Data centers are swapping water for forever chemicals to keep AI cool

ChemSec reported that AI-driven data-center cooling, semiconductor manufacturing and battery demand are driving major PFAS producers including Chemours, Daikin and Arkema to expand capacity, intensifying environmental and regulatory risks around "forever chemicals." ChemSec estimates that 50%-67% of Chemours' $5.8B revenue is tied to PFAS, while Daikin plans to triple PFAS capacity for semiconductor demand and Arkema is expanding in North America and Asia. The growth opportunity is shadowed by EPA opposition to Chemours' proposed Opteon 2P50 cooling chemical and a French lawsuit against Arkema and Daikin alleging PFAS pollution.

Analysis

The investable issue is not incremental cooling-material demand; it is whether PFAS-linked products retain a license to operate in mission-critical cooling. For CC, a negative EPA decision or extended review would impair the highest-multiple portion of the AI-cooling narrative while reinforcing a pre-existing liability discount, creating downside through both lower growth expectations and a higher terminal legal-reserve assumption. AKE is less exposed to a single U.S. regulatory binary, but European litigation can raise remediation provisions, insurance costs, and capex requirements before any final judgment.

Over the next 1-3 months, activist pressure and election-season scrutiny create headline volatility disproportionate to near-term revenue exposure. The critical verification point is whether hyperscalers and chiller OEMs commit to specific fluorinated-fluid platforms; without disclosed volume contracts, capacity-expansion announcements should not be capitalized as durable earnings. A regulatory delay would also favor cooling architectures that reduce reliance on specialized dielectric fluids, benefiting direct-to-chip liquid-cooling and thermal-management suppliers such as VRT, MOD, NVT and JCI.

The contrarian view is that a closed-loop application may face lower realized emissions than consumer or open-system PFAS uses, so an outright ban is not the base case. However, even a conditional approval can add monitoring, disposal, containment and reporting requirements that erode customer total-cost-of-ownership, slowing adoption rather than eliminating it. The thesis is falsified if EPA clearance arrives without material operating restrictions and CC discloses contracted data-center cooling volumes sufficient to move segment guidance.

Key Decisions for Investors

• Maintain an underweight in CC versus the chemicals basket; add a 3-6 month put spread only on confirmation that the EPA review is delayed, challenged, or conditioned. Target a 10-15% relative downside versus XLB if AI-cooling growth assumptions are removed; exit on unconditional approval plus raised Thermal & Specialized Solutions guidance.
• Pair trade: long AKE / short CC over a 6-12 month horizon, sized modestly. The trade isolates CC's more concentrated U.S. regulatory and liability sensitivity while retaining exposure to industrial-materials demand; stop out if CC's cooling-product order disclosures materially exceed expectations or AKE records a significant new PFAS provision.
• Use VRT, MOD, or NVT as watch-list beneficiaries rather than immediate buys: initiate only after OEM or hyperscaler commentary indicates direct-liquid cooling is gaining share at the expense of dielectric/immersion designs. The catalyst window is the next two earnings cycles; valuation risk is high after AI-infrastructure outperformance.
• Set alerts for EPA docket updates, state PFAS restrictions, and disclosed customer qualification wins by CC. Treat these as more decision-relevant than advocacy-report estimates, which do not establish revenue realization or product-level margin contribution.