What are the key learnings from the rapid growth of the managed inference offering, and how does CoreWeave decide to allocate capacity to it versus traditional take-or-pay deals? Additionally, what factors contribute to the 5-10% better margins on recent deals?

The managed inference product's rapid growth (from $1M to $100M ARR in a quarter) is seen as an opportunity to offer bleeding-edge compute and maximize value from GPUs coming off contract. The market is deep, and CoreWeave has an embedded advantage due to its control over silicon. The improved margins are a combination of the platform's higher value to customers (quality, reliability, TCO) and customers' willingness to pay more as they monetize their products. The market is pricing CoreWeave's premium product in a premium fashion. “The infrastructure that we deliver through the CoreWeave cloud is more valuable to our customers than any other solution that they can encounter.”

Given increasing local opposition to data centers, what is CoreWeave's confidence level in deploying over 3 gigawatts of active power by the end of next year and reaching 8 gigawatts by 2030?

CoreWeave believes moratoriums will shift where infrastructure is built, not impact demand. Their approach involves collaborative engagement with communities, local governments, utilities, and policymakers, ensuring grid upgrades are paid for and contributing to local economies through jobs and taxes. They are comfortable with their current roadmap, with 4.2 gigawatts contracted and 1.5 gigawatts of powered land options already secured, putting them well on track for the 8 gigawatts by 2030 target. “Ultimately, at the end of the day, it is in our interest, and it is in their interest for us to be good neighbors of their community.”

How does CoreWeave's fleet need to evolve to support the growing importance of inference, which requires more CPU and storage? Also, how does CoreWeave view the market evolving between edge, smaller clouds, neoclouds, and hyperscalers?

CoreWeave builds 'AI infrastructure' that seamlessly supports both training and inference, ensuring all necessary components are available for the full AI loop. They believe their infrastructure can effectively serve both edge workloads and those less latency-sensitive. The company leverages its position at the epicenter of information flow from various customer types (hyperscalers, labs, enterprises) to inform its strategy. Despite increased competition, demand, pricing, and margins are all expanding, signaling growth in their product and the overall massive TAM. “when you're building infrastructure, you don't build infrastructure for training, and you don't build infrastructure for inference. You build AI infrastructure.”

CRWV

CoreWeave (CRWV) Q2 FY26 earnings call: guidance and Q&A

Q: What is the typical customer intent for contract periods on renewed older-generation equipment, and what is the magnitude of this opportunity? Also, how is CoreWeave managing its broader supply chain beyond the Solidigm agreement to ensure future supply?

Management noted that older GPU generations retain significant value for various AI use cases, citing a 2020 A100 GPU contract extending to 2029 at an attractive price. The capacity coming up for renewal is a limited part of the fleet, and ASPs for older generations remain high. They also highlighted the managed inference product as a way to monetize these older fleets. For supply chain, CoreWeave maintains deep relationships with ODMs, OEMs, NVIDIA, and memory suppliers, strategically sourcing inputs to ensure product delivery. The value of CoreWeave's output has outpaced input cost increases, leading to margin expansion. “the fact that we have been able to go ahead and sell a GPU whose architecture was from 2020 in a contract that was fully priced out to 2029, really provides some insight into what the future is going to look like as this infrastructure comes off contract.”

Earnings call· Jun 2026 (Q2 FY26)· Information Technology· Aug 11, 2026

Source: earnings call transcript

Executive summary

CoreWeave Q2 FY26 — Record Revenue and Expanding Margins Driven by AI Demand

CoreWeave delivered an exceptional quarter, marked by record financial performance and significant capacity expansion, driven by intensifying demand for AI cloud infrastructure. The company demonstrated strong operating leverage and rapid growth in its managed inference platform, while strategically managing its supply chain and securing substantial financing. Management expressed confidence in its long-term growth trajectory and ability to meet evolving customer needs across the full AI life cycle.

Highlights

Generated record revenue of $2.6 billion, up 112% year-over-year.
Increased revenue backlog to $104.2 billion, with an additional $25 billion in net new customer commitments in early Q3.
Reached 1.5 gigawatts of active power, adding nearly 500 megawatts in Q2, the most in company history.
Adjusted operating income grew to $128 million, with margins expanding meaningfully.
Managed inference platform booked ARR grew from $1 million to over $100 million in a single quarter.

