Rounds 2–3 and mediator determination

Cutoff 2026-08-15. Prepared from the four independent Round 1 position papers in this directory.

Post-run mediator refresh, 16 August 2026. The original bargaining positions below are preserved. After correcting the working-capital premise, the clearing estimate moves from 17.5% to 15% upfront creditor warrants, stepping to 22.5% if agreed cash-conversion, liquidity or PIK milestones fail. The lower upfront grant reflects better expected operating conversion; the contingent step-up and monthly working-capital bridge protect creditors if that improvement does not appear.

0. Reconciled claim pool (US$m)

Corporate gross debt US$10.3B incl. leases; adjusted net debt US$9.5B; leverage 6.74x ⇒ implied LTM EBITDA ~US$1,409M. Consistent.

OCR correction adopted (from Braskem agent, cross-checked to q2release): deck p.030 '41 = ~579–587 not 877; '81 = ~231–250 not 207. The raw deck total was overstated.

1. Round 1 openings - convergence map

2. Round 2 - concessions and counteroffers

Creditors → all. Withdraw Path A (US$3B equitization) as a live demand but keep it as the RJ alternative. Concede: the LC pay-to-play is dropped (it splits our own coalition - the LC banks are the swing bloc and we need them for the EJ threshold). Concede: warrants at 20% rather than 30% conditional on all sponsor money being art. 171 permanent equity, which Braskem and IG4 have now both offered. Hold absolutely: 4.0% cash floor, second lien, Transforma Rio suspension, CRO, diligence.

New creditor demand introduced in R2: '28 treatment. US$1,250M matures Jan-28, inside any 24-month relief period. Demand either a 10% cash paydown at close or 20% amortisation in years 3–5.

Petrobras → creditors. Moves from US$454M to US$500M cash (rights participation + PNA-only backstop) and formalises the commercial leg: 60-day payment terms on feedstock supply, worth US$225–280M of permanent working capital, delivered under the existing arm's-length supply contracts. Critical: this requires no capital injection approval, no independent laudo, no TCU-exposed related-party capital event - it is ordinary commercial terms. Total Petrobras value ≈ US$750M, at its stated ceiling, but only ~US$500M of it is a cheque.

Petrobras concedes the lien crux with its R1 innovation: a pre-signed, pre-perfected second lien over the Brazilian crackers that springs automatically at net leverage >4.5x or liquidity <US$800M. Documentation executed and escrowed at close.

Braskem → all. Accepts the AHG's June 19 protection list in full and unconditionally. Concedes creditor 3 of 11 board seats, CRO, weekly 13-week cash. Concedes new guarantees from Braskem Netherlands, Netherlands Finance, America Finance and Trading & Shipping - closing the structural gap where the RCF is guaranteed by Braskem S.A. "only". This is real consideration at near-zero cost and it buys bondholder structural parity. Holds: no lien on the crackers at close.

IG4 → all. Moves to US$300M cash + US$100M backstop. Accepts 18–20% initial warrants. Crucially, offers what no one asked for: a contingent equitization right of up to US$1.5B at Dec-2028 if net leverage >5.0x, at 30-day VWAP with a R$4.00 floor. IG4's own BATNA work shows why - at US$1.5B EBITDA it is better off crammed down (US$275–315M) than winning a clean extension (−US$154M net of its cheque). IG4 is buying a levered option on ≥US$2.0B EBITDA and is willing to pay for it with contingent equity.

3. Round 3 - final bargaining, narrowest acceptable packages

Each party's narrowest package preferred to its realistic alternative:

Creditors. Will sign at: 20% warrants + 4.0% cash floor + springing second lien + US$1.0B+ permanent sponsor equity + Transforma Rio suspended + CRO + 50% sweep. Below that, RJ. Their own arithmetic: mark ~58c, RJ ~35c net of a 15–25% process discount and 2–4 year delay, consensual package ~65–73c. Indifference ≈ 65c. They are settling well above their BATNA and they know it; the hostility is about extracting the top of the zone, not about preferring RJ.

Petrobras. Ceiling US$750M total value against verified burden-sharing, of which ≤US$550M cash. Voting must stay at 47.03% - non-negotiable, because crossing 50% triggers consolidation of ~US$10B of debt onto the Petrobras balance sheet. Warrants must therefore settle into non-voting preferred.

IG4. Will fund US$300M + US$100M backstop, accept 20% initial / 30% stepped, accept the contingent equitization. Will not go below ~35% of ON. Its hard lever: authorized capital headroom is only 355.7M shares of 1,152,937,970 - ~30.85% of a post-money. Anything larger requires an EGM, which requires an ON majority, which only Shine can deliver. Consensual equitization above ~31% is legally impossible without IG4's affirmative vote. Creditors who want more equity must go through RJ.

Braskem. Needs ≥US$900M closing liquidity and ≥US$900M of LC capacity (monthly gross feedstock/utility outflow ~US$700–900M on US$13.0B COGS). Will not grant liens on the crackers at close because that destroys the unencumbered base supporting LC and factoring - RJ damage without RJ benefits. Springing lien resolves this.

