CoreWeave Earnings Call: 40% Reduction in This Year's Capital Expenditure Guidance Mainly Attributed to Delivery Delays; Prices for Older-Generation GPUs Remain Firm

After the US stock market closed on Monday, $CoreWeave(CRWV.US)$ a mixed Q3 earnings report was released, showing strong revenue and backlog performance. The third-quarter revenue reached $1.4 billion, increasing by 134% year-on-year, with the order backlog surging to $55.6 billion, nearly double that of the second quarter.

However, due to delivery delays, short-term capital expenditure guidance has sharply declined. The company expects its 2025 capital expenditure to be between $12 billion and $14 billion, a steep 40% drop from the previously anticipated range of $20 billion to $23 billion. Nevertheless, long-term capital expenditure remains solid, with 2026 spending projected to exceed “more than twice that of 2025.”

During the subsequent conference call, CoreWeave’s management provided detailed explanations regarding the reasons for the reduction in capital expenditure as well as the issue of overcapacity. Below are the key highlights from the call:

• Contracted power capacity increased to 2.9 gigawatts; however, delivery delays of “power shells” by a third-party data center developer led the company to significantly lower its full-year 2025 capital expenditure guidance to a range of $12 billion to $14 billion, down approximately 40% from previous market expectations. Affected customers have agreed to adjusted delivery schedules while retaining the total contract value.

• Given the strong growth in backlogged orders and “insatiable” demand, the company anticipates that 2026 capital expenditure will be “well over twice that of 2025,” suggesting CapEx for 2026 may exceed $24 billion or even higher.

• The delays primarily impacted the delivery of power shells, rather than GPUs or electricity itself. This is a temporary, systemic supply chain issue rather than a demand-related problem. The company operates in a “supply-constrained” environment where customer demand for AI computing power far exceeds current production capacity.

• The customer base has become more diversified, with nine out of the top ten clients having signed multiple contracts. Currently, no single client accounts for more than approximately 35% of total revenue, down from about 50% in the previous quarter and significantly below roughly 85% at the beginning of the year. Moreover, as of the end of the third quarter, more than 60% of revenue was generated from investment-grade clients.

• Market demand for AI cloud technology remains robust across all GPU generations. For instance, in the third quarter, the first contract exceeding 10,000 H100s was nearing expiration. The client proactively renewed the infrastructure contract two quarters ahead of schedule at a price only 5% lower than the original contract terms.

• Financing costs are declining significantly (e.g., interest rates on DDTL 3.0 have dropped by 900 basis points compared to prior levels), thanks to the company's first-class execution capabilities and stable cash flow visibility driven by “take-or-pay” contracts. This lays the groundwork for continued low-cost financing in the future.

• The company emphasized that it provides 'full-stack services from hardware to software,' including its proprietary orchestration software, the newly launched AI object storage (claimed to reduce customer costs by more than 75%), and an enhanced software toolchain through acquisitions such as OpenPipe and Marimo.

Below is the full transcript of the conference call:

Operator:

Good afternoon, everyone, and welcome to CoreWeave's Third Quarter 2025 Earnings Conference Call. Joining me today to discuss the results are Chief Executive Officer Mike Intrator and Chief Financial Officer Nitin Agarwal. Before we begin, I would like to take this opportunity to remind you that our remarks today will include forward-looking statements. Actual results may differ materially from those projected in these forward-looking statements. Factors that could cause actual results to differ materially are outlined in today’s earnings press release and in our Quarterly Report on Form 10-Q, which we will file with the U.S. Securities and Exchange Commission. Any forward-looking statements made during this call are based on assumptions as of today, and we undertake no obligation to update these statements due to new information or future events.

During this conference call, we will present both GAAP and certain non-GAAP financial metrics. A reconciliation table between GAAP and non-GAAP metrics is included in today’s earnings press release. The earnings press release and accompanying investor presentation are available on our website at investors.coreweave.com. A replay of this conference call will also be available on our investor relations website.