Concerns

Reported a net loss of $626 million for Q2, compared to $290 million in Q2 FY25.
Interest expense for Q2 was $640 million, significantly up from $267 million in Q2 FY25 due to increased debt.
Capital expenditures in Q2 totaled $9.4 billion, slightly above the guided range due to accelerated customer deliveries.

Guidance & targets

Category	Target	Timeline	Confidence
Active power	more than 1.85 gigawatts high materiality	FY26	High
Q3 Revenue	$3.45 billion to $3.6 billion high materiality	Q3 · FY26	High
Q3 Adjusted operating income	$200 million to $260 million medium materiality	Q3 · FY26	High
Adjusted operating margin	low teens high materiality	Q4 · FY26	High
Q3 Interest expense	$860 million to $940 million medium materiality	Q3 · FY26	High
Q3 CapEx	$11.5 billion to $13.5 billion high materiality	Q3 · FY26	High
Full-year 2026 Revenue	$12.4 billion to $13.2 billion high materiality	FY26	High
Full-year 2026 Adjusted operating income	$960 million to $1.15 billion high materiality	FY26	High
Full-year 2026 CapEx	$35 billion to $39 billion high materiality	FY26	High
End of year 2026 Annualized Run Rate Revenue	$18.5 billion to $19.5 billion high materiality	FY26	High
Managed inference platform ARR	at least $250 million medium materiality	FY26	High
Active power	at least 8 gigawatts high materiality	FY30	High
Demand vs. Supply	meaningfully exceed supply high materiality	multi-year horizon	High
International markets growth	major driver of growth medium materiality	future	Medium

CRWV operating KPIs by quarter

CRWV operating KPIs stated on its earnings calls, by fiscal quarter
KPI	Mar 2025 Q1 FY25	Jun 2025 Q2 FY25	Sep 2025 Q3 FY25	Dec 2025 Q4 FY25	Mar 2026 Q1 FY26	This call Jun 2026 Q2 FY26	Change vs prior quarter
Data centers	33 Today, we benefit from a network of 33 purpose-built AI data centers across the US and Europe, supported by 420 megawatts of active power. Source: earnings call transcript	—	—	43 We ended the year with more than 850 megawatts of active power, adding approximately 260 megawatts in the fourth quarter alone across 43 active data centers, up from 32 at the start of the year. Source: earnings call transcript	<50 We have navigated this complexity across close to 50 data centers, consistently delivering best-in-class dock to live times for customers. Source: earnings call transcript	—	—
Active power	420 MW Today, we benefit from a network of 33 purpose-built AI data centers across the US and Europe, supported by 420 megawatts of active power. Source: earnings call transcript	<470 MW To that end, we ended the quarter with nearly 470 megawatts of active power, and we increased total contracted power approximately 600 megawatts to 2.2 gigawatts. Source: earnings call transcript	~590 MW We expanded our active power footprint by 120 megawatts sequentially to approximately 590 megawatts, while growing our contracted power capacity over 600 megawatts to 2.9 gigawatts. Source: earnings call transcript	—	—	—	—
Total contracted power	~1.6 GW Our total contracted power extends to approximately 1.6 gigawatts, providing us with a durable multiyear runway in power capacity. Source: earnings call transcript	2.2 GW To that end, we ended the quarter with nearly 470 megawatts of active power, and we increased total contracted power approximately 600 megawatts to 2.2 gigawatts. Source: earnings call transcript	2.9 GW We expanded our active power footprint by 120 megawatts sequentially to approximately 590 megawatts, while growing our contracted power capacity over 600 megawatts to 2.9 gigawatts. Source: earnings call transcript	3.1 GW+ We contracted close to 2 gigawatts of additional power in 2025, ending the year with more than 3.1 gigawatts of contracted capacity, virtually all of which we expect to come online by the end of 2027. Source: earnings call transcript	3.5 GW+ Three, we have reached hyperscale with more than 3.5 gigawatts of contracted power, up more than 400 megawatts this quarter alone, with the substantial majority expected to be online by the end of 2027. Source: earnings call transcript	3.7 GW Contracted power grew to 3.7 gigawatts in Q2, and since quarter end, we have added roughly 500 megawatts, bringing contracted power to 4.2 gigawatts as of today. Source: earnings call transcript	—
Incremental contracted power	~300 MW In Q1, we added approximately 300 megawatts of incremental contracted power to our portfolio. Source: earnings call transcript	~600 MW To that end, we ended the quarter with nearly 470 megawatts of active power, and we increased total contracted power approximately 600 megawatts to 2.2 gigawatts. Source: earnings call transcript	—	—	400 MW+ Three, we have reached hyperscale with more than 3.5 gigawatts of contracted power, up more than 400 megawatts this quarter alone, with the substantial majority expected to be online by the end of 2027. Source: earnings call transcript	—	—
Active power	—	—	—	—	1 GW+ We generated approximately $2.1 billion of revenue, up 32% quarter-over-quarter and 112% year-over-year, and surpassed 1 gigawatt of active power. Source: earnings call transcript	1.5 GW We continue to execute on our power strategy, reaching 1.5 gigawatts of active power, adding nearly 500 megawatts, more than any quarter in our history and more than tripling year-over-year. Source: earnings call transcript	—