4. Why the springing lien is the Nash-improving trade

The lien dispute is not a dispute about value, it is a dispute about states of the world. Creditors want the lien for the normalization state (EBITDA US$1.5B). The company needs the assets unencumbered in the persistent state, because unencumbered assets are what support the LC and factoring lines that are its actual binding constraint. A pre-signed, pre-perfected lien that springs at 4.5x costs the company nothing in the state where it doesn't spring, and delivers creditors nearly full protection in the state where it does. Both sides gain relative to any fixed rule. Petrobras proposed it independently; IG4 proposed a variant at 5.0x. It clears.

The same logic governs the warrant ladder: the parties disagree about the spread outlook and cannot resolve that disagreement by argument. A contingent ladder prices the disagreement instead of requiring agreement. All four parties independently proposed step-ups.

5. The binding execution problem - under-weighted by every agent

An art. 171 rights offering of US$1.0–1.2B requires: independent committee valuation; EGM convocation (30 days); authorized-capital increase; a ≥30-day preemptive window; a sobras round; and for any US tranche, registration that is impractical on this timeline. Minimum ~90–120 days. The stay expires ~25 Aug. The election runoff is 25 Oct. The RCF/stand-by wall is 31 Dec.

Therefore the equity cannot be delivered before the stay lapses. The only coherent sequencing is:

1. Late Aug: EJ filing (or negotiated stay extension) to hold the ring - the creditors said in June they would support a "plan to a plan" EJ subject to protections, and Braskem has now accepted those protections in full.
2. Sept–Oct: term sheet + lock-up; Petrobras delivers the commercial leg (60-day terms) immediately, since it needs no capital approval and is politically invisible.
3. Q4/Q1: EGM, rights offering, homologation, closing.

This sequencing is what makes a deal possible at all - and its fragility is what keeps RJ probability high.

6. Brazil political economy as a binding constraint

• Election 4 Oct / 25 Oct. Petrobras cannot write a large discretionary cheque into a distressed private group in the campaign's final weeks. Window is now→late Sept, then dead until Q1-27 - which straddles the Dec-26 RCF wall. This is the single strongest argument for the commercial-terms leg over a capital injection.
• Lei 13.303 / CVM Res. 81 / TCU. Related-party review, independent laudo, minority challenge risk. Petrobras must participate strictly pro rata on terms identical to every other holder - which the art. 171 structure delivers by construction.
• "No statization, no debt consolidation." Caps Petrobras and pushes burden onto IG4, which defends by ratio (Petrobras ≥1.25× Shine) rather than absolute number.
• Industrial policy. REIQ raised to 5.8% (LC 228, Mar-26); PRESIQ effective 2027–31 (Law 15,294/25); definitive 5-yr PE antidumping duties vs US/Canada (GECEX 876/2026); 20% resin import tariff expires 16 Oct 2026 - between the election rounds. Renewal is likely on employment politics but not automatic, and is worth roughly the gap between 70% and 75%+ Brazilian utilization. It gets its own warrant step-up trigger so creditors are paid for that specific risk rather than pricing it into a control demand.
• Employment/regional politics (Camaçari, ABC Paulista, Triunfo/RS) and Alagoas make a pre-election RJ of Brazil's largest petrochemical producer politically costly - which raises the chance of an official-sector bridge (BNDES) and of judicial protectiveness in an RJ. Court protectiveness of large employers blunts the creditors' speed threat and worsens their RJ recovery.

7. Mediator determination - the stable compromise

The package below is the narrowest each party prefers to its realistic alternative, tested pairwise:

• Creditors get ~65–73c of PV against a ~58c mark and a ~35c RJ. Accept.
• Petrobras spends US$750M of value (US$500M cash) to protect a 36.15% stake, joint control, and an integrated feedstock relationship, while staying at 47.03% voting. Accept.
• IG4 pays US$300M + backstop, retains >35% ON and joint control, and buys a levered option on ≥US$2.0B EBITDA. Accept.
• Braskem gets ~US$2.2B of closing liquidity, US$1.2B of LC capacity, cash interest cut from ~US$714M to ~US$430M for 24 months, and keeps the crackers unencumbered unless it underperforms. Accept.

No party is better off at its BATNA. The compromise is stable - conditional entirely on the sequencing in §5 holding.

Revised warrant implementation: 15% fully diluted at close, stepping to 22.5% on objective misses. Required creditor reporting includes a monthly EBITDA-to-cash bridge separating operational working capital, LC runoff, supplier-finance changes, capex, interest and Alagoas cash use.

8. Leverage grid - most likely deal

Pro forma at close: gross ≈ 9,492 + 400 new money + ~90 fees = US$9,982M; cash ≈ US$2,150M ⇒ net ≈ US$7,832M. End-2028 after ~US$700M PIK accretion: gross ≈ 10,682, net ≈ 8,900.

At US$1.5B the structure still does not work - which is precisely what the warrant ladder and the Dec-2028 contingent equitization are for. That contingency is not a sweetener; it is the load-bearing element.