CoreWeave, Inc. Class A Common Stock (CRWV) Earnings

Management highlights

- Broadening Customer Demand and Sector Diversification * AI infrastructure demand has expanded beyond frontier AI model labs to embedded use cases across enterprise workflows, industrial operations, life sciences, financial services, public sector/national security, and sovereign cloud clients. * Notable new customer wins include Caterpillar (industrial physical AI), Isomorphic Labs (life sciences), Flow Traders and IMC (financial services quantitative trading), Leidos (U.S. public sector defense/intelligence), and the first CoreWeave Omni sovereign/enterprise client (to scale in 2027). * Pricing for new Blackwell and Vera Rubin SKUs hit new record highs, while pricing for prior generation SKUs remains at or above levels seen years ago, demonstrating persistent strong demand across all GPU generations. - Platform Innovation for End-to-End AI Development * Launched 7 new AI platform capabilities in Q2, became the first cloud provider to validate NVIDIA's Vera Rubin NVL72 leveraging custom software-defined liquid cooling technology, and set new ML Perf performance records for NVIDIA Grace Blackwell architecture. * Recognized as a Visionary in Gartner's 2026 Magic Quadrant for Cloud AI Infrastructure, validating the company's AI-native platform approach. * The managed inference business has grown extremely rapidly, with growth constrained only by near-term available capacity; the platform is already leader in cost per token and speed to first token for open source models. * Exceeded 1 billion cumulative model training runs tracked on the platform, with growing adoption of higher-margin AI development services that create natural long-term customer expansion paths as workloads scale. - AI-Native Architecture and Competitive Economics * CoreWeave's purpose-built AI architecture (custom power, cooling, rack design, networking, orchestration, and tooling) delivers an estimated total cost of ownership (TCO) up to 47% lower than general-purpose hyperscalers, per independent third-party analysis. * New contracts signed in Q2 have contribution margins 5 to 10 percentage points higher than contracts added in recent quarters, driven by product differentiation, premium pricing, and operating leverage from scale. - Power and Supply Chain Capacity Expansion * Added 500 megawatts of active power in Q2, the largest quarterly addition in company history, reaching 1.5 gigawatts of total active power at quarter-end, more than tripling active power YoY. * Total contracted power grew to 3.7 gigawatts by Q2 end, and an additional 500 megawatts was added post-quarter to bring total contracted power to 4.2 gigawatts as of the call; excluding unexercised options and LOIs, the company controls more than 1.5 gigawatts of additional potential power from accumulated powered land. * Expanded globally, with over 1 gigawatt of contracted power outside the U.S., including 360 megawatts in Indonesia (APAC) that will come online in ~18 months. The firm remains firmly on track to reach at least 8 gigawatts of active power by 2030. * Secured long-term supply agreements for critical components, highlighted by a new partnership with Solidigm to de-risk input access.

Guidance

- End-of-2026 active power guidance was raised to more than 1.85 gigawatts, up from the prior target of more than 1.7 gigawatts, with the company still on track for at least 8 gigawatts of active power by 2030. - Q3 2026 guidance: Revenue is expected in the range of $3.45 billion to $3.6 billion; adjusted operating income is expected between $200 million and $260 million, with margins continuing to expand sequentially and reach the low teens in Q4 2026; Q3 interest expense is projected between $860 million and $940 million; full-year 2026 capex guidance is set to $35 billion to $39 billion. - Full-year 2026 revenue guidance was raised to $12.4 billion to $13.2 billion, and adjusted operating income guidance was raised to $960 million to $1.15 billion. - End-of-2026 annualized run-rate revenue guidance was raised to $18.5 billion to $19.5 billion. - Managed inference ARR is expected to reach at least $250 million by the end of 2026.

