US refiners at peak margins: a deep dive on PBF Energy

Research note, data as of 24-25 June 2026. Not individual investment advice. US refining is running at abnormally high margins, and unlike past spikes this one is largely structural. We cover how much refiners earn now, how long it lasts, how much free cash flow PBF can generate, and a backtest of buying cheap on crack-adjusted EV/EBITDA. The lead name is PBF Energy (PBF) — the most concentrated way to play the theme.

The crack spread is ~$46/bbl now — double the norm, but not at its peak

The 3-2-1 crack spread — the notional margin of turning three barrels of crude into two of gasoline and one of diesel — sits around $46/bbl, versus a historical norm of ~$10-12. The headline USGC 3-2-1 averaged ~$42 in April 2026 (+95% YoY) and spiked above $52 in spring on Hormuz tension. In 2024-2025 the crack sagged to ~$20-24, which is why refiner profits collapsed.

Crack spread: where we are in the cycle

The high crack has lasted only ~4 months of a typical 3-9

The acute phase of the rally began in late February 2026 (Iran/Hormuz), so the high crack has lasted only about 4 months. Historically, elevated margins after a geopolitical shock persist 3-9 months — so the cycle is only partly through, and permanent capacity closures can extend it. Key nuance: PBF has barely monetized this crack yet, because its best plant (Martinez) was down — the bulk of the earnings is still ahead.

Strikes on Russian refineries and Hormuz add fuel — but it's reversible

Ukrainian strikes have knocked an estimated 20%+ of Russia's refining capacity offline; Russia began importing gasoline by sea and curbed product exports, and as a major diesel exporter its lost flows push global cracks up. The Hormuz/Iran conflict added a risk premium and Middle East outages. Both factors are reversible — cracks have eased before when Russian supply recovered — which argues the moment is cyclical, not a permanent new normal.

~900 kb/d shut for good — that lengthens the cycle

Since 2023 the US has permanently retired about 900 kb/d of refining capacity. Shut plants don't come back, which structurally tightens supply regardless of demand. Base case: elevated margins persist another 12-18 months, compressing 20-30% from peaks; bull case 24+ months; bear case a return to 2023 levels within 6-12 months.

US capacity permanently closed since 2023

PBF is a pure-play refiner: maximum sensitivity to the crack

PBF is a pure-play refiner — no midstream, no chemicals to smooth the cycle. Six plants, ~1.0 million b/d, weighted Nelson complexity 12.7. That makes it the most crack-geared large name — amplitude in both directions.

PBF refineries

PBF's profit swings wildly: $4.3bn at the peak, ~zero at the trough

A pure-play cuts both ways: adjusted EBITDA was $4.3bn in 2022, collapsed to $0.09bn in 2024 — a 48x swing in two years — and was near zero in 2025 (Martinez outage). At full utilization and current spreads the potential is $2.5-3.5bn a year.

PBF adjusted EBITDA by year

PBF's margin tracks the crack — depressed now by Martinez downtime

PBF's realized refining margin moved from +$11.7/bbl (Q1'24) to -$3.9 (Q4'24) to +$11.2 (Q4'25), then dipped to +$6.0 in Q1'26 — low despite a high crack because Martinez was down. As Martinez returns to full load in Q2 2026, the margin should re-converge toward the crack environment.

PBF realized refining margin by quarter

+$1/bbl of crack ~ +7% of market cap — that's the core leverage

At ~900 kb/d PBF processes ~330 million barrels a year, so every +$1/bbl of durable margin is ~$0.33bn of pre-tax EBITDA, ~$2.8/share, ~7% of market cap. Per $1bn of market cap PBF carries ~188 kb/d of capacity vs ~41 for VLO/MPC — 4.6x more barrels per dollar (but also more debt).

Leverage to the crack: capacity per $1bn market cap

At current cracks PBF can earn ~20-40% of its market cap in FCF in a year

Assuming 2026 capex ~$0.9bn, net interest ~$0.19bn and 21% tax: mid-cycle (EBITDA ~$1.8bn) yields ~11% FCF on the $4.8bn market cap; at current high cracks with Martinez running (~$2.5bn) ~22%; at a 2022-style peak (~$3.5bn) ~39%; at the trough FCF is negative. So a sustained high crack can return a fifth to two-fifths of the market cap in free cash flow in a single year.

Estimated PBF FCF yield by state of the cycle

Debt is manageable, and Martinez restarts into the high-spread window

Net debt is ~$2.3bn (net debt/cap ~36%, debt/EBITDA ~1.9x), with no maturity wall — 2026 refinancing pushed the 2028 notes out to 2034. Dividend ~2.7% plus a buyback; priority is deleveraging. Martinez (fire 1 Feb 2025, ~$0.9bn insurance recoveries) completes its restart in Q2 2026 — full load returns into the high-spread window.

Backtest: cheap on EV/(crack-adjusted EBITDA) = historically a good entry

We tested buying when a refiner is cheap on EV/EBITDA computed at the current crack (not reported earnings) across VLO/PBF/MPC/PSX/DINO, 2020-2026, monthly, with point-in-time ranking. Cheap names returned +47% on average over the next 12 months (median +45%) versus just +5% for expensive ones.

12-month forward return by cheapness on crack EV/EBITDA

Statistical weight comes from our Global Commodities strategy, whose refiner signal is built on the same crack model (crack to EBITDA to EV/EBITDA to upside): 2016-2026 it returned ~28% CAGR vs ~14% for the S&P, with a ~+11%/yr cross-sectional ranking edge over an equal-weight basket of the same names.

For cyclicals, multiples on reported earnings mislead — use anchors

For a cyclical, any multiple on reported/peak earnings misleads (P/E and EV/EBITDA alike) — at the margin peak, earnings are huge and the multiple looks low right at the top. Better anchors: EV/EBITDA recomputed at the current crack (as in the backtest), EV/mid-cycle EBITDA, and price-to-book. PBF's P/B range is 0.25x (trough), 1.12x median, 2.55x (peak); now ~0.90x, around or below book (~$44/share). Supercycle history: 2004-2005 saw Valero +239% and HF Sinclair +265%; 2022 saw HF Sinclair +106%, Valero +58%, Marathon +53%.

PBF price / book value (P/B)

Risk/reward skews up: base +45%, bull +112%, bear -28%

Combining FCF-driven equity build with multiple normalization toward mid-cycle, rough 12-24 month price scenarios for PBF (current ~$40): base (crack ~$30-35) ~$58 (+45%); bull (crack ~$40+, holds 24 months) ~$85 (+112%); bear (crack to $20) ~$29 (-28%). The risk/reward skews up, but this is a high-volatility cyclical bet — both upside and downside are large. Not a forecast.

PBF illustrative price scenarios