Chemours: Buying the PFAS Discount, Then Moving Out the Curve

The Trade

Chemours' senior unsecured paper was trading wide of BB/B materials credits carrying similar leverage. The gap was compensation for two risks (PFAS litigation and a TiO₂ trough) that the balance sheet could already absorb, and the thesis didn't require a recovery.

I pitched this credit twice, four weeks apart and at two points on the curve, the first time with a team at UCLA Anderson's Fink Center Credit Pitch Competition and the second time at UNC Kenan-Flagler's Alpha Challenge.

Both pitches made the same case at the same seniority, and the December one is what you get when you act on the November analysis.

Moving Out the Curve

In the UCLA deck we priced the PFAS premium separately at each point on the curve, and it wasn't flat, coming out at roughly +150bp on the 2028s, +200bp on the 2029s and +150bp on the 2033s against the fitted market curve. We pitched the 2028s in November because the shorter tenor made the carry easier to defend in a bear case, even though the analysis put the largest mispricing one maturity further out.

By early December there was more to work with, because the Q3 numbers had landed. LTM EBITDA was $793mm against $789mm the prior quarter, net income had turned positive at $151mm after a $412mm LTM loss, cash was up to $613mm and net leverage was down to 4.6x. Opteon refrigerant sales grew 80% year over year in the quarter, against 65% the quarter before, and the 2026E Opteon revenue mix moved from 60% to 70%, so the deleveraging leg of the thesis wasn't a projection anymore.

That mattered more for the 2029s than for the 2028s. The market was charging its premium for a litigation tail that would resolve over years, so the premium was concentrated in tenor, and with the fundamentals visibly improving, the additional duration was worth owning. By December the 2028s stood at +432, which meant the 2029s offered 88bp more OAS for less than a year of additional spread duration.

The Alpha Challenge didn't allow a free pick. It screened its eligible universe down to US public issuers rated BB+ to CC+ with at least $300mm outstanding, yielding above 5% and maturing between 30-Oct-2029 and 31-Oct-2035, and the 2028s missed the maturity floor by eleven months, so they weren't eligible. Chemours had two bonds on the list, the 4.625% 2029s and the 8% 2033s, and the November work had already measured the wider residual on the 2029s, at +200 against +150 on the 2033s. The screen removed the 2028s, and the November residuals picked the replacement.

In December, then, I moved out the curve and pitched the 4.625% 2029s as a buy at 87.1 and +520, with a target of +428 for an 11.2% total return, which meant taking more spread and more duration for the same 92bp of compression.

How It Played Out

Marked as of 08-Aug-2026.

Both bonds are through their target price with months still left on a one-year horizon. The 2028s returned about 8.3% in nine months, made up of 3.61 points of price and 4.33 of carry on a 95.27 basis, and the 2029s returned about 12.0% in eight months, from 7.34 points of price and 3.15 of carry on 87.10, which annualizes to something near 18%.

Holding the start date constant gives a cleaner test, so the table below prices them off the December cap table, when both bonds were live options on the same screen:

The issuer, the seniority and the window are identical, so the 390bp difference comes down to the tranche decision alone. The curve also compressed in the shape the regression predicted, with yields falling 209bp on the 2029s (where the measured premium was widest), 126bp on the 2028s and 66bp on the 2033s.

Part of what the tranche decision bought was headroom. The 2028 indenture calls at 101.917% from November 2025 and 100.958% from November 2026, so the upside on that paper was bounded near 101 regardless of how the credit performed, and the covenant appendix in the November deck flagged this at the time, in a note that the near-term call premium caps upside on a short-term hold. The ceiling never bound (the 2028s reached 98.88 against a 100.958 call), but it was there from the start, and the 2029s at 87.1 had nothing like it above them. The XPO memo raises the same objection to that issuer's unsecured notes, where the return is capped by the call schedule rather than by the credit.

Two caveats, though. Those are yield moves rather than spread moves, so part of the compression is the Treasury curve rather than credit, and a clean attribution would need the curve on both dates (the price and total return figures don't depend on that). Separately, the 2028 tranche has shrunk from $782.6mm to $594.6mm, which means Chemours has been retiring this maturity with cash, and that's the deleveraging path the base case required.

Why the Bonds Were Cheap

Opteon economics. Thermal & Specialized Solutions sells low-GWP refrigerants into a patent-protected duopoly with Honeywell that runs through 2030, against mandated phase-downs of high-GWP HFCs under the U.S. AIM Act and the EU F-Gas Directive. Customers' HVAC and automotive systems are engineered around a specific refrigerant, so switching costs are high and it's regulation that schedules demand. The Corpus Christi expansion was already complete and had lifted Opteon capacity 40%, with the capex spent and the added revenue still to come, and we modeled TSS going from $571mm of segment EBITDA in 2024 to $738mm in 2026E at a 33% margin.

The cyclical segment was already at the bottom. Titanium Technologies averaged 23% EBITDA margins from 2013 to 2022 and was running near 7% on pigment prices at cycle lows, with housing starts down roughly 28% from the 2022 peak. The base case assumed no recovery and held TT margins at about 10%, against that 23% historical average, and the credit still deleveraged to mid-3x on TSS growth alone, with help from a $250mm cost-reduction program and the Taiwan plant closure. Every $100/ton move in pigment is worth about $110mm of annual EBITDA, so the upside was material, but the bonds worked without it.