Now, I will turn the call over to Mike.

Michael:

Good afternoon, everyone, and thank you for joining us. CoreWeave has once again delivered an outstanding quarterly performance, demonstrating continued acceleration in our business as artificial intelligence rapidly proliferates across industries globally. We remain in a highly supply-constrained environment, with demand for CoreWeave’s premier artificial intelligence cloud platform significantly outpacing current capacity.

The strong demand for CoreWeave clearly indicates that leading global enterprises trust CoreWeave and rely on its platform to support their most critical AI workloads. In the third quarter, we exceeded expectations, achieving revenue of $1.4 billion, representing a year-over-year increase of 134%. In just the third quarter alone, we added over $25 billion in backlog revenue, bringing the total backlog at the end of the third quarter to more than $55 billion—almost double that of the second quarter and nearly four times the amount since the beginning of the year. Additionally, CoreWeave’s remaining performance obligations have reached $50 billion, surpassing the growth rate of any cloud platform in history.

These results demonstrate deep confidence in CoreWeave, as customers recognize us as the indispensable cloud service provider in the AI domain. Despite ongoing industry-wide capacity constraints, we continue to expand aggressively. Our active power capacity increased by 120 megawatts quarter-over-quarter to approximately 590 megawatts, while contracted power capacity grew by over 600 megawatts to reach 2.9 gigawatts. This lays a solid foundation for our future growth, with over 1 gigawatt of contracted capacity available for sale to customers, most of which is expected to come online within the next 12 to 24 months.

In the third quarter, we signed large-scale computing contracts with numerous major clients, including Meta and OpenAI. Each contract represents a significant expansion of our existing partnerships and diversifies our client base beyond reliance on a single customer. Additionally, we deepened our relationship with a leading hyperscale data center operator, signing our sixth contract with this client to date. In fact, nine out of our top ten clients have entered into multiple agreements with us, the sole exception being a new client who joined us in Q3. CoreWeave empowers pioneers in artificial intelligence innovation by doubling their capabilities, accelerating their path to breakthrough advancements.

These are among the most advanced AI institutions globally, and once they experience the performance, flexibility, and reliability of CoreWeave Cloud, they consistently expand their cooperation with us. This is the highest recognition we could receive. Our exceptional growth underscores that the adoption of artificial intelligence is rapidly surpassing the cutting-edge AI labs and hyperscale data centers, achieving breakthroughs. The broader global demand, along with our recent significant successes, is driving the diversification of our revenue sources.

For instance, the number of clients generating over $100 million in revenue in the past 12 months has tripled year-on-year. AI-native enterprises and companies across various industries are adopting CoreWeave to transform operations and unlock new potentials for innovation, productivity, and growth. Poolside, a leader in foundational model development, chose CoreWeave to support its mission of building general artificial intelligence and deploying enterprise-grade intelligent agents. Meanwhile, Periodic Labs leverages CoreWeave to push the boundaries of scientific discovery and computational research. At the application layer, we have added AI-native clients like Jasper, who selected CoreWeave as their cloud partner to revolutionize the digital marketing landscape.

We are also witnessing remarkable growth momentum within our enterprise client base. CrowdStrike has chosen CoreWeave to advance the development of AI-powered cybersecurity agents, while Rakuten utilizes our platform to transform its vision-language models, achieving greater transparency, reproducibility, and speed in AI workloads. Furthermore, we have expanded collaborations with various enterprise clients, including a leading software design platform and a major U.S. telecommunications operator. Today, our services have extended into the public sector, a market with unique requirements for performance and security. We recently launched CoreWeave Federal, aimed at providing cloud services to U.S. government agencies and the defense industrial base.

Currently, NASA is utilizing our services at its Jet Propulsion Laboratory to advance scientific exploration. We are honored to strengthen America's AI infrastructure, enabling agencies to accelerate innovation, address critical missions, and safeguard national interests. These recent successes also highlight our readiness to serve enterprises. As our client base continues to grow across verticals and geographies, we are thrilled to welcome John Jones to our team as our first Chief Revenue Officer. John previously served as Global Head of Startups and Venture Capital at AWS.

