Refining Capacity Is the New OPEC

TO: Risk Committee
RE: Diesel, and why everyone is still modeling the wrong barrel
STATUS: Not for distribution

Somebody on the desk keeps pulling up WTI and calling it the oil story. WTI is $85. It's been $85-ish, give or take a Hormuz headline, for weeks. That is not where the damage is. The damage is in a number most of this book has never had to price before: the diesel crack spread, which broke $106 a barrel on September 1st. Pre-conflict baseline on that spread was about $20. We are running a refining margin more than five times its historical norm, and it has nothing to do with how much crude is coming out of the ground.

Write this down, because it's the whole memo: for the first time in a generation, the binding constraint on distillate prices isn't crude supply. It's refining capacity. And refining capacity is being destroyed by drones, not depleted by decline curves.

Two wars, one product. Ukraine has run more than 70 strikes on Russian refineries since January, and Moscow responded in July by banning diesel exports outright, extended now through September 30. Russia was moving something like 800,000 to a million barrels a day of diesel into the seaborne market. That's roughly 10% of global seaborne diesel supply, gone, by policy choice, because the refineries that would have made it are on fire. Separately, the Strait of Hormuz standoff has cut Persian Gulf refined product exports to something like 40% of pre-war volume, per Bloomberg's reporting, with the IEA estimating close to 3 million barrels a day of regional refining capacity knocked offline by the strikes themselves. Add the Houthis threatening the Bab el-Mandeb corridor and the Caspian Pipeline Consortium's Black Sea loadings getting paused intermittently, and you get to the figure Goldman and others are now using: 7 to 8 million barrels a day of global refining capacity effectively offline. Simultaneously. From two unrelated conflicts that happen to share a target list.

U.S. refiners are running at 98% operable utilization trying to plug the gap. Ninety-eight percent is not a number with much room in it. There is no swing capacity sitting idle waiting to be switched on if a Gulf Coast refinery has a bad week. Every barrel of distillate capacity in this country is already spoken for, and it's being outcompeted internally by jet fuel, which pays a better margin right now and is soaking up crack capacity that would otherwise go to diesel. That's not a geopolitical problem. That's just what happens when you run a fixed-capacity industry at redline for seven months straight.

Retail diesel hit $5.85 a gallon nationally on September 4th, an all-time high, with California over $7.70. This matters to the book two ways, and only one of them is obvious. The obvious one: domestic transportation burns roughly 120 million gallons of diesel a day with no near-term substitute, so every dollar of pump-price increase is a direct, inelastic tax on freight, agriculture, and construction margins across every name we hold with distribution exposure. Model that straight through to Q3 guidance calls — Tyson already took one down citing input pressure, and they're not alone.

The less obvious one is upstream of the Fed. Diesel is a headline PPI and core-goods input in a way gasoline mostly isn't anymore, because gasoline demand has some seasonal and behavioral give to it and freight doesn't. August PPI already printed hot. The diesel crack spread hitting a fresh record on the exact same days the CPI print came in hotter than consensus is not a coincidence anyone on this desk should be treating as noise. It's a mechanical transmission line from a refinery in Ryazan getting hit by a drone to a rate-hike probability on CME FedWatch. Warsh doesn't need a wage-price spiral to justify hiking into this. He just needs the freight bill.

Positioning-wise, the market has actually gotten this half right for once, which should worry us more than comfort us. Valero, Phillips 66, HF Sinclair, and PBF have quietly been among the year's better performers while the tape stayed obsessed with AI capex, because pure-play crack exposure has been printing. Goldman's revised 2027 U.S. diesel margin forecast — up to $63 a barrel from $27 — tells you this desk isn't alone in believing the structural premium outlasts the news cycle. The refiners are pricing a multi-year regime change in their own margin structure. The equity market broadly is not yet pricing a multi-year regime change in the CPI basket that regime change feeds into.

Two caveats before anyone gets too attached to a linear "crack spreads only go up" trade. First, a chunk of the crude side of this move is arguably overdone — the 150-million-barrel release tied to the U.S.-Iran memorandum, erratic Chinese buying, and reports of irregular trading activity all point to crude being mispriced $10-15 relative to physical stock levels, meaning a chunk of any resolution gets captured on the crude leg, not the crack. Second, cracks reverse hard and fast when capacity comes back — ask anyone who was still long product exposure into the 2022-23 unwind. Refining fires get put out. Sanctions regimes get renegotiated. Drone campaigns end.

Until one of those things happens, though, model the diesel crack as a policy input, not a commodity residual. It's currently doing more work on the inflation print than crude oil is, and nobody on the macro side of this book has fully repriced for that yet.