PFAS exposure was bounded and documented. The MOU with DuPont and Corteva caps shared liability at $4B gross with a 50/50 cost split, and $2.6B of that was already spent, which left roughly $1.4B of shared headroom plus $225mm of unused insurance. The $875mm New Jersey consent order is a nominal figure over 25 years, roughly $500mm NPV with half of it to Chemours, and it requires no out-of-pocket cash before 2030 because insurance monetization and existing escrow fund it. Our base case had litigation and remediation cash at $50mm a year and the bear case had it at $125mm, and against $1.5B of liquidity neither number threatens a 2028 or 2029 maturity.

Isolating the PFAS Premium

The relative value argument had to survive the objection that Chemours trades wide because it deserves to, so instead of eyeballing a comp table we regressed bond spread on net leverage and interest coverage across the specialty chemicals and materials complex, correcting for maturity.

The model fit at an R² of 0.841 and predicted +242 for the 2028s, against +394 actual. The residual of about 150bp is what the market was charging for PFAS and TiO₂ headline risk beyond what the credit metrics justified, and the trade rests on that number. Measuring it made the position falsifiable, since the bonds only needed the premium to compress toward the fitted curve, and it's also what made the comparison across the curve possible, which is where the December switch came from.

Chemours traded wide of what its credit metrics justified

Running the same model forward on 2026 base case metrics gave +201, but we weren't willing to assume the premium had disappeared, so we added back 100bp of residual PFAS charge, and that produced the +301 target on the 2028s. The bear case ran the same model in reverse, at 6.5x leverage and 1.2x coverage for a fitted +497, then widened the premium to 200bp for +697, and it still cleared a positive total return, because a 5.75% coupon on a three-year bond carries a lot of income against a price decline.

Downside

The downside has two independent floors, since the litigation tail is the one risk that doesn't respond to operating performance.

Going concern. At 6.5x on $910mm of base case EBITDA, enterprise value is $5.9B against $1.5B of secured claims, and the unsecured recovers par. If you stress the multiple to 5.5x on $577mm of bear case EBITDA, the unsecured recovers 50%. And if you capitalize $2.8B of PFAS claim value ahead of the notes as an unsecured pari claim, that still leaves 87% recovery on base case enterprise value, and 33% in a bear case with the tail on top.

Liquidation. If you mark assets down to 70 to 80% recovery on inventory and 35 to 45% on net PP&E, and run trustee fees, wind-down and contingency through the waterfall, the senior secured is covered in full and total unsecured recovers 32 to 44%. That's the floor, and neither 95 nor 87 is close to it.

Documentation

The notes are senior unsecured, structurally subordinated to the credit facilities and guaranteed by all material domestic restricted subsidiaries. What protects the unsecured cushion is the secured debt cap, which limits secured debt to the greater of $3.20B or a 2.50:1.00 consolidated net secured leverage ratio, with a general lien basket that permits a further 15% of consolidated net tangible assets outside the primary facility capacity. In the bear case secured net leverage peaks near 1.9x, which is inside that test, so the cap leaves room, and because the cap is a defined limit, secured debt can only rank ahead of the notes up to the basket.

Two of the events of default sit at low thresholds for a credit that's carrying active litigation. Cross-acceleration triggers at $100mm of other debt accelerated on default, and judgment default triggers at a $100mm unstayed judgment, net of insurance, that goes unpaid for 60 days. That's why the settlement structure mattered as much to the thesis as the settlement amounts did, since 25-year payment schedules and insurance-funded near-term obligations keep any single adverse outcome away from the judgment threshold instead of relying on the total staying small.

The change of control put is weaker than it reads, because although it requires a repurchase offer at 101% plus accrued interest, it only does so on a double trigger, meaning a change of control followed by a downgrade or ratings review within 60 days. A reporting failure becomes an event of default after 120 days.

What Would Have Broken It

The most direct bear path was TiO₂ going from depressed to recessionary. TSS at $738mm of 2026E segment EBITDA more than covers TT at $237mm, though, and anti-dumping duties in the U.S. and EU put a floor under pigment pricing against Chinese supply.

The real tail is the MOU cap. Below $4B of gross claims Chemours pays half, and above it Chemours pays all of the excess, with no further cost-sharing and only insurance as a cushion. The unsettled matters that could reach it are AFFF personal injury claims in the MDL, roughly 100 opt-out water systems and state natural resource damage claims. All of them are disclosed as reasonably possible but not accruable, which is why the market charged a premium for them, and why our argument was that the premium was too large rather than that it was unwarranted.

Extending from the 2028s to the 2029s meant buying more of that tail, and that's the cost of the December decision. A 3.5-year spread duration compounds a widening as efficiently as it compounds a tightening, so the bear case on the 2029s is worse than on the 2028s.

Attribution

I pitched this as part of Team Delta for UCLA Anderson at the Fink Center Credit Pitch Competition on 07-Nov-2025, with Hailee Arst, Gus Guenther and Andrew McAllister.

The competition is national, with a field that includes Wharton, Chicago Booth, Columbia, Yale, NYU Stern, Kellogg, Haas, Darden, Tuck and London Business School, and each team pitches a single long or short in a specific corporate credit instrument rather than a company. The Fink Center records the 2025 to 2026 result as UCLA Anderson first, Columbia second and USC Marshall third.

I pitched the credit again at UNC Kenan-Flagler's Alpha Challenge on 03-Dec-2025, and presented that one solo. The Alpha Challenge is also national and in its 22nd year, and it runs credit and equity tracks for MBA teams from top programs. I can speak to every number in both decks.

Both decks are published with this piece. The UCLA deck is on the 2028s and has the quarterly model, the covenant analysis, the settlement history and a doomsday cash-drain scenario, and the UNC deck is on the 2029s.