Segment performance

CoreWeave reports consolidated financial results for Q2 2026, with segment breakdown only provided for non-GPU ancillary businesses: these businesses (including storage, CPU, networking, and software) had annual recurring revenue (ARR) exceeding $400 million as of Q2 2026, representing approximately 15.4% of total quarterly revenue. The core AI cloud infrastructure segment accounted for the remaining ~84.6% of Q2 revenue. Core consolidated results: Q2 2026 total revenue was $2.6 billion, up 112% year-over-year (YoY) and 24% sequentially. Adjusted operating income was $128 million (5% adjusted operating margin), up from $21 million in Q1 2026. Adjusted EBITDA was $1.5 billion, up 100% YoY, with a 59% adjusted EBITDA margin. Net loss for Q2 2026 was $626 million, compared to a $290 million net loss in Q2 2025. Capital expenditures (capex) totaled $9.4 billion, slightly above the prior guided range. Total revenue backlog at quarter-end was $104 billion, up 46% YoY, with over 50% of backlog tied to contracts where customer delivery has already commenced. The new managed inference offering (launched a few months prior to Q2) saw booked ARR grow from $1 million to over $100 million by Q2 end, already contributing ~3.8% of total Q2 revenue.

Risks & headwinds

- Global supply chains for critical AI infrastructure components (GPUs, memory, networking) remain complex and constrained, requiring aggressive long-term relationship management and strategic sourcing to secure sufficient inputs to meet customer demand. - Increasing local regulatory opposition and NIMBY sentiment (moratoriums on new data center development) create permitting and deployment risks for new capacity expansion. CoreWeave notes that to date, its pipeline of contracted power and expansion projects is not impacted by these headwinds, as the company engages proactively with local communities and policymakers. - Capex is front-loaded for long-term customer contracts, requiring ongoing access to capital markets to finance build-out, though the company recently completed multiple oversubscribed financing transactions to secure sufficient near-term capital. - Sustained AI demand growth has created a multi-year period of market disequilibrium where demand structurally outpaces supply of AI infrastructure, creating capacity allocation challenges even as CoreWeave expands rapidly.

Analyst Q&A

• Q: How large is the opportunity from recontracting GPUs that roll off initial long-term contracts, and what customer contract duration trends are you seeing for these assets? /

  A: Only a limited share of CoreWeave's current GPU fleet is up for renewal over the next few years. ASPs for older generation GPUs remain at or above levels seen one year ago. Off-contract older GPUs are ideal for the fast-growing managed inference business, which already has a clear path to $250 million in ARR by end of 2026. A recent example shows CoreWeave signed a 2020-vintage A100 contract extending to 2029 at an attractive market price, demonstrating strong recontracting upside. (249 characters)

• Q: How is CoreWeave addressing ongoing supply chain constraints, and will it pursue more long-term supply agreements to secure inputs? /

  A: Aggressive supply chain management is embedded in CoreWeave's DNA, with long-standing deep relationships with NVIDIA, OEMs, ODMs, and component suppliers already in place. The growth in value of CoreWeave's cloud output has outpaced input cost increases, allowing the company to maintain expanding margins even as component costs rise, with new Q2 contracts carrying 5 to 10 percentage points higher contribution margins than recent quarters. (261 characters)

• Q: How has the successful financing of shorter-duration inference contracts changed CoreWeave's target contract duration strategy to capture higher margins? /

  A: The successful closing of the DDTL 5.5 financing, the first public infrastructure-backed delayed draw term loan to include shorter-duration contracts, unlocked the ability to serve more enterprise clients with shorter contracts. This now lets CoreWeave choose the most profitable mix of contract durations across its fleet, allowing it to capture higher ASPs and margins from shorter-term deals and managed inference for off-contract older GPUs. (258 characters)

• Q: Can you confirm your confidence in hitting your 2027 and 2030 power deployment targets amid increasing regulatory pushback on new data centers? /

  A: CoreWeave currently has 4.2 gigawatts of contracted power, plus an additional ~1.5 gigawatts of power from options and executed LOIs, putting the firm near 6 gigawatts of secured power by mid-2026. CoreWeave proactively engages with local governments, utilities, and communities, covering grid upgrade costs to avoid shifting burden to local ratepayers and highlighting job and tax benefits, so none of the current expansion pipeline is impacted by regulatory pushback to date. (272 characters)

• Q: When will the 300 megawatts of active power added in late June start contributing to revenue, and what is the early outlook for Vera Rubin monetization? /

  A: The 300 megawatts added in June 2026 were back-loaded in Q2, so most revenue contribution will start flowing through in Q3 and Q4 2026. Early pricing and margin for Vera Rubin SKUs are very strong, and much of the 5 to 10 percentage point step-up in contribution margins for new Q2 contracts comes from Vera Rubin, making it a very successful high-margin SKU for CoreWeave. (227 characters)