CoreWeave CEO Describes AI Compute Demand as "Relentless," Holds $66.8 Billion Order Backlog, Sees Long-Term Profit Margins Stabilizing at 25%

Deep News09:12

AI infrastructure provider CoreWeave saw its stock decline after issuing a weaker-than-expected first-quarter forecast, falling over 9% in after-hours trading, though its full-year outlook remains supportive.

After the U.S. market closed on February 26th, CoreWeave reported a fourth-quarter adjusted loss per share of 56 cents, wider than the consensus estimate of 50 cents. The net loss expanded significantly to $452 million from $51 million in the same period last year.

Although the first-quarter revenue guidance range of $1.9 to $2.0 billion fell far short of analyst expectations of $2.29 billion, dampening market sentiment.

CFO Nitin Agrawal explained that as the company undergoes massive expansion, data center leasing costs, power expenditures, and depreciation expenses commence ahead of revenue recognition. Agrawal later added that once the business and growth normalize, the company is confident in achieving long-term profit margins of 25% to 30%.

CEO Mike Intrator described AI compute demand as "relentless and never-ending," driving the average customer contract length to 5 years. Furthermore, an explosion in inference demand has caused prices for A100 chips to increase rather than fall.

The company's capital expenditures are projected to double to over $30 billion by 2026, with expected revenue of $12-13 billion, targeting annualized revenue exceeding $30 billion by 2027.

For the 2025 fiscal year, CoreWeave's full-year revenue reached $5.1 billion, a surge of 168% year-over-year. CEO Mike Intrator stated unequivocally during the call that 2025 was a decisive year for CoreWeave, marking the fastest cloud platform in history to reach $5 billion in annual revenue.

Intrator, looking ahead, stated that the company's 2026 capacity is largely sold out, and it continues to sign new contracts, which will be allocated upon the activation of new capacity in 2027.

Following the Thursday earnings report, CoreWeave's stock dropped nearly 9% in after-hours trading. Despite this, CoreWeave's year-to-date gain remains over 36%, significantly outpacing the S&P 500's approximate 1% gain.

Key takeaways from the earnings call include:

* **Short-term margin pressure, long-term optimism:** Impacted by upfront infrastructure deployment and depreciation, Q1 2026 is expected to be the low point for margins, with adjusted operating profit anticipated at only $0 to $40 million, before climbing quarter-by-quarter. Management reaffirmed a long-term operating margin target of 25%-30% as the business matures. * **Staggering order backlog:** The signed contract backlog reached $66.8 billion by year-end, increasing by $11.2 billion quarter-over-quarter and over $50 billion year-over-year. * **Diverse client base and extended contract terms:** Demand has spread from hyperscalers to AI-native companies and traditional enterprises. The average customer contract length extended from about 4 years to about 5 years. * **Strong demand for previous-generation GPUs:** Demand for previous-generation GPU architectures surged, primarily for inference. During Q4, the average price of the H100 remained above 90% of its年初 level, while the average price of the A100 actually increased during 2025. * **Significant product cross-selling:** Customers are no longer just leasing GPUs; 80% of clients with annual consumption over $1 million adopted CoreWeave's storage products. Bundled sales with software like Weights & Biases contributed hundreds of millions in contract value in the second half. * **Doubling power and data center scale:** By the end of 2025, the company operated 43 active data centers with an active power capacity of 850 megawatts, adding 260 MW in Q4 alone. Active power capacity is expected to double to 1.7 gigawatts by the end of 2026. * **Becoming an Nvidia "Exemplar Cloud":** In Q4, CoreWeave became the first platform to receive Nvidia GB200 "Exemplar Cloud" status. It expects to be among the first to bring Nvidia's next-generation Rubin GPU to market in the second half of 2026. * **Aggressive 2026 guidance:** Revenue for 2026 is projected to be $12-13 billion, representing approximately 140% growth at the midpoint. Annualized revenue is expected to reach $17-19 billion by the end of 2026 and exceed $30 billion by the end of 2027.

