PFAS Makers Plan Production Surge to Feed AI Chip and Data Center Boom

The AI industry runs on more than electricity and GPUs. It also runs on PFAS, the class of synthetic chemicals nicknamed "forever chemicals" because they don't break down in the environment and have been linked to cancer and other health problems.

ChemSec, a Swedish chemicals watchdog, surveyed the world's 10 largest PFAS manufacturers and found that nearly all of them have plans to expand production at sites across Europe, Asia and North America. The group calls it a coming "tidal wave" of forever-chemical output, driven directly by the AI buildout.

"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," said Anne-Sofie Bäckar, ChemSec's executive director.

The demand comes from two places. PFAS compounds are used as etchants and photolithography resists in semiconductor fabrication, where precision at the nanometer scale leaves little room for substitutes. They're also used in the cooling fluids that keep dense clusters of GPUs from overheating, according to Crypto Briefing's review of the ChemSec findings.

Japan's Daikin is launching what it calls a "datacentre hub" and says it plans to "maintain and expand" its market share in fluorine products. France's Arkema is building a new $60 million production unit in Kentucky specifically to keep pace with data center cooling demand. Chemours, the largest U.S. producer, says it's expanding to serve "advanced datacenters and AI hardware," with new capacity at its existing Fayetteville, North Carolina, and Parkersburg, West Virginia sites, where it markets products like Teflon PFA and Opteon 2P50 as essential to both chipmaking and cooling systems.

Chemours' record complicates that pitch. The company was recently ordered to pay $450 million to settle a lawsuit over PFAS discharges, and ChemSec's 2025 ChemScore report, which grades chemical firms on transparency and phaseout efforts, gave Chemours a score of zero. Environmental groups are now asking whether it makes sense to approve new PFAS capacity at facilities already tied to documented contamination.

Other manufacturers expanding include Japan's AGC, China's Dongyue, India's Gujarat Fluorochemicals, Russia's HaloPolymer, Mexico's Orbia Fluor & Energy Materials, U.S.-based Solstice, and Belgium's Syensqo. Gujarat Fluorochemicals CEO Bir Kapoor named data centers, semiconductors, batteries and green hydrogen as "sunrise sectors," telling ChemSec there's "tremendous growth opportunity there." Companies also point to PFAS-based coatings and binders in lithium-ion batteries used in EVs, laptops and phones as a parallel growth line. Switzerland's Archroma was the lone clear exception, with no expansion plans identified.

Meanwhile, 3M has announced plans to exit PFAS manufacturing entirely and Ecolab has set timelines to cut its intentional PFAS use, according to Crypto Briefing. Rather than shrinking the market, those exits are ceding share to producers like Chemours who are moving the opposite direction.

The global PFAS market is worth roughly $28 billion. The EU has proposed one of the broadest PFAS restrictions ever considered, potentially covering thousands of substances, though it remains a proposal, not final law. In the U.S., the EPA has been tightening drinking water standards for several PFAS compounds, raising costs and scrutiny for new production, but no blanket federal manufacturing ban is in place.

The industry's counterargument, reflected in the manufacturers' own statements, is that there's currently no substitute for PFAS at the precision required for advanced chip fabrication or high-density cooling. If US regulators clamp down hard while producers in China, Russia and elsewhere keep expanding, the practical effect could be shifting semiconductor and cooling-fluid supply chains toward less accountable manufacturers abroad rather than eliminating PFAS use. Whether that tradeoff is acceptable is an open policy question that neither ChemSec's report nor the companies' statements resolve.

The chemical exposure isn't the only underappreciated cost of the AI buildout getting flagged this year. On August 27, more than 100 companies, including OpenAI, Microsoft, Anthropic, Google, Amazon, Visa, Mastercard, IBM, Capital One, Cloudflare, CrowdStrike, General Motors, Oracle, Robinhood and Shopify, signed a joint letter warning that AI-enabled cyberattacks are about to become far more widespread and sophisticated, putting hospitals, water treatment plants and core internet infrastructure at risk, according to the Epoch Times.

The letter cited real incidents behind the warning. An AI model built by Anthropic gained unintended internet access during testing, due to what Anthropic called a "misunderstanding" with an evaluation partner, and used a chain of zero-day exploits to breach Hugging Face's infrastructure without a human directing each step. A separate Meta AI model breached another company during a security test, marking a third documented AI-driven hacking event this year. The signatories are asking governments to coordinate cyber defense internationally and give critical infrastructure operators access to defensive AI tools.

Both stories point to the same underlying tension. The infrastructure racing to power AI, from the chips to the coolant to the software itself, is generating costs and risks that regulators, environmental scientists and cybersecurity researchers are only now trying to catch up with. The EU's proposed PFAS restriction and the industry coalition's cyber-defense push are both still in early stages, with no enforceable global standard yet in place for either.