The Nobel Prize That Pulls Water Out of Thin Air

There is something strangely inefficient about an AI data centre. Thousands of processors consume enormous amounts of electricity while producing equally enormous quantities of heat, which operators then spend more energy trying to remove. Depending on the cooling system, getting rid of that heat can also consume significant quantities of water. US data centres directly consumed roughly 66 billion litres of water in 2023, while their electricity demand is rising quickly enough that researchers at Lawrence Berkeley National Laboratory estimate the sector could represent close to 12% of total US electricity consumption by 2030 under a central scenario. The industry is therefore producing a resource nobody particularly wants (heat) while consuming another resource that is becoming increasingly difficult to secure: water.

Atoco thinks those two problems may fit together. The California company is developing machines that pull water directly from air using a class of highly porous materials engineered to attract individual molecules. In a recent demonstration in Irvine, warm air designed to replicate the waste heat coming from a small data centre was passed through one of Atoco's systems. Water vapour in the surrounding air became trapped inside the material, the low-grade heat helped release it again, and liquid water eventually emerged from the machine. The current prototype can produce around 300 litres per day, while Atoco is developing an off-grid commercial system targeting as much as 1,000 litres per day and an electrically powered version designed for several thousand litres. The company has already deployed prototype trials with partners in the United States and Gulf countries and says it wants to begin taking commercial orders before the end of 2026.

The idea creates an almost perfect industrial loop. Data centres currently use water partly to manage heat. Atoco wants to use the heat to make water. Instead of treating the thermal energy pouring out of servers as waste, it becomes part of the input required to regenerate the material inside Atoco's water-harvesting system. A facility could potentially reduce part of its dependence on municipal water supplies while finding a productive use for energy it was already trying to dispose of. And because the captured water can be extremely pure, it could potentially be reused inside industrial systems before being treated and mineralised for other purposes.

Yet describing Atoco simply as a company making water from air misses most of what is interesting about it. The same underlying scientific platform is also being developed to capture carbon dioxide, both from industrial exhaust and directly from the atmosphere. The molecule changes, but the principle remains similar: design a material whose microscopic structure selectively attracts something valuable or problematic, then build an industrial system capable of capturing and releasing that molecule efficiently. Atoco's real ambition is not to build a better water machine. It is to make the atmosphere itself usable infrastructure.

A Nobel Prize turned into a startup

Atoco was founded in 2020 by Omar Yaghi, a chemist at the University of California, Berkeley who spent decades helping create and develop the field of reticular chemistry. In 2025, Yaghi shared the Nobel Prize in Chemistry with Susumu Kitagawa and Richard Robson for the development of metal-organic frameworks, usually shortened to MOFs. The award did not suddenly invent Atoco's science; it recognised work that had been developing for decades. What it did do was place the molecular architecture at the heart of the company under one of science's brightest spotlights.

MOFs sound extremely technical, but the core idea is surprisingly visual. Imagine constructing a three-dimensional frame from metal atoms connected by organic molecules, then shrinking the structure to the molecular scale. The result looks solid to us but internally contains enormous numbers of microscopic pores. Those pores create an extraordinary internal surface area. Some MOFs have such large effective surface areas that only a small amount of material can contain the equivalent of an entire football field's worth of internal surface. More importantly, chemists can change the building blocks used to construct them, altering the size and chemical personality of those microscopic cavities.

A MOF is therefore not simply a sponge. It is a sponge whose pores can be designed around the molecule you want to catch. One structure might be particularly attracted to water molecules. Another can favour carbon dioxide. Other frameworks are being studied for storing hydrogen or methane, filtering pollutants, separating chemicals and accelerating industrial reactions. Atoco also works with related covalent organic frameworks, or COFs, giving the company a broader library of materials that can be designed from the atomic level upward.

That explains the company name: Atoco essentially comes from “atom company.” The logic is that industrial systems do not necessarily have to begin at the level of the machine. They can begin at the level of the atom, designing exactly how a material interacts with the world before scaling that chemistry into a device large enough to matter commercially. This makes Atoco fundamentally different from many climate-tech startups. It is not taking an existing industrial machine and trying to improve its efficiency by 10%. It is starting with the molecular interaction itself and trying to redesign the system from there.

