Here’s a scenario that keeps semiconductor CEOs up at night.
You’ve just broken ground on a $40 billion advanced logic fab. The politicians came for the ribbon-cutting. ASML confirmed delivery of your EUV systems. Applied Materials has your deposition tools in production. Wall Street analysts are modeling your 2028 revenue ramp.
Three months into construction, your chief engineer walks into your office with a problem.
The municipal water authority can deliver 6 million gallons per day to your site. Your ultrapure water consultant says you’ll need 14 million once you’re at full capacity. The system to close that gap—assuming you can source all the equipment—won’t be operational for 32 months. Optimistically.
Your $40 billion fab just became a very expensive paperweight until 2029.
This isn’t a hypothetical scenario. This is happening right now across the United States, Europe, and Asia as the semiconductor industry attempts the largest manufacturing expansion in its history. And most investors have no idea it’s coming.
Everyone’s obsessing over EUV lithography, geopolitical chip wars, and China’s technology ambitions. Meanwhile, the industry is quietly hitting a constraint that could derail the entire buildout: it’s running out of ultrapure water capacity.
The companies that solve this problem are about to make a fortune. And the market hasn’t figured it out yet.
Start with a basic fact that should reshape how you think about semiconductor manufacturing: a modern advanced fab consumes 10 to 15 million gallons of water every single day.
To put that in perspective, that’s roughly equivalent to the daily water consumption of a small city. But the comparison actually understates the problem, because municipal water and semiconductor water exist in completely different universes.
City water needs to be clean enough to drink—roughly 99.9% pure. Semiconductor manufacturing requires water at 99.9999999% purity. Nine nines. At that level, you’re measuring contaminants in parts per trillion. A single stray ion—literally one molecule in a billion—can brick an entire silicon wafer worth tens of thousands of dollars.
TSMC alone burns through approximately 63 million tons of water annually across all facilities. That’s more than many countries consume. And they’re planning to double capacity by 2028 to meet AI chip demand.
Here’s the insight most people miss: producing ultrapure water at semiconductor specifications is harder than manufacturing many of the chips themselves. The infrastructure required to deliver 15 million gallons per day at nine-nines purity represents one of the most complex industrial processes humans have ever scaled.
And right now, every major chipmaker building new capacity is discovering the same thing: water infrastructure is the longest lead-time item in the entire fab. Longer than cleanrooms. Longer than most process equipment. In many cases, longer even than ASML’s infamous EUV machines.
Let me explain why this is so technically challenging, because it’s not intuitive.
Regular water purification—the kind used for drinking water or even pharmaceutical manufacturing—removes particles, bacteria, maybe some dissolved solids. Standard stuff. Multi-stage filtration, maybe some UV treatment, you’re done.
Semiconductor ultrapure water is something else entirely. You’re removing everything: dissolved atmospheric gases, organic molecules down to the molecular level, metallic ions, particles smaller than most viruses, bacteria obviously, endotoxins, silica. The final product must have electrical resistivity of 18.2 megohm-cm, which is the theoretical maximum for H2O.
Achieving this requires a treatment train that can include a dozen or more stages: pre-treatment, reverse osmosis, electrodeionization, degasification, UV oxidation, ultrafiltration, final polishing. Each stage targets specific contaminants. Each stage requires equipment from specialized suppliers. Each stage must be redundant because any component failure means immediate production shutdown.
The capital cost for a complete ultrapure water system serving an advanced fab runs $200 to $400 million. That’s before installation, commissioning, and qualification. And the lead time from purchase order to first qualified water? Two to three years under normal conditions.
But these aren’t normal conditions. Right now, with Intel, TSMC, Samsung, and a dozen other chipmakers simultaneously building out massive new capacity, lead times have stretched to 30 to 40 months. The specialized equipment suppliers can’t scale fast enough. The engineering firms that design these systems are completely booked. The construction contractors who install them are overcommitted.
And here’s the multiplicative factor that’s accelerating this crisis: every process node shrink increases water consumption exponentially. Moving from 7nm to 3nm doesn’t just require moderately more water—it requires three times as much per wafer. More cleaning cycles. More rinsing steps. Higher purity requirements because smaller features are more sensitive to contamination.
