A multiyear cloud contract can add billions to a provider’s backlog long before its data centers can run the workloads.
CoreWeave reported a $132.5 billion backlog at the end of September and $3.66 billion in third-quarter revenue. The AI cloud provider released those figures on November 10. Its November 12 quarterly filing gives a timetable for when the company expects to recognize revenue from most of that backlog.
About $47.6 billion in the first two years
CoreWeave reported $132.3 billion in remaining performance obligations at September 30. That figure represents the value assigned to work the company still owes customers, after estimates such as potential service credits. Its broader backlog also includes other amounts CoreWeave expects to recognize under customer commitments.
CoreWeave expects to recognize 36% of the $132.3 billion in the first 24 months, through September 2028. Applying 36% to the balance yields about $47.6 billion. The filing allocates another 38%, or about $50.3 billion, to the following 24 months. That leaves roughly $34.4 billion for months 49 through 86. These amounts use the rounded percentages in the filing and describe expected revenue recognition. The filing includes billed and unbilled amounts in the same balance.
From a power contract to a GPU reservation
In its earnings release, CoreWeave reported more than 1.7 gigawatts of active power and more than 4.6 gigawatts under contract. Both are company figures for power. A team booking a training run still needs to confirm the hardware, location, and date for its own reservation.
For a run that needs 256 GPUs on one high-speed network, ask for a date when the whole allocation will be available together. Check whether billing starts when the reservation is signed or when the cluster passes acceptance tests. If delivery slips, the contract should tell you whether you can move the reservation or cancel it. For more on the financing behind that buildout, see our analysis of CoreWeave’s second quarter.