There is also a much more personal connection between Yaghi and the company's first major application. Yaghi was born in Amman, Jordan, to Palestinian refugees, and has described growing up in a household where water arrived intermittently. As a child, he would help fill containers whenever the water supply became available because the family might then wait days or weeks before the next delivery. Decades later, his research group began creating MOFs capable of extracting water even from extremely dry air. Researchers eventually demonstrated MOF-based water harvesting in places including Death Valley, proving that appropriately designed materials could collect moisture even when relative humidity was extraordinarily low.

That matters because conventional atmospheric water systems usually behave somewhat like dehumidifiers. Air is cooled below its dew point until water condenses. In humid environments this can work reasonably well. In a desert, where there is very little moisture in the air, cooling enormous volumes of air becomes increasingly energy intensive. Adsorbent materials provide a different approach: instead of cooling everything, selectively capture the few water molecules that are present and then use heat to release them later. The difficulty is balancing those two behaviours. Capture water too weakly and the system collects very little. Capture it too strongly and enormous amounts of energy are required to get it back.

Atoco CEO Samer Taha has described the ideal material beautifully: it has to like water, but not love it. That tiny distinction may determine the economics of the entire company.

AI may have accidentally created Atoco's perfect market

When Atoco began, data centres were not the obvious first application. The commercial opportunity changed as generative AI exploded. Training and running increasingly large models requires enormous amounts of computing power. That produces enormous heat loads, and the facilities containing those processors require increasingly sophisticated cooling infrastructure. At the same time, technology companies are running into another constraint: the communities surrounding new data centres do not necessarily want enormous industrial facilities taking large quantities of local electricity and water.

This has turned something once treated as an operational detail into a strategic issue. Access to water can influence where data centres are built, what cooling systems they use, whether planning permission becomes politically controversial and how credible technology companies appear when making environmental commitments. Companies including Google, Microsoft and Amazon have all made public water-replenishment commitments while simultaneously expanding AI infrastructure. Water is becoming part of the permitting and social licence required to build the next generation of computing.

That makes the sector unusually attractive for Atoco. The people suffering most severely from water scarcity often cannot afford expensive emerging technology. Hyperscale data-centre operators can. They have enormous capital budgets, high costs associated with downtime, strong incentives to secure reliable infrastructure and growing pressure to demonstrate that their facilities are not simply draining resources from surrounding communities. A litre of water used by a data centre is therefore not valued only at the municipal price of that litre. Reliable access to water can help protect billions of dollars of infrastructure.

This creates a clever route to market. Atoco can sell first to customers for whom water scarcity is economically painful, use those customers to industrialise the technology, and potentially drive costs down enough for broader applications later. Tesla once used a very expensive Roadster to begin developing technologies that ultimately moved into cheaper vehicles. Atoco's equivalent could be the AI data centre: an affluent early customer willing to pay for technology before it reaches true mass-market economics.

The other advantage is energy. Atoco says its water-harvesting materials can be regenerated using relatively low-temperature heat, with some systems operating from sources around 38°C. That is important because data centres already produce large quantities of exactly this kind of low-grade thermal energy. Most industries struggle to find economically useful applications for heat at those temperatures. It is too cool for many conventional industrial processes. Atoco is effectively designing a process around a resource nobody else particularly values.

That changes the business proposition completely. The company is no longer merely selling a machine that manufactures water. It is potentially offering customers a way to convert waste energy into a strategically valuable resource. The less electricity required to operate the system, the stronger the economics become.

Atoco has now begun positioning itself much more explicitly around this market. It was recently selected for the 2026–27 Net Zero Innovation Hub for Data Centers after being evaluated by a panel involving organisations including Google, Danfoss and the Technical University of Denmark. The prize itself is insignificant compared with the strategic access: Atoco needs relationships with exactly the companies building and operating the infrastructure where its system could eventually be deployed at scale.

The company is really selling molecular selectivity

Water is only one molecule Atoco wants to control. The second is carbon dioxide, and the science reveals the broader platform hiding underneath the company.

Carbon capture comes in several forms. Post-combustion systems attempt to remove CO₂ from exhaust before it reaches the atmosphere. Direct air capture is much more difficult because ordinary air contains only a little over 400 carbon-dioxide molecules for every million molecules of air. A machine must therefore move through huge volumes of atmosphere to collect relatively small amounts of CO₂. Existing systems often rely on liquid chemicals such as amines that bind strongly to carbon dioxide but can require substantial energy to regenerate.