While margins face short-term pressure, future potential exists. CoreWeave's core business involves deploying Nvidia GPUs in its data centers and renting compute power to enterprise clients and AI model developers for training and running large language models.

The company's Q4 adjusted operating profit margin was 6%. Addressing short-term margin pressure from massive capital expenditures, CEO Intrator stated that their margins reflect the cost of building tomorrow's revenues.

CFO Nitin Agrawal elaborated that as the company scales massively, data center leasing costs, power costs, and depreciation start before associated revenue is recognized. This is particularly pronounced in 2026, as the company deploys roughly double the new capacity compared to 2025. He emphasized that Q1 represents the low point for annual margins. Agrawal reiterated confidence in achieving 25%-30% long-term margins once business normalizes.

Regarding future growth potential, CoreWeave is evolving beyond a pure "GPU renter" towards higher-margin software and ecosystems. Agrawal disclosed that approximately 80% of CoreWeave cloud clients with annual spending over $1 million have adopted at least one of the company's storage products. The storage business achieved an annual recurring revenue exceeding $100 million in Q3. Additionally, cross-selling with Weights & Biases added hundreds of millions in total contract value in the second half.

On the software front, CoreWeave has developed proprietary cloud technology stacks and reference architectures, including Sunk and Mission Control. In January, Nvidia announced its intent to test and validate CoreWeave's platform for inclusion in its reference architectures for other cloud, enterprise, and sovereign clients. The CEO summarized that broadly distributing their proprietary cloud stack is expected to become a growing source of high-margin revenue over time, representing tangible long-term upside not reflected in the current 2026 guidance.

The "relentless" demand is matched by a "massive" order backlog. CoreWeave reported that as of December 31, 2025, its contracted revenue backlog surged to $66.8 billion, up $11.2 billion quarterly and over $50 billion year-over-year. CEO Mike Intrator stated the demand environment remains relentless, driving rapid adoption from hyperscalers, AI-native firms, to traditional enterprise customers.

A notable change is in customer behavior. CFO Nitin Agrawal pointed out that customers are entrusting CoreWeave with their foundational AI workloads for longer periods, with the weighted average contract length increasing from about 4 years to about 5 years. Furthermore, customer diversification continued throughout 2025, with the number of clients committing to spend at least $1 million on the CoreWeave cloud platform growing nearly 150%. New clients included AI-native and enterprise names like Cognition, Cursor, Mercado Libre, MidJourney, and Runway, while relationships with two existing hyperscale cloud clients were expanded.

Addressing prior market concerns about "older chip oversupply," management revealed that demand for previous-generation GPU architectures is increasing significantly due to the rapid proliferation of AI inference use cases. Intrator noted that related contracts are being signed ahead of capacity availability, aiming to dispel concerns about weak demand for older GPUs. On pricing, Intrator stated that GPU pricing remained generally stable throughout 2025. H100 average pricing in Q4 was down less than 10% compared to the beginning of the year, while A100 pricing trended upwards.

On the execution front, CoreWeave cleared concerns about data center delays discussed last quarter. The CEO revealed the company has delivered over 50,000 Grace Blackwell units to affected clients and became the first cloud platform to achieve Nvidia GB200 exemplar cloud status.

To address the massive backlog, CoreWeave expects 2026 capital expenditures to be $30-35 billion, more than double the approximately $15 billion spent in 2025, with almost all of it directly linked to signed customer contracts. The CFO emphasized that nearly all 2026 capital expenditures are directly tied to already-signed customer contracts, expecting to double active power capacity to over 1.7 GW by year-end.

For 2026 performance guidance, the company expects full-year revenue of $12-13 billion, representing approximately 140% growth at the midpoint. It also anticipates annualized run-rate revenue reaching $17-19 billion by the end of 2026, growing to over $30 billion by the end of 2027.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap
Leave a comment