to around 100 liters per kilogram of hydrogen, with one seawater project even requiring 240 liters per kilogram. For wind-based PEM electrolysis, I found estimates of 52 liters per kilogram. This is a significant volume.

The global green hydrogen market is projected to surge from USD 12.31 billion in 2025 to USD 17.28 billion in 2026, and then to a staggering USD 231.32 billion by 2035, expanding at a compound annual growth rate (CAGR) of 34.09% from 2026 to 2035. Other reports I've reviewed show slightly different, but still explosive, growth, with estimates of USD 11.4 billion in 2025 reaching USD 173.5 billion by 2035 at a 31.2% CAGR, or even USD 2.79 billion in 2025 growing to USD 247.26 billion by 2035 at a 56.7% CAGR. Regardless of the exact figures, the demand for this specialized, ultrapure water will undeniably skyrocket. Major players like Electric Hydrogen, which achieved unicorn status in three years and opened a 1.2 GW annual production gigafactory in Devens, Massachusetts, are driving this expansion, backed by investors like Microsoft and United Airlines. Linde plc is also a significant force, building a 100-megawatt renewable hydrogen plant for Shell in Germany and supplying a $4 billion low-carbon ammonia facility in Louisiana.

The Desalination Dilemma and the Brine Bomb

Meeting this escalating demand for ultrapure water, particularly in arid regions rich in solar and wind resources ideal for green hydrogen production, often necessitates desalination. While modern seawater reverse osmosis (SWRO) technologies have become more energy-efficient, consuming typically 2.5-4.0 kWh per cubic meter, and sometimes as low as 1.83-2.67 kWh/m³ in advanced projects, the true challenge lies elsewhere: brine. For every tonne of fresh water produced by desalination, roughly a tonne of highly concentrated brine is generated. This brine, laden with salts and chemicals, is toxic to marine life, and its disposal poses a significant environmental threat.

In 2020, approximately 1.6 billion cubic meters of brine were produced globally, primarily from freshwater desalination. I anticipate the rapid scaling of green hydrogen and ammonia projects, many planned in water-stressed coastal areas like Chile, Australia, and Saudi Arabia, will increase this figure several-fold. For example, the NEOM Green Hydrogen Company (NGHC) in northwest Saudi Arabia, an $8.4 billion joint venture with partners like Air Products and ACWA Power, aims to be the world's largest utility-scale green hydrogen and ammonia facility by the end of 2026 or 2027. This mega-project, which expects to produce 600 tonnes of green hydrogen per day, relies entirely on seawater desalination. While NEOM has ambitious plans for zero brine discharge, converting it into industrial materials, the sheer scale of such operations emphasizes the monumental challenge of managing this byproduct.

Australia, too, is seeing major green hydrogen projects like the Cape Hardy project in South Australia, which plans for a massive desalination plant to supply its 5 million tons per year hydrogen production target. Even pilot projects, such as Yarra Valley Water’s $1.7 million initiative at its Aurora treatment plant in Melbourne, highlight the growing need for dedicated water infrastructure for green hydrogen.

Green Ammonia's Water Footprint and the Food Security Nexus

Beyond green hydrogen, I've also examined the water demands of its derivative: green ammonia. Ammonia is a crucial hydrogen carrier and a vital component in fertilizers, directly linking green energy ambitions to global food security. My findings indicate that green ammonia production presents a distinct water challenge. While the direct water usage for electrolysis in green ammonia production is lower than for hydrogen alone, ranging from approximately 1.58 to 2.45 tonnes of water per tonne of NH3 for electrolysis and process needs, these figures still represent a substantial demand for ultrapure water. When desalinated water is used, I found an estimate of 1.6 tonnes of water per tonne of ammonia.