Operating figures the company stated on at least two of the calls shown, checked against each call's transcript. Click a figure to read the sentence. A dash means it was not stated that quarter.

Orderbook & backlog

Revenue backlog $104.2 billion Q2 FY26 end

up 246% year-over-year

More than 50% is attached to a contract where customer delivery has commenced; expected to reach more than 2/3 of Q2 backlog by end of FY26.

Net new customer commitments $25 billion early Q3 FY26

Added in the early weeks of Q3, not included in the Q2 backlog figure.

Product announcements

Product	Type	Timeline
AI platform capabilities	launch	Q2 FY26
CoreWeave ARIA	launch	—
CoreWeave Omni	launch	scaling in 2027
NVIDIA Vera Rubin NVL72	milestone	Q2 FY26
MLPerf records	milestone	Q2 FY26

Deals & partnerships

Caterpillar Deploy NVIDIA's Vera Rubin platform for physical AI training and inference.

Caterpillar will use CoreWeave's AI cloud infrastructure as its data factory to train specialized models for autonomous construction equipment.

Isomorphic Labs New customer in life sciences.

CoreWeave will support Isomorphic Labs in their mission to solve diseases.

Flow Traders New customer in financial services (systematic trading firm).

Using CoreWeave to develop and deploy next-generation AI models for quantitative trading.

IMC New customer in financial services (systematic trading firm).

Using CoreWeave to develop and deploy next-generation AI models for quantitative trading.

Leidos Collaboration for CoreWeave Federal.

Working to accelerate the delivery of secure AI capabilities for defense, national security, and intelligence missions.

Decart Using CoreWeave platform to develop Oasis 3.

Decart is developing Oasis 3, the first API accessible world model for physical AI, on CoreWeave's platform.

IBM Using CoreWeave for AI experimentation.

IBM is using CoreWeave to experiment securely with reinforcement learning, agent tool use, and model evaluation.

Nissan Monolith field engineers work alongside customers.

Through Monolith, CoreWeave's specialist field engineers work directly with Nissan to accelerate the development of enterprise-ready AI applications.

ZF Monolith field engineers work alongside customers.

Through Monolith, CoreWeave's specialist field engineers work directly with ZF to accelerate the development of enterprise-ready AI applications.

Solidigm Long-term agreement for critical inputs.

Agreement to derisk access to critical inputs needed to serve customers, illustrating supply chain management.

Capital programs

Self-build data centers underway

First several self-builds are well underway, with the first site expected to come online later this year, forming the foundation for deeper vertical integration.

APAC expansion (Indonesia) underway

Benefit:360 megawatts

Contracted 360 megawatts in Indonesia, part of over 1 gigawatt of power contracted outside the United States, expected to begin coming online in approximately 18 months.

Risks & headwinds

Regulatory backdrop for data centers (local opposition/moratoriums) Current and ongoing

Not quantified, but noted as impacting where infrastructure gets built.

Mitigation:Collaborative engagement with communities, local governments, utilities, and policymakers; ensuring payment for grid upgrades; creating construction and long-term jobs; contributing to the tax base.

Global supply chain complexity Ongoing

Not quantified, but noted as a challenge.

Mitigation:Operational discipline and leveraging partner relationships, including new ones like Solidigm, to strategically source required inputs.

Increased interest expense Ongoing, expected to be $860 million to $940 million in Q3 FY26.

$640 million in Q2 FY26, up from $267 million in Q2 FY25.

Mitigation:Securing financing at lower weighted average cost of debt (reduced by 300 bps over past year, saving ~$1.1 billion annually).