Atoco is developing solid-state systems based on porous materials engineered to capture CO₂ selectively while reducing some of the energy required to release it afterwards. The company has reported laboratory results suggesting significantly lower energy requirements than some conventional approaches, although these results remain far earlier than commercial-scale validation. Atoco itself acknowledges that its water products are closer to the market than its carbon-capture technology.

Still, the combination is strategically fascinating. Data centres increasingly unable to secure enough grid power are exploring more on-site electricity generation, sometimes using natural gas. That helps solve an energy constraint while creating additional carbon emissions. Atoco can imagine a future where the same facility uses one family of materials to harvest water using waste heat and another to capture CO₂ produced by its power systems.

Suddenly Atoco is not selling a water machine and a carbon-capture machine. It is selling molecular infrastructure around computing.

And the underlying capability could theoretically extend further. MOFs and COFs are being researched for hydrogen storage, gas purification, chemical separation, pollutant removal and other applications. Atoco would probably destroy itself strategically if it attempted to enter all of those markets simultaneously, but the potential matters because it clarifies what the company actually owns.

It does not fundamentally own “water from air”. It owns expertise in selectivity.

Industrial chemistry is full of expensive separation problems. Huge amounts of global energy are spent separating useful molecules from unwanted ones: carbon dioxide from exhaust, oxygen from nitrogen, water from salts, chemicals from complex mixtures. If a material can become much more selective about what it attracts, entire processes can become cheaper.

That may eventually make Atoco's molecular design expertise more important than any one machine it builds.

The economics will decide whether extraordinary science becomes ordinary infrastructure

The scientific story is captivating. The commercial story is less forgiving.

Atoco believes its water systems could eventually produce water at a levelised cost of around $5 per metric tonne, roughly one cubic metre. That could be competitive with some older desalination facilities and potentially attractive in locations where transporting water is expensive. However, the best modern desalination plants can achieve costs closer to $2 per tonne or below, depending on local conditions. Atoco believes it may require several more years to approach that range.

But this comparison needs context. Atmospheric water harvesting does not always compete directly with the world's cheapest desalination facility. Desalination generally requires access to seawater or another saline source, large fixed plants, substantial energy, pipelines and infrastructure for disposing of concentrated brine. Atoco can theoretically operate in locations where there is no ocean at all. A mine in a remote region may currently receive water by truck. A desert greenhouse may have limited groundwater. A data centre may face severe restrictions on municipal supply. A military site or isolated hospital may value independence much more than the absolute global commodity price of water.

The competitor is not necessarily the cheapest litre of water on Earth. It is the cheapest reliable litre at the place where the customer actually needs it.

That distinction helps explain how Atoco could initially charge more than desalination and still have a viable business. But scale remains the much harder problem. The prototype producing hundreds of litres per day sounds impressive until compared with the millions of litres that large industrial facilities can potentially use. Atoco's proposed solution is modularity: install many systems together. Yet multiplying a machine also multiplies manufacturing requirements, maintenance, capital investment, space and opportunities for failure.

The company must therefore prove more than the molecular chemistry. It needs to manufacture large quantities of specialised porous materials cheaply. Those materials must retain their performance after thousands of capture-and-release cycles. Airflow systems must survive dust, heat and industrial environments. The machines must operate continuously with minimal maintenance. Production needs to expand from prototypes into standardised hardware. Supply chains must form around a category that barely exists today.

Atoco has largely proved that the physics works. It now has to prove that the factory works.

That transition destroys many deep-tech companies. A material can perform beautifully in a laboratory while becoming too expensive, fragile or difficult to manufacture at industrial scale. The Atoco story from this point onward will therefore be less about scientific discovery and increasingly about process engineering, manufacturing discipline and unit economics.

AI is both the problem and part of the solution

There is another beautiful circularity inside the company: Atoco uses artificial intelligence to help discover the materials that could eventually reduce AI infrastructure's environmental footprint.

The design space for metal-organic frameworks is enormous. Chemists can alter metal centres, linkers, pore geometries, functional groups and structural arrangements, creating a huge number of theoretically possible materials. Testing every combination experimentally would be impossibly slow. Atoco uses machine-learning systems to predict the properties of candidate structures and decide which materials are worth synthesising and testing physically.

This changes the research process. Instead of chemists manually exploring thousands of possibilities, computational models can narrow an enormous molecular search space into a much smaller set of promising candidates. Experiments then generate new data that can improve the models. The company is effectively creating its own materials-discovery flywheel: AI proposes molecules, laboratories test them, results improve the data, and better data helps the AI identify the next generation of materials.

