AI Boom May Be Next Wave of Forever Chemical Production

CLIFF NOTES

• AI growth is increasing demand for PFAS used in semiconductor manufacturing and advanced data-center cooling.
• ChemSec reports that most major PFAS producers are expanding or planning additional production.
• Semiconductor manufacturers say some PFAS applications currently lack practical drop-in substitutes.
• U.S., European and UK regulators are tightening or evaluating PFAS controls as industrial demand grows.
• EPA research identifies activated carbon, ion exchange and reverse osmosis as important technologies for reducing PFAS in drinking water.

The rapid growth of artificial intelligence is creating a new environmental problem far beyond the electricity demands of data centers. Chemical manufacturers are expanding production of PFAS, the persistent substances used in semiconductor manufacturing, advanced cooling systems and batteries. Regulators, meanwhile, are trying to limit PFAS contamination in water and the environment.

That collision — between technological expansion and chemical regulation — is becoming one of the less visible consequences of the AI boom.

Why Is AI Increasing Demand for PFAS?

Artificial intelligence depends on physical infrastructure.

AI models require powerful computer chips. Those chips sit inside data centers packed with servers that generate enormous amounts of heat. Producing the chips and controlling that heat require specialized materials that can tolerate extreme chemical and thermal conditions.

PFAS, short for per- and polyfluoroalkyl substances, can provide those properties.

The same carbon-fluorine bonds that make many forever chemicals resistant to heat, water and chemical attack also make them extremely persistent after they enter the environment.

That creates the central problem: some of the characteristics that make PFAS useful to advanced technology also make PFAS contamination difficult and expensive to manage.

Swedish chemicals organization ChemSec reported in September 2026 that most of the world’s major PFAS producers are expanding or planning to expand PFAS-related production. ChemSec identified AI and data-center infrastructure, semiconductor manufacturing and lithium-ion battery materials as three important sources of new demand.

Anne-Sofie Bäckar, executive director of ChemSec, warned:

“These companies’ expansion plans show that unless governments and regulators insist on a rapid phaseout, we will face a new tidal wave of forever chemicals.”

Semiconductor Manufacturing Is a Major Part of the Story

The connection between AI and PFAS begins before a server reaches a data center.

Advanced AI processors require extremely sophisticated semiconductor manufacturing. PFAS-containing materials can appear in photolithography, wet chemistry, heat-transfer fluids, manufacturing equipment, lubricants, packaging and other parts of semiconductor fabrication.

The Semiconductor Industry Association’s PFAS Consortium has documented seven broad areas where PFAS chemistry supports semiconductor manufacturing: photolithography, wet chemistry, fluorinated gases, heat-transfer fluids, chip packaging, fabrication tools and associated equipment, and lubricants.

The industry argues that substitution is not simple.

The Semiconductor Industry Association says PFAS-containing surfactants can be important for uniform coatings, resist removal, reducing defects and controlling extremely small structures during lithography. It also says there is no single replacement that works across every application.

This distinction matters.

“PFAS” does not describe one chemical. It describes a very large family of fluorinated substances with different properties and uses. Any discussion about restricting them therefore quickly becomes a debate over which uses can be eliminated, which can be replaced and which industries claim still require exemptions.

The semiconductor industry has said qualification of substitutes can take years because changing one chemical can affect a highly controlled chain of manufacturing steps. According to the Semiconductor Industry Association, a typical semiconductor process technology change can require 10 to 15 years for full qualification and integration.

That means growing AI chip demand could lock some PFAS uses into supply chains for years unless workable alternatives emerge faster.

AI Data Centers Are Creating Another PFAS Market

The second connection comes after the chips are manufactured.

AI servers can produce far more concentrated heat than conventional computing equipment. Traditional air cooling becomes increasingly difficult as rack power density rises, pushing data-center designers toward liquid cooling.

One technology is two-phase immersion cooling.

Servers or computing components are exposed to a specialized dielectric fluid. Heat from the electronics causes the fluid to vaporize. The vapor rises, condenses and returns to the system, allowing the process to repeat.

Chemical companies see a substantial market.

In February 2026, Chemours announced a partnership with 2CRSi involving its Opteon two-phase immersion cooling fluid for high-density servers and AI infrastructure. Chemours said the technology can reduce cooling energy and water consumption compared with traditional systems.

Chemours had already announced manufacturing plans in 2025 designed to provide new capacity for its two-phase cooling fluid beginning in 2026, explicitly connecting the expansion to the heat, energy and water demands of AI hardware and advanced data centers.

The company continued expanding its data-center cooling portfolio in August 2026, announcing additional Opteon refrigerants for chillers serving AI-driven data centers.

Daikin is also pursuing the data-center market. In July 2026, Daikin and NTT DATA announced an AI-driven data-center cooling project aimed at predicting server heat conditions and optimizing cooling operations. Daikin noted that generative AI servers consume more power and generate more heat than conventional servers.

