Complex Water Systems Shortages

Water has become a hot topic of conversation, with many headlines flagging the enormous amount of water that is being used by things like AI datacenters. But this is not the shortage that we think is most interesting. Regulators and startups are helping to address the use of the water, but we need people to build the systems in the first place. This week, instead of focusing on the volume of water consumed by these facilities, we want to talk about the systems themselves. Where water systems are in the most demand, what they are actually doing, and the supply challenges the industry is facing.

Which Industries Are Using Complex Water Systems?

While AI datacenters take the headlines, they are not the only consumer of complex water systems. Semiconductor fabs and their use of ultra-pure water (UPW) are another large consumer, and lithium processing is an emerging use case we expect to grow rapidly in the coming years:

• Datacenters: Datacenters use lower quality water primarily for cooling. Their wastewater is less toxic, and closed-loop systems are radically reducing the total amount of water that needs to be processed/treated.
• Semiconductor Fabs: Semiconductor fabs use ultrapure water (UPW) to rinse wafers and remove chemicals and microscopic particles that could short-circuit the chips. Removing these chemicals and particles from the water after the cleaning is also a highly complex process.
• Other: Lithium/batteries, pharma & biotech, mining, oil & gas, and the newly emerging green hydrogen industries are also consumers of complex water systems.

Breaking Down Water Systems

While the exact system demands vary for each market, you can broadly think about water systems by their function (water in, usage, water out), however in practice they are all tightly intertwined:

Water in (where water comes from): Systems in this bucket tend to focus on water quality. For example, fabs and biotech require extremely purified water while datacenters are less stringent. Ultra-pure water (UPW) is the most extreme version of this and is virtually free of organic and inorganic compounds, particles, and gases.

Water usage (how it is used): This includes cleaning processes used in fabs, cooling processes at datacenters, or even the use of water as an ingredient in biopharma.

Water out (removing or recycling water from the system): Systems in this bucket focus on the waste that the system produces. They work to remove toxic chemicals and particles, and make the water clean enough to dispose of or recycle back into the system. This is called wastewater treatment (WWT) and wastewater reclaim (WWR). The fact that a large majority of semiconductor fabs and datacenters are being built in areas with limited access to fresh water has caused regulators to push factories towards zero liquid discharge (ZLD). This is an extension of WWT/WWR where no water is discarded and is instead all recycled back into the system.

Rapidly Growing Demand

Due to the rapid adoption of AI, the demand for semiconductor fabs to create AI chips, and datacenters to manage training and inference, has increased dramatically.

• Datacenters: Major hyperscalers will spend $800Bn in capex in 2026, increasing to $1.2T in 2028 (50% increase) (Wedbush). Estimating water cooling systems as ~3.5% of total capex (Jefferies) or $1.5M per MW (HSBC) of power added, both get you to an estimated revenue opportunity of ~$28Bn for complex water systems in 2026, increasing to ~$43Bn in 2028.
• Semiconductor Fabs: Capex by the major fab builders is expected to reach $159Bn in 2026 and could surpass $230Bn by 2028 (~45% increase). Veolia, one of the largest water systems providers, was awarded a $177 million contract on a $17Bn fab in 2022, representing ~1% of the total capex. Applying this to 2026 fab capex would imply a ~$1.7Bn market for complex water systems increasing to $2.4Bn in 2028. Regulatory scrutiny is also pushing fabs in water-stressed environments to enhance waste management to include zero liquid discharge, maximally reducing water waste. ZLD alone, across industries, is a ~$7Bn market and expected to grow to over $10Bn by 2030.

Supply Chain Pressure

While there are dozens of players going after this market, we are hearing that the industry is not capable of servicing the booming demand. Veolia and Gradiant have consistently come up in our diligence as the largest incumbent and the largest new entrant, respectively, who are winning RFPs for complex water systems in datacenters and semiconductor fabs. Ecolab’s acquisition of Ovivo and CoolIT has also helped it capture substantial business. However, these three players combined are only generating ~$2.5Bn in revenue, <10% of the estimated 2026 revenue opportunity. Including some of the other U.S based suppliers like Aquatech and Saltworks gets you to just over 11%.

U.S. Headquarters:

• Ecolab ($80Bn mkt cap): Acquired Ovivo ($1.8Bn) and CoolIT ($4.75Bn) to target fabs and datacenters. They have grown this business to nearly $1.5Bn of annualized revenue for 2026.
• Gradiant ($2Bn enterprise value; Headquarters: U.S. / Singapore): Estimated $200-400M.
• Aquatech: (Private): Estimated $750M revenue
• Saltworks: (Private) Estimated $120M revenue
• Xylem: ($29Bn mkt cap): Unknown revenue for datacenter and fab business

International Headquarters:

• Veolia ($30Bn mkt cap; Headquarters: France):  Announced a new target of €1bn revenue from AI-adjacent sectors (datacentres + semiconductors) by 2030. This compares to FY25 'AI revenue' of €560m, with only ~€50M currently from datacentres.
• Kurita: ($6.5Bn mkt cap; Headquarters: Japan)
• Organo: ($4.3Bn mkt cap; Headquarters: Japan)
• VA Tech Wabag ($1.2Bn mkt cap; Headquarters: India)

The combination of increased regulatory focus and rapid demand growth across multiple verticals has put enormous pressure on the supply chain for complex water systems, making lead times a major decision-making factor when competing on RFPs.

Most commonly, we hear that these lead times can largely be attributed to a lack of skilled labor on the design and implementation side of these systems, not the equipment.

For U.S. manufacturers, the problem is compounded by geopolitical and logistics risks, including tariffs, which can create pressure to use American design, integration, and fabrication, which puts additional pressure on supply by limiting the use of foreign suppliers. This is further exacerbated by the concern that some of the legacy international players are failing to modernize in the face of increased demand, causing quality issues.

The industry has responded with multiple solutions, including using AI to improve the efficiency of existing teams and using modular fabrication to reduce lead time, but there is still room for further innovation.

Companies like Saltworks have shown that it is possible to utilize technology to beat out incumbents. Companies like Gradiant have suggested that there is a venture-scalable path to water system design and integration. And the market is so supply-constrained that large incumbents are not able to pick up smaller contracts, leaving the door open for smaller teams that can move quickly and utilize technology to reduce costs and improve delivery times.

Takeaway: The AI boom is driving a ~$30Bn+ demand wave for complex water systems across datacenters and semiconductor fabs, but the largest suppliers combined can service ~10% of the 2026 opportunity. The bottleneck is not equipment, it is skilled labor for design and implementation, compounded in the U.S. by tariffs and pressure to keep work domestic. That supply gap makes lead times a deciding factor on RFPs and leaves the door wide open for fast-moving, tech-enabled startups to win contracts incumbents can not even bid on.