In contrast, traditional Haber-Bosch ammonia production, which dominates the market, typically consumes between 1,500 to 3,000 liters of water per metric ton of ammonia. While green ammonia offers a significant reduction in carbon emissions, the overall water footprint remains a critical consideration, especially when compared to the 18 tons of water per ton of NH3 required for conventional grey ammonia, which includes demineralized, cooling, and utility water. If we consider blue ammonia, which incorporates carbon capture and storage (CCS), I've learned that the CCS process itself can add 2-5 tons of water per ton of CO2 captured for cooling and steam generation.

The geographical overlap of high renewable energy potential and water scarcity creates a direct conflict with other vital water uses, such as agriculture and drinking water. In Chile, for example, a country with immense potential for green hydrogen, I've seen reports highlighting the risk of distribution conflicts if export-oriented green hydrogen projects are prioritized over local needs for drinking water and agriculture. This creates a complex water-energy-climate nexus, where the pursuit of decarbonization could inadvertently strain already limited freshwater resources and impact food production in vulnerable regions.

Innovations and the Path Forward: Brine-to-Value and Beyond

I believe that addressing this hidden water cost requires a multi-pronged approach. One promising area I've explored is advancements in brine management. Technologies like Zero Liquid Discharge (ZLD) and Minimal Liquid Discharge (MLD) are gaining traction. ZLD systems aim to maximize water recovery and eliminate brine disposal, although they are currently expensive and energy-intensive. MLD, on the other hand, reduces brine volume and recovers some water, offering a more cost-effective alternative while still requiring some brine disposal. Companies like NuWater are developing advanced technologies, including 'de-supersaturation' modules, that can achieve water recovery rates as high as 95% from brine.

I've also seen a growing focus on "brine-to-value" initiatives, which seek to extract valuable salts and minerals from the brine, transforming a waste product into a resource. The NEOM project, with its zero brine discharge ambition, is a prime example of this. Furthermore, integrating water treatment with renewable energy systems, as seen in Iberdrola's projects in Spain, or using recycled wastewater, as Plug Power is doing in California, are innovative solutions that I believe can reduce the strain on freshwater sources.

What This Means For Investors/Entrepreneurs/Professionals

For investors, I see a clear opportunity in companies at the forefront of sustainable water management technologies. This includes firms specializing in advanced desalination, ZLD/MLD systems, and brine valorization. Look for companies like NuWater, or those mentioned in reports on ZLD market growth, which are developing cost-effective and energy-efficient solutions. The demand for these solutions will only grow as green hydrogen projects scale. Additionally, I believe investing in companies like Bloom Energy, which is partnering with Oracle to power AI infrastructure with green hydrogen, offers exposure to the direct intersection of AI and green energy.

Entrepreneurs should identify niches in the green hydrogen and ammonia value chains where water efficiency is paramount. This could involve developing novel pre-treatment technologies for diverse water sources, creating modular and decentralized water purification units for remote green hydrogen sites, or innovating in resource recovery from brine. I also see potential in consulting services that specialize in water-energy nexus planning, helping large-scale projects navigate regulatory complexities and ensure social license to operate, especially in water-stressed regions like Chile.

Professionals in engineering, environmental science, and policy development will find themselves in high demand. Engineers with expertise in water treatment, membrane technology, and process optimization for green hydrogen and ammonia production will be critical. Environmental scientists will be essential for conducting impact assessments and developing sustainable brine disposal or reuse strategies. Policy professionals will play a vital role in shaping regulations that balance ambitious decarbonization targets with the imperative of water security and local community needs. I believe cross-disciplinary collaboration will be key to unlocking the full potential of green computing without incurring an unacceptable environmental cost.

Bottom Line

The hidden water cost of green computing, particularly for green hydrogen and ammonia production, represents a significant challenge that I believe is being underestimated. While the shift to renewable energy is crucial for AI, ignoring the massive water footprint embedded in its supply chain risks exacerbating global water scarcity and undermining the very sustainability goals we aim to achieve. I contend that a concerted effort towards water-efficient technologies and integrated resource management is not just an environmental necessity, but a critical economic imperative for the future of AI and green energy.