How can a barrel of oil be worth less than nothing?

On 20 April 2020, the expiring WTI light sweet crude contract for May delivery settled at −$37.63 a barrel, having traded as low as roughly −$40. Real contracts, really cleared, at a negative number. The contract expired the following day.

It is usually told as a story about panic. It is better understood as the storage model running out of road, and every ingredient was visible in advance.

Ingredient 1 — physical delivery, in a landlocked place

WTI is physically delivered at Cushing, Oklahoma. A long holder at expiry is obliged to take 1,000 barrels per contract during the delivery month and must have somewhere to put them. Cushing is a tank farm in the middle of a continent — the barrels cannot be left floating offshore while you think about it.

Ingredient 2 — storage was gone, not merely expensive

Demand had collapsed in the early weeks of the pandemic while supply adjusted far more slowly. Inventories built fast.

Cushing's working storage capacity is roughly 76 million barrels. By mid-April stocks were around 55 million and climbing — and, decisively, essentially all the remaining capacity had already been leased. This is the distinction that made the difference: empty capacity is not available capacity. A tank you cannot rent, because someone else has already contracted for it, is a full tank as far as you are concerned.

So the cash-and-carry trade that normally caps a contangoed curve — buy spot, store, sell the future — could not be executed by anyone. Unit 1's ceiling did not rise. It ceased to exist.

Ingredient 3 — a compulsory obligation with no bid

In the final hours of trading, the remaining longs were participants who could not take delivery: they had no tank, no lease, no barge. Their only option was to sell. The natural buyers would have been anyone with somewhere to put the oil, and there was no such person at any price.

With a compulsory obligation on one side and no bid on the other, the price did the only thing arithmetic permitted: it went below zero. Sellers paid buyers to accept the delivery obligation. That payment is the negative price.

Two further mechanical points, both factual and both usually omitted:

• The exchange plumbing did not break. In the weeks beforehand, CME Group had publicly confirmed that its systems and its options pricing model could handle negative prices. The infrastructure worked exactly as designed; the price simply expressed what the physical situation implied.
• Volume was thin. The great majority of open interest had already rolled or closed in the ordinary way. The −$37.63 settlement is a genuine cleared price on which real money moved, and it priced a small residual of contracts with no physical outlet.

The control experiment: Brent did not go negative

On the same day, Brent traded in the mid-twenties, and the June WTI contract — one month later — settled above $20. The market was not saying oil was worthless. It was saying storage at Cushing in May was unobtainable, which is a much narrower and much more precise statement.

Brent is waterborne, and its final settlement is against a published index rather than a landlocked tank farm. A cargo at sea has options — sail somewhere else, wait — that a delivery obligation at Cushing does not. The specific ingredient was landlocked physical delivery into exhausted storage, not oil in general.

The general principle

The intuition that prices have a floor at zero comes from the idea that you can always throw a thing away for free. For a physical commodity that must be stored, cannot legally be dumped, and comes from wells that cannot be switched off and on without cost and damage, disposal is expensive — and a negative price is simply the market pricing disposal.

It is not unique to oil:

• Electricity, which essentially cannot be stored and where some generation cannot be cheaply turned down, prints negative wholesale prices routinely in several markets.
• Natural gas at constrained production hubs has repeatedly settled negative when associated gas has no pipeline capacity to leave by.

Tie it back to the identity: F = S + financing + storage + insurance − convenience yield. When storage capacity is exhausted, the storage term stops being a rate and becomes a hard constraint. The model that generated an upper bound has no solution, and the price is set by whoever must not take delivery.

This is a factual account of a documented market event. It is not a prediction that anything similar will recur, and it is not a comment on any instrument.

In the data

Two different crude series cover that week, and only one of them went below zero. The daily WTI spot assessment at Cushing carries the name of a place rather than of a contract month; the continuous front-month futures series had already rolled from the expiring May contract to June. Both, as published, measured by us on 29 September 2026:

The spot assessment followed the expiring contract through zero, because both described a barrel that had to be delivered into Cushing that week. The rolled series never printed a negative number, and a chart of it, like the one below, hides the most famous price in the history of oil. Which series you are reading decides which history you think happened.

Try it now

1. The full continuous crude history is the chart above. Navigate to April 2020 — switch to Daily bars and use the range buttons to get there — and find 20 April. The famous −$37.63 is not on this chart. Read the low of that day's candle, set it beside the spot assessment in the table above, and say which contract month each of the two numbers belonged to.
2. Brent over the same days is below. Put the two weeks side by side and Measure the worst move on each. One of them is landlocked; the other floats.

3. Write one sentence explaining the difference between the two that week, using only the words delivery, storage and waterborne.