Summary

  • CoreWeave, Inc put "revenue backlog" at approximately $132.5 billion as of 30 September 2026 and defined it in a footnote as remaining performance obligations plus other estimated amounts under committed customer contracts, subject to delivery and availability.
  • Its Form 10-Q prints $132.3 billion of unsatisfied RPO at the same date — 36 per cent expected within 24 months, 38 per cent in months 25 to 48, the rest to month 86, and no twelve-month figure anywhere.

Ask CoreWeave how much future revenue is under contract and the company gives two answers, two days apart, both as of 30 September 2026. On 10 November its results release put revenue backlog at approximately $132.5 billion. On 12 November its Form 10-Q put unsatisfied remaining performance obligations at $132.3 billion. The difference is about $0.2 billion, and the more useful fact is structural rather than numerical: the release's own footnote defines its headline as containing the filing's figure.

These are not rival numbers and not a contradiction. They are one book of contracted business printed twice — once as a management aggregate whose inclusion rules the company chooses, once as the accounting obligation that aggregate is built on — with the joining left to the reader. The joining matters because the conditions that decide whether $132.3 billion ever becomes revenue are printed in the same note the headline never touches.

What the release says its number is

The sentence is short; the footnote carries the weight. "Revenue backlog 1 was approximately $132.5 billion as of September 30, 2026." Footnote 1 reads: "Revenue backlog includes remaining performance obligations, plus other amounts we estimate will be recognized as revenue in future periods under committed customer contracts, in each case, subject to the satisfaction of delivery and availability of service requirements." https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Third-Quarter-2026-Results/

That is three decisions in one sentence. The population is RPO plus amounts the company estimates will become revenue under committed contracts — so the accounting measure is a stated component, not a comparator. The condition attaches to both halves: delivery, and availability of service. And "approximately" sits in front of a figure carried to four significant digits, a precision signal that matters the moment anyone subtracts one printed number from the other.

What the filing says its component is

The 10-Q's definition is the tighter instrument, and its tightening is mechanical. "RPO represents the aggregate amount of the transaction price, net of estimated variable consideration, allocated to performance obligations not delivered, or partially undelivered, as of the end of the reporting period." The variable consideration is then itemised: estimates of future potential credits to customers under availability of service agreements, amounts that may not be recognised as revenue due to delivery delays, and estimates of committed cloud computing capacity that the company has the right to resell. "RPO includes both billed and unbilled consideration from the Company's committed contracts." https://www.sec.gov/Archives/edgar/data/1769628/000176962826000465/crwv-20260930.htm

Read those clauses as a list of ways the accounting number is smaller than the contracted promise. Availability credits reduce the transaction price. Delivery delays reduce it. Capacity the company may resell is treated as variable rather than committed. The filing does not present these as warnings; it presents them as measurement. It is the difference between counting what customers have promised and counting what the company expects to be paid for delivering.

The subtraction is mine, and it is small

$132.5 billion less $132.3 billion is about $0.2 billion. Expressed against the obligation, that is roughly 0.15 per cent. I am labelling both because neither appears in either document, and because the residual carries the rounding built into the word "approximately" — which is why I will not read it as the value of the "other amounts" bucket. No document read itemises that bucket. A reader who wants its size needs the company to print it.

So the headline and the component are, on the printed figures, almost the same number. That is the first surprise: the famous gap between management measures and accounting measures is not where this story lives. The gap is in what the two numbers are conditioned on, and in what neither document will do with them.

The calendar stops at month 86

The filing prints duration in three bands and no more: 36 per cent of the obligation expected to be recognised over the initial 24 months ending 30 September 2028, 38 per cent between months 25 and 48, and the remaining balance between months 49 and 86. My own subtraction puts that remainder at 26 per cent, and 64 per cent of the obligation beyond the initial 24 months. Both are my arithmetic from printed percentages; the filing prints 36, 38 and "the remaining balance".

Notice what the band structure refuses to give anyone. There is no twelve-month figure. Anyone quoting a next-four-quarters number from this disclosure is manufacturing it, because the shortest printed window is two years. The obligation is also dated to month 86 from the balance-sheet date — a seven-year horizon printed as a percentage ladder, not as a contract schedule.

The client list is measured on revenue, not on the book

Here the disclosure changes basis entirely. The concentration table is headed "The following customers accounted for 10% or more of the Company's revenue for the periods presented." Customer A is printed at 24 per cent of revenue for the quarter and 33 per cent for the nine months; Customer B at 30 per cent and 26 per cent; the same letters show 67 per cent and 70 per cent for A a year earlier, and an asterisk for B, which the filing defines as "Customer did not represent 10% or more of revenue." Customers C and D carry asterisks in every period.

Accounts receivable tells a different split again: 18, 33, 12 and 14 per cent across A to D.

Then the caveat that forbids the obvious year-on-year story: "The customer references of A through D may represent different customers than those reported in a previous period." Letters are labels, not identities. A 70 per cent to 33 per cent move in Customer A's revenue share is a fact about a label, and the filing says the label may have changed occupant.

And the sentence that ties the client base to the contracted book without ever measuring one against the other: "Revenue recognized related to customer commitments, including revenue from delivering capacity prior to commitment start dates, represented 98% of total revenue for the three and nine months ended September 30, 2026 and 2025." Nearly all revenue arrives under commitments — some of it from capacity delivered before the commitment's start date. The $132.3 billion, by contrast, is never broken down by customer in any passage read. Concentration is measured on flows, not on the stock.

A quarter in which $1.9 billion of the book became cancellable

The sharpest line in the filing is not about size. "Additionally, as of September 30, 2026, $1.9 billion of the deferred revenue balance from the beginning of the period was reclassified from deferred revenue to customer liabilities, which are included in other current liabilities, as contracts became cancellable during the quarter."

Follow the movement, not the amount. Money that had been sitting as deferred revenue — a liability that converts into revenue as service is delivered — moved into customer liabilities inside other current liabilities, because the contracts behind it became cancellable. The passage names no contracts, no customers and no trigger. It simply records that, inside one quarter, $1.9 billion of the contracted book changed accounting class, and the destination implies a different future: a cancellable customer liability rather than revenue waiting to be recognised.

Anyone reading the aggregate as a schedule of revenue to come has to hold that sentence beside it.

What the documents qualify, and what they do not

Both documents flag the forward-looking character of what they print.

The release's forward-looking list ends with "other estimated amounts included in our revenue backlog figure", and adds that expectations "may not materialize, and actual results in future periods are subject to risks and uncertainties that could cause actual results to differ materially from those projected." The filing lists "the percentages of remaining performance obligations that we expect to recognize as revenue over respective future periods" among its forward-looking statements and states: "We cannot assure you that the results, events, and circumstances reflected in the forward-looking statements will be achieved or occur."

What I did not find matters as much, and I will state it as a boundary rather than a discovery. In the pages read, no sentence says in terms that backlog or RPO is not a guarantee of future revenue. The filing's risk factors and MD&A were not returned as full text in any retrieval, so the absence holds only for the passages actually read. Likewise, no document read prints customer concentration against the obligation itself.

That leaves a number the market quotes and a number an accountant recognises, separated by two definitions, one date, a rounding word, and a footnote about cancellable contracts. CoreWeave's contracted book is real and enormous on either measure. What it is not is self-explaining: the release gives the total, the filing gives the terms, and no page gives both. For the company's directory entry on this desk, see https://btw.media/en/directory/coreweave-inc.