The 2nm nodes that will start production in 2026-2027? Even more water-intensive. The industry can’t engineer its way out of this—it’s fundamental to the physics and chemistry of chip manufacturing.
Now let’s zoom out to the macro picture, because the numbers are staggering.
The U.S. CHIPS Act allocated $52 billion in government funding for semiconductor manufacturing. When you add private sector co-investment, that translates to over $200 billion in new fab construction through 2030. Similar programs are underway in Europe, Japan, and Korea.
TSMC is building facilities in Arizona with commitments exceeding $65 billion. Intel has massive projects in Ohio and Arizona. Samsung is expanding in Texas. Micron is constructing in upstate New York. Every one of these fabs will require 10 to 15 million gallons of ultrapure water daily at full production.
Add it all up, and the semiconductor industry needs to more than double its global ultrapure water infrastructure capacity by 2030. Current annual capacity for new ultrapure water systems globally? Roughly $3 to $4 billion. Required capacity to meet projected demand? North of $8 billion annually.
That’s not a minor gap. That’s a structural shortage that will take years to resolve even if every supplier maxes out production capacity.
And we haven’t even addressed the elephant in the room: water availability.
Here’s where the semiconductor industry’s expansion plans run headfirst into physical reality.
The economically optimal locations for advanced fabs—places with skilled workforces, existing tech ecosystems, favorable business environments, strong infrastructure—increasingly happen to be water-scarce regions.
Arizona, where TSMC and Intel are making multi-billion dollar bets, is experiencing its worst drought in over a millennium. The Colorado River, which supplies much of the state’s water, is at historically low levels. Agricultural water allocations are being cut. Groundwater tables are declining. And now the state is trying to support semiconductor fabs that each consume as much water as a small city.
Texas faces similar dynamics. The state has dealt with recurring droughts, and its electrical grid fragility is well-documented. Now add extreme water demand from semiconductor facilities on top of existing stress.
Even Taiwan—home to TSMC’s most advanced production—experiences periodic water shortages that have forced production adjustments. In 2021, Taiwan faced its worst drought in 50 years, and TSMC had to truck in water to keep fabs running.
The industry is caught in a fundamental tension: the best places to build chips from a business perspective are increasingly the worst places from a water availability perspective. And relocating to water-abundant regions means sacrificing workforce quality, supply chain proximity, and infrastructure advantages.
This isn’t a cyclical challenge that resolves when commodity prices adjust. This is a permanent structural constraint that fundamentally alters the economics of semiconductor manufacturing.
Faced with this reality, the industry has been forced into extremely aggressive water recycling programs. Not because of environmental pressure—though that helps—but because it’s the only way to make the math work.
Modern advanced fabs now recycle 85% to 95% of water consumption. TSMC publicly commits to 90%+ recycling rates. Intel targets 95%+ at new facilities. Samsung makes similar claims.
But recycling ultrapure water is harder than creating it in the first place. After use in semiconductor manufacturing, that water is contaminated with photoresists, etching chemicals, metals, solvents, and countless other compounds. Bringing it back to nine-nines purity requires even more sophisticated treatment than starting with municipal water.
The cost implication is significant: comprehensive water reclamation adds $100 to $200 million to the ultrapure water system capital budget. It doubles the treatment complexity. It increases maintenance requirements exponentially. And it’s becoming a non-negotiable requirement for permitting.
No major fab gets environmental approval anymore without demonstrating 90%+ water recycling capability. The industry has no choice but to pay.
While the semiconductor industry worries about single-source dependencies on ASML for EUV lithography, there’s an equally severe concentration in ultrapure water equipment that nobody discusses.
The specialized equipment required for semiconductor-grade water production comes from a small number of manufacturers globally. We’re talking about industrial-scale reverse osmosis systems with semiconductor-specific specifications, ion exchange systems capable of achieving 18+ megohm-cm resistivity, UV oxidation systems, ultrafiltration membranes with sub-10 nanometer pore sizes.