The environmental tradeoff is becoming clearer: liquid cooling may help reduce electricity or water requirements in some data centers, but the chemistry used to achieve those efficiencies also has to be considered.

PFAS Producers Are Expanding Beyond Data Centers

ChemSec’s research suggests AI is only part of a larger expansion.

Arkema has invested roughly $60 million in a new production unit at its Calvert City, Kentucky, facility. The company has also expanded PVDF capacity serving lithium-ion batteries and other markets.

Battery production is important because fluoropolymers such as PVDF can serve as binders and coatings in lithium-ion batteries used in electric vehicles, electronics and energy-storage systems.

ChemSec also identified expansion activity involving companies including AGC, Chemours, Solstice and Syensqo. Its September 2026 report said Daikin planned to more than triple fluoropolymer production capacity in response to semiconductor demand.

Not every chemical company is moving in the same direction.

ChemSec reported that Archroma has been reducing PFAS substances while developing PFAS-free alternatives. BASF has also announced plans to phase out most products formulated with PFAS by 2028, with pesticides treated separately.

The divide is becoming sharper: some manufacturers see PFAS-intensive technologies as growth markets, while others see substitution as the safer long-term strategy.

Why Are PFAS Called “Forever Chemicals”?

The nickname comes from persistence.

Carbon-fluorine bonds are exceptionally strong. That stability makes many PFAS useful where ordinary materials would degrade, but it also means some PFAS can remain in the environment for long periods.

Once released, PFAS can move through soil, groundwater, surface water and food systems.

The U.S. Environmental Protection Agency says people can encounter PFAS through contaminated drinking water, food, soil, dust, air, workplaces and PFAS-containing products. The agency also states that certain PFAS can accumulate in the body over time.

Research has associated exposure to certain PFAS with adverse health outcomes. The evidence and level of risk differ by compound, dose and exposure pattern, which is why regulators increasingly distinguish between specific PFAS instead of treating every substance as biologically identical.

For households, the most direct concern is often PFAS in drinking water.

The United States Already Has Federal PFAS Drinking-Water Limits

The regulatory landscape has changed significantly.

In 2024, the EPA established the first nationwide legally enforceable drinking-water limits for several PFAS. The enforceable maximum contaminant levels for PFOA and PFOS were set at 4.0 parts per trillion each.

Federal policy is still evolving.

As of September 2026, the EPA is maintaining the PFOA and PFOS standards but has proposed allowing qualifying public water systems two additional years, until 2031, to comply. The agency has separately proposed rescinding the existing regulations for PFHxS, PFNA, HFPO-DA — commonly called GenX — and the Hazard Index mixture, while potentially evaluating additional PFAS through future rulemaking. Those proposals should not be confused with final changes already in force.

The EPA also continues to require monitoring under the existing federal framework. This means communities are receiving increasingly detailed information about drinking water contaminants that historically were not routinely measured.

For consumers, concentrations measured in parts per trillion can seem abstract. One part per trillion is one part out of one trillion parts of water. The tiny numbers involved show how low federal limits have become for PFOA and PFOS.

Europe Is Moving Toward Broader PFAS Restrictions

Europe is considering a different approach.

The European Chemicals Agency, or ECHA, has been evaluating a broad PFAS restriction proposal under the European Union’s REACH chemicals framework.

In March 2026, ECHA’s Risk Assessment Committee adopted its opinion, while the agency’s Socio-Economic Analysis Committee continued evaluating the proposal. ECHA has indicated support for restriction combined with targeted derogations where alternatives may not yet be technically or economically available.

That creates an important conflict for industries such as semiconductor manufacturing.

Regulators want to reduce long-term PFAS releases. Chip manufacturers argue that certain uses cannot yet be replaced without disrupting advanced semiconductor production.

AI growth raises the stakes because semiconductor demand is moving in the opposite direction from the regulatory pressure.

The United Kingdom Has Also Strengthened Its PFAS Strategy

The UK government published its first national PFAS Plan in February 2026.

The plan focuses on identifying PFAS sources, understanding how the chemicals move through society and the environment, and reducing exposure. The government describes the long-term objective as minimizing harmful PFAS effects while moving toward safer alternatives.

The UK is also considering PFAS controls through UK REACH and implementing international restrictions agreed under the Stockholm Convention.

This means manufacturers now face overlapping pressures in several major markets: expand production for semiconductors, batteries and data centers while preparing for increasingly restrictive chemical policies.

Why More PFAS Production Could Eventually Become a Water Problem

PFAS production does not automatically mean PFAS will enter a particular community’s drinking water.

The route matters.

Contamination can occur through industrial releases, wastewater, disposal sites, firefighting foam, landfills and other pathways. The EPA identifies manufacturing and processing facilities among known contributors to PFAS releases into air, soil and water.

This is why production volume matters even when the chemicals are being manufactured for technologies that appear far removed from drinking water.

Persistent chemicals can create long-lived waste-management problems. Once contamination reaches groundwater or a drinking-water source, removing it can require treatment for years.

