CoreWeave, Inc. Class A Common Stock Q2 FY2025 earnings call

• Revenue surged 107% YOY to $1.2 billion with adjusted operating income of $200 million, achieving the first quarter with over $1 billion in revenue and $200 million in adjusted operating income. • Ended the quarter with nearly 470 megawatts of active power and increased total contracted power to 2.2 gigawatts, on track to deliver over 900 megawatts by year-end. • Revenue backlog reached $30.1 billion, up $4 billion from Q1 and doubling year to date, including new customer wins and expansion contracts with hyperscalers. • Continued investing in the platform, including Mission Control for reliability, a private preview of object storage, integration with Weights and Biases, and support for third-party storage systems. • Proposed acquisition of Core Scientific to accelerate value creation, expecting cost savings and enhanced infrastructure scaling. • Secured $6.4 billion in capital markets through high-yield offerings and a delayed draw term loan, lowering the cost of capital.

CoreWeave, Inc. Class A Common Stock reported revenue of $1.2 billion for the second quarter, marking a 107% year-over-year growth. Adjusted operating income stood at $200 million. The revenue backlog was $30.1 billion, which was a $4 billion increase from the first quarter and doubled year to date.

• Q3 revenue is expected to be in the range of $1.26 billion to $1.3 billion. • Q3 adjusted operating income is anticipated to be between $160 million to $190 million. • Full-year 2025 revenue guidance is raised to $5.15 billion to $5.35 billion. • Full-year adjusted operating income is expected to be $800 million to $830 million, unchanged from prior guidance. • CapEx for 2025 is expected to be $20 billion to $23 billion, unchanged from prior guidance.

• The market is structurally supply-constrained with demand outpacing supply, particularly in powered shells, GPUs, and transformers. • Potential fluctuations in tax rates due to non-deductible items and valuation allowances on net deferred tax assets. • Uncertainty regarding the reception of US-based technology by sovereign governments in some jurisdictions.

Q: Could you talk about the renewal of hyperscaler contracts and tweaks to achieve better return on assets?

A: Michael Intrator stated focusing on expansion rather than renewals, with clients typically upgrading to new hardware architectures. Nitin Agrawal discussed acquisitions like Weights and Biases and the proposed Core Scientific acquisition to drive cost savings and scale.

Q: How to think about the mix of training vs. inference and supply constraints?

A: Michael Intrator mentioned infrastructure is fungible for training and inference. Supply constraints are mainly in powered shells, with structural supply issues across components like GPUs and transformers.

Q: Developments with sovereign governments and backlog from hyperscaler contracts?

A: Michael Intrator noted sovereigns are seeking best-in-class tech, with expansion in the sovereign cloud universe. Nitin Agrawal clarified one hyperscaler contract was in Q2 backlog, the other in Q3.

Q: Economics of inference vs. training and flexible capacity products?

A: Michael Intrator said economics for inference and training are identical in long-term contracts. Flexible capacity products like spot are in preview to attract new users and use cases.

Q: Repurposing older GPU clusters and CapEx timing?

A: Michael Intrator said older GPUs are recontracted for inference. CapEx is backloaded in Q4 due to infrastructure build-out timing.

Q: Cost timing and end segments of demand?

A: Nitin Agrawal discussed costs for infrastructure before revenue. Michael Intrator noted broad-based demand from labs, enterprises, and new sectors like VFX and finance.

Q: Cost of debt and Weights and Biases sales tracking?

A: Michael Intrator highlighted lower cost of debt through capital raises. Integration with Weights and Biases is driving new clients and product developments.

Q: On-demand/spots and Weights and Biases sales tracking?

A: Michael Intrator said on-demand/spots help new players, and Weights and Biases integration is driving product development and client traction.