These aren’t catalog items you can source from multiple vendors. Each system requires custom engineering for the specific fab’s requirements. Integration with process tools. Redundancy architecture. Qualification protocols that can take six months.
And lead times are brutal. Pre-pandemic, 18 months was standard. Now? 30 to 36 months minimum for major systems. Some specialized components are pushing 40+ months.
The companies that can deliver get whatever pricing they ask. This isn’t equipment where chipmakers can bargain hard or play suppliers against each other. If your ultrapure water system is delayed six months, your $20 billion fab sits empty burning cash. You pay the premium and thank them for fitting you into the schedule.
Here’s what makes this opportunity particularly attractive from an investment perspective.
Traditional semiconductor equipment companies are cyclical. When chip demand surges, equipment orders explode. When demand softens, capex gets slashed. Stock prices yo-yo accordingly.
Ultrapure water companies have fundamentally different dynamics:
New fab construction drives initial system sales—these are huge, lumpy orders worth hundreds of millions.
Existing fab expansions require capacity additions—smaller but frequent upgrade projects.
Aging infrastructure requires replacement—equipment doesn’t last forever, especially under 24/7 operation.
Service and maintenance generates recurring revenue at 40%+ margins—once a fab installs your system, you’re locked in for 15 to 20 years.
The result is much steadier revenue than typical semiconductor equipment suppliers, with built-in recurring revenue streams that provide downside protection during industry downturns.
And the growth trajectory is stronger than the semiconductor industry itself because each new process node requires more water per wafer. The TAM is expanding independent of chip unit volumes.
Pricing power is exceptional. Ultrapure water system pricing has increased 30% to 40% since 2020 despite general industrial equipment deflation. This is specialty equipment where reliability and technical capability matter infinitely more than cost. You don’t bargain shop when production shutdown costs millions per hour.
Semiconductor investors fixate on obvious bottlenecks: lithography, deposition tools, the latest process node announcements. Water treatment is infrastructure—boring, assumed to be commoditized, not exciting enough for equity research coverage.
This perception creates the opportunity.
The companies providing ultrapure water solutions exhibit characteristics that typically command premium valuations: long-term contracted revenue, high customer switching costs, 40%+ service margins, secular growth independent of demand cycles, limited competition in advanced applications.
Yet many trade at 12x to 18x earnings—industrial conglomerate multiples. Compare that to semiconductor equipment companies at 25x to 35x despite having inferior business models in many respects.
The disconnect exists because water treatment isn’t sexy. It doesn’t get keynote presentations at industry conferences. Analysts don’t write breathless reports about the latest purification technology breakthrough.
But while everyone watches ASML’s order book, the ultrapure water suppliers are quietly becoming indispensable infrastructure for the entire industry’s expansion. And that gap between perception and reality creates investable opportunities.
This isn’t a two-year trade on semiconductor capex cycles. This is a decade-long structural change in how the industry approaches manufacturing infrastructure.
Water scarcity isn’t temporary. Climate patterns are shifting. Population growth continues. Industrial competition for water resources intensifies. Every major semiconductor-producing region faces increasing water stress.
The industry’s response—aggressive recycling, efficiency improvements, advanced treatment—requires more sophisticated water management solutions, not simpler ones. Capital intensity increases. Technical requirements become more demanding. The moat around suppliers with proven capabilities widens.
Government policy accelerates the trend. Environmental regulations tighten. Permitting requires demonstrated water stewardship. Public pressure forces transparency. All of this drives demand for premium water treatment solutions.
And every new fab that breaks ground locks in 20+ years of service revenue for the supplier that wins the initial contract. Market share gains today compound into durable competitive advantages tomorrow.
The investment thesis is clear: ultrapure water infrastructure is becoming the critical constraint for semiconductor manufacturing expansion. Companies that can deliver these systems at scale have multi-year backlogs, exceptional pricing power, and minimal competition in advanced applications.
Most trade at valuations that ignore these dynamics entirely.
Here are three positions I’m building—a European industrial conglomerate with global dominance, a Japanese specialist that’s the default choice for Asian fabs, and a North American pure-play levered to CHIPS Act spending.
Ready to see the specific opportunities?