John’s addition will significantly enhance our team’s capabilities, playing a key role in scaling global revenues and driving the next phase of growth. Moving forward, while discussing our expanding data center footprint, I would like to briefly mention our previously announced plan to acquire Core Scientific, which was terminated in October. While the transaction was strategically beneficial for both companies, the valuation demanded by its shareholders was not suitable for CoreWeave, especially considering that the outcome of this deal would not adversely impact our ability to achieve growth targets in the coming years. We will continue to collaborate closely with Core Scientific, jointly operating approximately 590 megawatts of leased data center capacity.

We are expanding our capacity with rigor to meet the growing global demand for CoreWeave cloud services. As mentioned earlier, this quarter we increased our contracted power capacity to 2.9 gigawatts, enhancing the resilience and flexibility of our overall offering through diversification in scale, geography, and developers. As of the end of Q3, no single data center provider accounted for more than 20% of our contracted power portfolio. Last quarter, we added eight new data centers across the U.S., further strengthening our domestic coverage. Additionally, we are expanding in Europe, including the construction of a large-scale data center in Scotland, developed in collaboration with the UK government.

As announced during the summer, we have initiated self-build projects to further expand our footprint and enhance operational control. Despite the robust demand for our platform, data center developers across the industry are facing unprecedented supply chain pressures. On our end, delays from third-party data center developers have temporarily impacted our progress. This will affect our expectations for Q4, which Nitin will elaborate on later.

Nevertheless, affected clients have agreed to adjust delivery schedules and extend contract expiration dates. Consequently, we have maintained the total value of the original contracts, allowing clients to retain their capacity throughout the initial agreement period—a testament to their confidence in our ability to deliver the highest-performing solutions in the market. We take immense pride in our technical achievements, and clients consistently report that CoreWeave is the best platform for running AI workloads. In Q3, we continued delivering initial large-scale deployments of GB200, while once again being the first to bring GB300 to market, further demonstrating our unmatched operational track record.

CoreWeave’s leadership in the industry is unparalleled. We are the only cloud service provider to submit MLperf inference results for GB300-class systems, setting a benchmark for real-world AI performance. Just last week, Semianalysis recognized our excellence for the second time, awarding CoreWeave the highest honor—the Platinum Cluster Max rating—outperforming over 200 suppliers, including hyperscale cloud providers and emerging cloud competitors. No other cloud service provider has ever received this distinction.

CoreWeave has successfully achieved this twice, further solidifying its leading position in the AI cloud domain. The demand for AI cloud technology remains robust across all GPU generations. For instance, in the third quarter, our first contract exceeding 10,000 H100 units is set to expire. The customer proactively renewed the infrastructure contract two quarters in advance at a price only 5% lower than the original contract.

This strongly indicates customer satisfaction, as well as the long-term utility and differentiated value of GPUs running on the CoreWeave platform. CoreWeave is the world’s first hyperscale AI cloud platform, encompassing compute, storage, networking, and software specifically designed for AI workloads. Our continuously expanding cloud portfolio is built on an increasingly sophisticated suite of software and services, enabling customers to build, train, and deploy new products faster. In addition to task control, we also possess proprietary orchestration solutions that are critical for autonomously operating our cutting-edge AI cloud.

We recently launched CoreWeave AI Object Storage, a fully managed storage service that eliminates any friction associated with moving data across regions, cloud platforms, and storage tiers, with no egress fees or transaction charges. CoreWeave’s AI Object Storage delivers the highest throughput for AI workloads while saving customers over 75% in costs. We have observed significant market interest in this service, attracting several new early adopters, including frontier AI labs such as Mistral. Our entire storage platform has also seen rapid customer adoption, with annual recurring revenue (ARR) surpassing $100 million in the third quarter.