The EPA has also noted that conventional wastewater treatment was generally not designed specifically to remove PFAS. Advanced processes may therefore be needed where significant contamination occurs.

The question surrounding the AI boom is consequently larger than whether PFAS help make better computer chips.

It is who manages the chemicals after they have served their industrial purpose — and who pays when they escape containment.

PFAS Treatment Is Possible, but It Does Not Make the Chemicals Disappear

Water treatment creates another important distinction.

Removing PFAS from drinking water is not necessarily the same thing as destroying PFAS.

Filtration usually transfers contaminants from the water into another material or waste stream. Carbon media eventually has to be replaced. Ion-exchange resin reaches the end of its useful life. Reverse osmosis creates a concentrated reject stream.

The EPA identifies three major technologies capable of reducing PFAS in drinking water:

• granular activated carbon,
• ion-exchange resin,
• and high-pressure membranes such as reverse osmosis.

EPA research has found granular activated carbon can be effective for PFAS, particularly longer-chain compounds such as PFOA and PFOS. Shorter-chain PFAS can be more difficult for carbon treatment.

High-pressure membranes, including reverse osmosis, have also demonstrated strong PFAS reduction performance.

The treatment system still has to match the contaminant.

That is why water testing and laboratory analysis matter when PFAS contamination is suspected. A general household water test should not automatically be assumed to include PFAS; PFAS analysis normally requires specialized laboratory methods.

What Should Homeowners Know About PFAS Filters?

Not every filter marketed for better-tasting water is designed to reduce PFAS.

The EPA advises consumers to look for independently certified products with PFAS-reduction claims. It identifies granular activated carbon, ion exchange and reverse osmosis among technologies that can substantially reduce PFAS when properly selected and maintained.

Maintenance is critical.

A filter that initially captures PFAS does not have unlimited capacity. Carbon, resin and membranes must be serviced or replaced according to their design and the quality of the incoming water.

Homeowners considering a PFAS filter should therefore look beyond the word “filter” and examine the specific contaminant-reduction certification and performance data for the system.

EPA guidance specifically points consumers toward NSF/ANSI 53 and NSF/ANSI 58 certifications when evaluating products with PFAS-reduction claims.

How Reverse Osmosis and Whole-Home Treatment Relate to PFAS

The AI-PFAS connection may seem distant from an ordinary kitchen faucet, but persistent industrial chemicals ultimately become a local issue when they reach a drinking-water source.

For drinking water, a properly designed reverse osmosis system can provide point-of-use treatment at a kitchen sink. EPA research identifies reverse osmosis as highly effective at reducing PFAS, although system performance, certification, maintenance and incoming water conditions still matter.

Whole-home treatment serves a different purpose. A dedicated whole house PFAS filter may use treatment media designed for PFAS reduction before water is distributed throughout a home.

A conventional whole house water conditioner, however, should not automatically be assumed to remove PFAS. Water softening or conditioning systems are commonly designed around hardness, chlorine, chloramine or other water-quality problems. PFAS reduction requires treatment specifically designed and verified for that purpose.

The distinction matters because there is no single device that solves every water-quality problem.

The Hidden Environmental Cost of the AI Race

Artificial intelligence is often discussed in terms of computing power, electricity consumption and water use. PFAS adds another layer.

The infrastructure behind AI begins with mines, chemical plants and semiconductor fabrication facilities long before an AI model reaches a data center.

PFAS can help manufacturers build extraordinarily small semiconductor structures. Fluorinated fluids can help manage the intense heat generated by powerful computing systems. Similar materials support batteries and other technologies tied to electrification.

Those advantages have economic value.

Their persistence creates a different kind of bill.

As AI infrastructure expands, regulators will have to decide which PFAS applications are truly essential, which can be replaced, how industrial releases should be controlled and who bears the cost of contamination that may remain long after the equipment that created the demand has been replaced.

ChemSec’s warning is therefore not simply about another chemical factory expanding production.

It exposes a larger contradiction in the AI economy: technology designed to build the future may increase dependence on chemicals whose environmental legacy can last far beyond the technology itself.

For communities concerned about safe drinking water, that is the part of the AI boom that may ultimately matter most.

Official Sources and Further Reading

Readers who want to explore the research, regulations, and industry developments behind this story can review these primary and official sources:

• U.S. Environmental Protection Agency (EPA): Current understanding of PFAS health and environmental risks
• EPA: PFAS drinking water regulations
• EPA: PFAS drinking water treatment technologies
• EPA: Reducing PFAS in drinking water with home filters
• EPA: Identifying drinking water filters certified to reduce PFAS
• European Chemicals Agency (ECHA): PFAS restriction and regulatory information
• UK Government: United Kingdom PFAS Plan
• Semiconductor Industry Association: PFAS use in the semiconductor industry
• ChemSec: Research on the world’s major PFAS producers
• Chemours: Two-phase immersion cooling for AI and high-density data centers
• Daikin: AI-driven data-center cooling project with NTT DATA