Aug 11, 2026 earnings call — AI demand, capacity build and margin expansion
CoreWeave reported record revenue of $2.6 billion, up 112% year-over-year.
Revenue $2.6B (est $2.6B) · EPS $-1.03 (est $-1.21) · Next-quarter revenue outlook $3.5B–$3.6B
Key quotes
"Demand continues to intensify as the market broadens across sectors, geographies, workloads and generations of GPU architecture."
"As I shared at the top of the call, we ended Q2 with 1.5 gigawatts of active power, adding close to 500 megawatts in the quarter alone."
"In the past few months since its launch, booked ARR for our managed inference platform has grown from $1 million to more than $100 million."
Analyst Q&A 6
contract lengths and supply chain access
Q (Samik Chatterjee · JPMorgan): He asked what contract lengths customers prefer now and how much of the installed equipment is coming up for renewal. He also asked about supply-chain agreements to secure more capacity.
A: Management said older-generation GPUs still have significant value and can be sold on long contracts. They said the capacity coming up for renewal is a very limited part of the fleet, ASPs on older generation remain high, and managed inference should reach about $250 million of ARR by the end of the year. They also described long-standing relationships with ODMs, OEMs, NVIDIA and memory suppliers to secure the parts needed.
"In terms of the capacity that's coming up for renewal, Samik, is a very limited part of our fleet." — Nitin Agrawal (CFO)
managed inference early lessons and margins
Q (Brad Zelnick · Deutsche Bank): He asked for initial learnings from the managed inference product and how capacity will be allocated between managed inference and long-term take-or-pay deals. He also asked how much of the 5%–10% better margin is from shorter-duration deals versus other factors and about pricing in the market.
A: Management said they are broadening the offering beyond long-term take-or-pay deals and that managed inference scaled very quickly (from $1 million to $100 million in a single quarter). They said managed inference lets them use GPUs coming off contract and can be high margin; the DDTL 5.5 financing shows capital markets will underwrite shorter-duration contracts. They also said the 5%–10% margin improvement is driven by multiple factors, including platform quality, reliability, security, TCO, and customers willing to pay a premium.
"The infrastructure that we deliver through the CoreWeave cloud is more valuable to our customers than any other solution that they can encounter." — Michael Intrator (CEO)
ARR target and data center permitting
Q (Jyhhaw Liu · Evercore): He asked whether the $18 billion to $19 billion ARR exit target still makes sense given pricing pressure. He also asked about confidence building more than 3 gigawatts by next year and the path to 8 gigawatts by decade-end.
A: Management confirmed they raised exit ARR guidance to $18.5 billion to $19.5 billion for 2026 and said that guidance is baked in. They said local moratoriums change where infrastructure gets built but not overall demand, and emphasized collaborating with communities, paying for grid upgrades and creating jobs. They added they have 4.2 gigawatts contracted today plus about 1.5 of powered land options, putting them near 6 gigawatts by mid-2026 and on track for greater than 8 gigawatts of active power by the end of 2030.
"Yes. So we increased the exit ARR number that we provided in guidance to you folks right now at $18.5 billion to $19.5 billion for 2026, so that is baked in our guidance that we provided to you." — Nitin Agrawal (CFO)
inference infrastructure and edge vs cloud
Q (Raimo Lenschow · Barclays): He asked whether the fleet needs to change (more CPU, storage) to support inference and if existing data centers can do that. He also asked how workloads will split between edge, smaller clouds and hyperscalers.
A: Management said they build AI infrastructure to serve the full AI loop, not separate systems for training and inference, so their existing infrastructure can evolve to serve inference. They said CoreWeave sees signals from hyperscalers, labs and enterprise and has built the cloud to serve both edge and non-edge workloads. They also noted that despite increased competition, demand, pricing and margins are expanding.
"We believe that when you're building infrastructure, you don't build infrastructure for training, and you don't build infrastructure for inference." — Michael Intrator (CEO)
new capacity timing and Vera Rubin impact
Q (Michael Turrin · Wells Fargo): He asked about the timing when the 300 megawatts added in June will reach steady-state contribution and the linearity of capacity adds. He also asked early signals on Vera Rubin monetization versus prior generations.
A: Management said Q2 added about 500 megawatts of power, with 300 megawatts added in June, so that backend-loaded capacity will start showing in Q3 and Q4. They said Vera Rubin is already driving margin expansion from the start, demand for the Vera Rubin platform is enormous, and much of the 5%–10% margin improvement is coming from the Vera Rubin SKU.
"Vera Rubin is a generation that is seeing the margin expansion right from the start." — Michael Intrator (CEO)
recontracting GPUs and DDTL 5.5 effect
Q (Brett Knoblauch · Cantor Fitzgerald): He asked how CoreWeave decides to recontract GPUs rolling off versus putting them on spot or inference and how far ahead they make that decision. He also asked if DDTL 5.5 changes target contract durations to capture more margin with shorter-term deals.
A: Management said they evaluate the economics and long-term obligations: some infrastructure that comes off-line is placed back into term contracts if the economics warrant it, while other infrastructure can be sold on shorter-duration contracts to extract additional margin. They said the DDTL 5 and 5.5 executions let CoreWeave populate the curve, support shorter-duration financing (clients wanting 2 years or 3 years), and open new markets for customers who prefer shorter contract cycles.
"We want to sell our compute on long-term contracts. We also want to sell it on shorter-term contracts to extract additional margin." — Michael Intrator (CEO)