Combined with our uniquely designed global network backbone for artificial intelligence, CoreWeave has become the central hub for customers and their critical AI workloads, ensuring consistently superior performance and seamless user experiences whether using CoreWeave Cloud or other service providers. We have further expanded our observability and security suites to ensure CoreWeave can optimally handle all customers’ critical workloads, regardless of use case or geography. Over the past few years, we have been committed to supporting pioneers in artificial intelligence with development and improvements. Today, we are expanding our role to facilitate the practical application of AI. From tools developers need to build AI, to solutions required for real-world AI adoption, we have utilized mergers and acquisitions as a key means to accelerate this process, including the recent announced acquisitions of OpenPipe, Marimo, and Monolith.

With OpenPipe, we have quickly integrated its solution into our broader fine-tuning product suite and launched the first publicly available serverless reinforcement learning tool. Through Marimo, we are expanding CoreWeave’s influence within the open-source community, starting with entry-level exploration and prototyping. Both OpenPipe and Marimo seamlessly complement the functionalities of Weights and Biases, and we are rapidly growing the developer community relying on CoreWeave’s comprehensive platform. With Monolith, we are extending these capabilities into the real world to unlock the monetization potential of AI, initially focusing on industrial use cases with established enterprise customer bases and mature workloads, including leading automakers like Nissan and Stellantis.

Through the rapid and successful launch of new products and services, we are expanding our target market and growing alongside our customers. We are fundamentally enhancing CoreWeave’s capabilities to establish footholds in new markets and expand operations, all aimed at further supporting the rapid evolution of artificial intelligence and helping AI developers and innovators bring products to market faster and more reliably, thereby improving return on investment. Our collaboration model is maturing, exemplified by our partnership with CrowdStrike, which will unlock and accelerate partner-driven growth. The launch of our new storage products and our collaboration with Vast Data further demonstrate our efforts to accelerate portfolio and partner market expansion, allowing us to compete in new markets where previous product offerings were limited or nonexistent. This fosters customer-driven platform adoption and product-led growth, creating favorable conditions for our business development.

Finally, I would like to emphasize what truly sets CoreWeave apart. We are an indispensable cloud platform in the field of artificial intelligence, backed by unparalleled technical and operational strength, as well as a rapidly growing customer base. We offer the industry’s most powerful infrastructure, fastest time-to-market, and most advanced functionalities. Leading global AI innovators choose CoreWeave because we enable them to act faster, scale smarter, and achieve results unattainable elsewhere.

Our momentum has never been stronger, and the opportunities ahead continue to expand. With exceptional products, an outstanding team, and unmatched execution, CoreWeave is well-positioned to enter its next phase of growth as a full-stack AI service provider and hyperscale solutions provider. The future is powered by CoreWeave, and we are just getting started. Next, let me hand it over to Nitin.

Nitin Agrawal:

Thank you, Mike, and good afternoon, everyone. Our remarkable third-quarter results once again underscore the strong market demand for CoreWeave and our exceptional achievements in building a mission-critical cloud platform for artificial intelligence. As Mike mentioned, we continue to operate in a highly supply-constrained environment, a situation expected to persist for the foreseeable future. We remain committed to delivering the best-performing solutions in the market and are continuously increasing investments across the entire technology stack. This has driven the growth and diversification of our customer base, from emerging enterprises and AI-native businesses to the expansion of existing clients, yielding significant results. Now, let us take a closer look at the third-quarter performance.

Revenue for the third quarter reached $1.4 billion, representing a year-over-year increase of 134%, primarily driven by robust customer demand and efficient execution. The order backlog at the end of the quarter stood at $55.6 billion, nearly doubling in just the third quarter. Demand for the Blackwell platform as well as our entire GPU portfolio remains strong. During the third quarter, we secured multiple orders for legacy GPU models, adding new customers and renewing existing contracts.

The broad-based demand for CoreWeave cloud services has enabled us to significantly reduce customer concentration. Today, no single customer accounts for more than approximately 35% of our total revenue, down from about 50% in the previous quarter and importantly, from around 85% at the beginning of the year. Additionally, by the end of the third quarter, over 60% of our revenue was derived from investment-grade customers. This reflects our successful efforts to diversify both our platform and customer base.

