CoreWeave, a company that rents computing power to artificial-intelligence businesses, is expanding the electricity capacity behind its services. Its November 10 results report more than 4.6 gigawatts of contracted power and more than 1.7 gigawatts active at the end of September.
Daylight Post found that the difference between these two measures remained large as the business expanded from June to September. The comparison shows how much of the growth story depends on future infrastructure, alongside the capacity already operating.
A contract is one stage of building a service
AI means software that can do tasks such as producing text or analyzing images. Developing and running it uses computers in data centers. A cloud company rents access to those computers, so its customers do not have to build the whole system themselves.
CoreWeave reports contracted power separately from active power. The first includes capacity secured through agreements; the second describes its active footprint. Gigawatts measure power capacity, not the electricity used over a year. The numbers should not be read as a household-style electricity bill.
Its June results listed about 3.7 gigawatts contracted and 1.5 active. Both measures rose in the next quarter, but they continued to describe very different-sized operations.
| Quarter end | Contracted power | Active power |
|---|---|---|
| June 30, 2026 | About 3.7 | 1.5 |
| September 30, 2026 | More than 4.6 | More than 1.7 |
These are the company’s reported figures, including their rounding. Future capacity does not all have to be operating today: contracts can support planned expansion. What customers ultimately need is an available service at the agreed time.
Delivery sits behind the order book
CoreWeave reported a $132.5 billion revenue backlog. That is expected future revenue under committed customer contracts, subject to delivery and service-availability requirements. It is not cash already earned. The company expects more than 1.85 gigawatts of active power by year-end.
Its earlier June regulatory filing explains the practical dependencies. Electricity companies, grid operators and other suppliers help determine whether expansion sites can obtain enough power. The company warned that shortages or delivery failures could affect its ability to serve customers.
For an organization buying AI services, this adds another supplier dependency to the software and chip choices examined in our reporting on control over AI infrastructure. Equipment, buildings and electricity have to work together before a customer can use the promised computing capacity.
Why the wider electricity system matters
The International Energy Agency, an intergovernmental energy research body, projects global data-center electricity consumption roughly doubling between 2025 and 2030. It identifies grid connections, equipment and financing among the constraints on expansion.
Its analysis also explains why rising demand does not always mean higher bills for everyone else. Steady new demand can spread existing system costs across more users. Where supply is tight, new demand can require investment that raises costs.
The agency says policymakers determine how those costs are shared, including through electricity tariffs. For households and public services using the same power system, the terms for funding new capacity matter alongside how quickly it is built.