What to watch in Q3 FY26

Managed inference platform ARR

by year-end 2026

Current $100 million

Target at least $250 million

Why it matters

Indicates the rapid monetization and scalability of CoreWeave's consumption-based AI services, a key growth driver.

We expect to exit 2026 with at least $250 million of managed inference ARR.

Find out whether CRWV delivers this. Inve re-checks your case every morning.

Q&A highlights

What is the typical customer intent for contract periods on renewed older-generation equipment, and what is the magnitude of this opportunity? Also, how is CoreWeave managing its broader supply chain beyond the Solidigm agreement to ensure future supply?

Management noted that older GPU generations retain significant value for various AI use cases, citing a 2020 A100 GPU contract extending to 2029 at an attractive price. The capacity coming up for renewal is a limited part of the fleet, and ASPs for older generations remain high. They also highlighted the managed inference product as a way to monetize these older fleets. For supply chain, CoreWeave maintains deep relationships with ODMs, OEMs, NVIDIA, and memory suppliers, strategically sourcing inputs to ensure product delivery. The value of CoreWeave's output has outpaced input cost increases, leading to margin expansion.

“the fact that we have been able to go ahead and sell a GPU whose architecture was from 2020 in a contract that was fully priced out to 2029, really provides some insight into what the future is going to look like as this infrastructure comes off contract.”

asked by Samik Chatterjee · answered by Michael Intrator

2 min read 5 chapters

Detailed narrative

Broadening Demand and Customer Adoption

CoreWeave is observing a significant broadening of AI demand across various sectors, geographies, and GPU architectures. AI is now embedded in software, industrial systems, financial markets, and national security missions, moving beyond frontier model labs. The company highlighted new customer engagements with Caterpillar for physical AI training, Isomorphic Labs in life sciences, Flow Traders and IMC in financial services, and a collaboration with Leidos for CoreWeave Federal.

Additionally, CoreWeave Omni secured its first deal, expected to scale in 2027, indicating expansion beyond core infrastructure offerings.

Continuous AI Development Tools and Platform Innovation

The company's AI-native platform is designed to support the continuous AI life cycle, where training, inference, evaluation, and improvement form an integrated loop. CoreWeave introduced seven new AI platform capabilities in Q2, including CoreWeave ARIA (AI Research and Iteration Agent) to accelerate experimentation.

The platform surpassed 1 billion model training runs tracked, demonstrating strong adoption of its developer tooling. This approach allows customers to compress the time from idea to production improvement, fostering deeper relationships with developers who are future AI cloud infrastructure customers.

Performance, Economics, and Industry Recognition

CoreWeave emphasizes its leadership in performance and cost efficiency, achieving new MLPerf records and the lowest cost per token for inference on NVIDIA Grace Blackwell. The platform offers a total cost of ownership estimated to be up to 47% lower than the average hyperscaler, integrating enterprise-grade security, reliability, and a broad portfolio of services.

Gartner recognized CoreWeave as a visionary in its 2026 Magic Quadrant for Cloud AI infrastructure, validating its purpose-built approach to AI cloud.

Power and Supply Chain Foundation for Growth

CoreWeave significantly expanded its active power capacity, reaching 1.5 gigawatts by adding nearly 500 megawatts in Q2, with 300 megawatts coming online in June alone. Contracted power grew to 4.2 gigawatts as of the call date, with an additional 1.5 gigawatts from powered land options.

The company is undertaking its first self-build data centers, with the first site expected online later this year, and is expanding globally with over 1 gigawatt contracted outside the US, including 360 megawatts in Indonesia. Strategic partnerships, like the long-term agreement with Solidigm, are crucial for derisking access to critical supply chain inputs.

Strategic Financing and Capital Structure

In Q2, CoreWeave raised approximately $18 billion through a combination of debt, convertibles, and equity, bringing total capital secured to date to over $32 billion. These transactions included inaugural Eurobonds and the first publicly syndicated delayed draw term loan (DDTL) backed by HPC infrastructure.

The recent DDTL 5.5 specifically supports shorter-duration customer contracts, enabling CoreWeave to serve the enterprise market at scale and accelerate its managed inference platform, while reducing its weighted average cost of debt by almost 300 basis points, resulting in approximately $1.1 billion in annualized interest savings.

Summary of the company's earnings call; see how call summaries are prepared. Not investment advice.