Operating expenses for the third quarter totaled $1.3 billion, including $144 million in share-based compensation expenses. We continue to ramp up investments in data centers and server infrastructure to address the growing order backlog, which resulted in increased cost of revenue and technology and infrastructure spending during the third quarter. Additionally, the rise in sales and marketing expenses was driven by investments in marketing activities and scaling our go-to-market organization to capitalize on the rapidly growing AI opportunities within enterprises and AI-native companies. The increase in general and administrative expenses was driven by higher professional service costs and staffing levels.

Adjusted operating income for the third quarter was $217 million, compared to $125 million in the third quarter of 2024. Adjusted operating margin for the third quarter was 16%. Adjusted operating income exceeded expectations, primarily due to revenue growth, reduced costs resulting from timely delivery schedules at third-party partner data centers, and enhanced fleet operational efficiency. Net loss for the third quarter was $110 million, compared to a net loss of $360 million in the third quarter of 2024.

Interest expense for the third quarter was $311 million, compared to $104 million in the third quarter of 2024, primarily due to increased debt to support infrastructure scale expansion, though this was partially offset by improved debt pricing as we further reduced capital costs. Adjusted net loss for the third quarter was $41 million, compared to near breakeven in the third quarter of 2024; adjusted EBITDA for the third quarter was $838 million, compared to $379 million in the third quarter of 2024, representing more than a twofold increase year-over-year. Adjusted EBITDA margin was 61%. Capital expenditures for the third quarter totaled $1.9 billion, below expectations due to delayed deliveries from third-party data center providers as mentioned by Mike.

Construction-in-progress projects grew significantly to $6.9 billion, an increase of $2.8 billion quarter-over-quarter, as a direct result of this situation. It is important to note that construction-in-progress refers to infrastructure not yet in service and is not included in capital expenditures until it becomes operational. Now, let's turn to our balance sheet and strong liquidity position. As of September 30, we held $3 billion in cash, cash equivalents, restricted cash, and marketable securities.

Rapid growth and scaled operations require strategic financing approaches. CoreWeave has emerged as a leading AI cloud platform and is at the forefront of innovative financing for the infrastructure required to support the most advanced workloads for enterprises and AI labs. We have made significant progress in strengthening our capital structure and reducing capital costs. In the third quarter, we amended our DDTL 2.0 financing facility, increasing its remaining available capacity by more than $400 million and creating a new $3 billion tranche with an interest rate of SOFR + 425 basis points, which is significantly lower than the initial cost of this financing arrangement.

As previously discussed, we completed the DDTL 3.0 financing in the third quarter, priced at SOFR + 400, representing a 900-basis-point reduction from the non-investment-grade portion of our prior financing arrangements. Looking ahead, as funding providers increasingly recognize our best-in-class execution and the stable cash flow visibility underpinning our take-or-pay customer contracts, we expect to continue securing financing at lower spreads. Additionally, in July, we issued $1.75 billion in senior notes at a cost 25 basis points lower than our initial issuance in May, further expanding our presence in the high-yield bond market. Year-to-date, CoreWeave has successfully executed $14 billion in debt and equity transactions to support our rapidly growing order backlog and efficient scaling, positioning us for long-term growth.

Aside from payments related to OEM financing and self-liquidating debt generated through committed contract payments, we have no other debt maturities until 2028. Regarding taxes, in the third quarter, we recognized a non-cash tax benefit primarily attributable to the impact of a substantial tax bill. While this tax bill had a one-time effect on the third quarter due to catch-up payments from the beginning of the year, we expect that changes in legislation will enable CoreWeave to realize cash tax savings in future periods. Concerning our outlook, as previously noted, delays in PowerShell deliveries from data center providers will impact our fourth-quarter results.