The irony is difficult to miss. Generative AI is increasing demand for data centres. Data centres create more waste heat and pressure on local water supplies. Atoco is using AI to design materials that can use that waste heat to harvest water. AI is helping build a solution to an infrastructure problem created partly by AI itself.

This also makes Atoco a good example of how AI could create value outside software. Much of the current AI conversation revolves around chatbots, agents and digital applications. Materials discovery is a very different proposition. A better model can eventually produce a better physical substance, which can then alter the economics of water, energy, manufacturing or carbon capture. The AI disappears into the research process while the output becomes something tangible.

Why now matters almost as much as why Atoco

Many important technologies spend years waiting for the market around them to become ready. Atoco may be experiencing exactly that moment. The science behind MOFs has been developing for decades. Atmospheric water harvesting did not suddenly become physically possible in 2026. What changed is the environment around the technology.

Water scarcity is becoming more economically important. Climate change is making drought more frequent in some regions. Data centres are expanding rapidly. AI is creating enormous new power and cooling requirements. Technology companies are facing pressure to improve their water footprints. Grid constraints are pushing some facilities toward on-site generation. Carbon capture remains strategically important for industries that cannot easily eliminate emissions. Machine learning is accelerating materials discovery. And Yaghi's Nobel Prize has dramatically raised awareness of the molecular frameworks Atoco is commercialising.

The scientific breakthrough and the business opportunity did not arrive at the same time. They may finally be meeting.

The Nobel Prize itself creates an unusual advantage. Scientific recognition does not guarantee commercial success; Nobel committees do not assess unit economics, manufacturing yields or customer-acquisition costs. But it removes one layer of scepticism. Corporate customers no longer need to be convinced that metal-organic frameworks are an obscure scientific curiosity. One of chemistry's highest honours has already validated the importance of the underlying field.

For a young deep-tech company, credibility can be almost as important as capital. Infrastructure buyers are extremely conservative because the cost of failure is high. A data-centre operator may be interested in an exciting new material, but it needs confidence that the underlying science is serious. Atoco still has to prove that its product works economically. It no longer has to prove that its science deserves to exist.

Interestingly, the company has disclosed far less about its financing than many startups at a similar stage. Atoco is backed by Revonence Technologies International, but its total funding and valuation have not been publicly detailed. That makes it unusual in an industry where companies often build their identity around the size of their latest funding round. With Atoco, the headline number is less obvious. The technology itself is currently carrying more of the narrative.

And perhaps that is appropriate for a company whose entire philosophy begins at a scale nobody can see.

From a molecule to a water tap

The most extraordinary thing about Atoco is the distance between the beginning and end of its product.

At one end, a scientist decides how atoms should be arranged. Those atoms become a molecular building block. The building blocks assemble into a porous crystal. The crystal is processed into an engineered material. That material is placed into a module. Modules become an industrial machine. Air flows through it. Invisible molecules become trapped inside billions of microscopic pores. Waste heat releases them again.

And water comes out of a tap. A different material could perform the same trick with carbon dioxide.

That is why Atoco's slogan, “atomic precision for the planet,” is unusually accurate for startup marketing. It describes the company's fundamental theory of change: problems measured in gigatonnes and billions of litres may ultimately be influenced by controlling matter at the scale of individual molecules.

The company is now reaching the point where great deep-tech businesses are either created or exposed. The scientific foundation is extraordinary. Working prototypes exist. Industrial trials are underway. Potential customers are appearing. A Nobel Prize has transformed awareness of the underlying technology.

What remains is the difficult part.

Can Atoco manufacture enough specialised material cheaply? Can its systems operate reliably for years rather than weeks? Can hundreds of litres become hundreds of thousands? Can water harvested from air compete with the alternatives available in the places that actually need it? Can carbon capture move from laboratory performance into industrial economics? And can something designed with atomic precision eventually be manufactured with the boring consistency required of infrastructure?

Those questions are harder than proving that a porous crystal can capture a water molecule. They are also the questions that matter. Because Atoco is no longer trying to demonstrate that humans can pull water from desert air. The underlying science has already shown that.

It is trying to prove that a Nobel Prize can become infrastructure.

And if it succeeds, one of the most ordinary things around us begins to look very different. The air stops looking empty. It starts looking like inventory.

Research updated 12 October 2026.