These delays are temporary, and as Mike indicated, affected customers have agreed to revised delivery schedules to ensure their capacity remains aligned with the original agreement terms and achieves the total value of the original agreements. Against this backdrop, we anticipate 2025 revenue to be between $5.05 billion and $5.15 billion. Additionally, we expect 2025 adjusted operating income to range between $690 million and $720 million, and we project active power generation to exceed 850 megawatts by year-end. In the fourth quarter, we will initiate some of the largest projects in the company’s history.

Due to the timing difference between the initial occurrence of data center costs and the commencement of revenue recognition, this will have a short-term impact on the adjusted operating profit margin. We anticipate interest expenses for 2025 to range between USD 1.21 billion and USD 1.25 billion, primarily driven by increased debt to support demand-driven capital expenditure growth, partially offset by declining capital costs. Regarding capital expenditures, we now project that our capital spending for 2025 will be between USD 12 billion and USD 14 billion. This reduction in spending compared to previous expectations is mainly reflected in corresponding increases in construction and progress, as infrastructure awaiting deployment is being built following the delivery of powered data center capacity.

Consequently, the majority of the remaining capital expenditures previously anticipated to be recognized in the fourth quarter will now be recorded in the first quarter. Moreover, given the significant growth in our backlog and the continued strong demand for our cloud services, we expect capital expenditures for 2026 to more than double those of 2025. These investments in infrastructure platforms will enhance our competitive advantage and support our sustained high growth. In summary, we achieved record-breaking third-quarter results, and we are more confident than ever in the long-term prospects of our business.

This year, we have made tremendous progress. Our revenue backlog has grown at an accelerated pace, now surpassing USD 55 billion. Additionally, we have diversified our customer base, strengthened and expanded our platform through strategic partnerships and acquisitions, secured new sources of funding, significantly reduced capital costs, and scaled up capacity and organizational size at an unprecedented rate. These advancements enable us to seize current opportunities while laying a solid foundation for growth in the coming years. With the widespread application of artificial intelligence across industries and use cases, along with prudent business decisions aimed at expanding our product portfolio and increasing market share, our target market continues to expand.

CoreWeave is growing at an astonishing rate, with rapid and efficient scale expansion that reinforces our leadership position as a key cloud platform for artificial intelligence. Thank you to all investors and analysts for your support and engagement. We look forward to updating you on our progress in the upcoming quarters. Next, we will move into the Q&A session.

Q&A Session

Q1 Mark Murphy: Every time we discuss developments in the artificial intelligence space, we hear about surging order volumes, and CoreWeave is no exception. However, bottlenecks related to power and labor are becoming increasingly severe. Could you elaborate on the situation with third-party suppliers? Specifically, is it a shortage of power or labor? Or are there other issues related to GPUs, memory, or storage? Additionally, have you communicated with other third-party suppliers to understand their scheduling status and whether they believe they can meet deadlines or deliver by early next year?

Michael Intrator: Let me address this question from several different perspectives. Alright. First, you're absolutely correct. This issue is indeed very frustrating for our clients.

It is deeply frustrating for us that there are numerous systemic challenges within the supply chain that underpin global infrastructure development, which are essential for the realization of artificial intelligence. Nevertheless, we have taken multiple measures to enhance our ability to address future challenges. We have invested significant time in diversifying our data center suppliers. Furthermore, we have established a substantial portion of our organization dedicated to facilitating infrastructure operations.

We have initiated our own self-build projects, including sites in Kenilworth and Lancaster, Pennsylvania. Thus, you can see that we are actively expanding our operations and doing everything possible to minimize losses or delays in infrastructure delivery, as these delays have placed immense pressure on the supply chain. When you have diversified channels for infrastructure delivery, the relative impact of each delay diminishes. You can flexibly call upon different data centers based on delivery timelines.

We view this as a critical infrastructure module, albeit with some delays in delivery. Ultimately, the end users—our clients utilizing this infrastructure—have extended contract terms to allow us to deliver the full contracted value despite the delays. This underscores the value clients derive from our infrastructure. Therefore, when engaging with CoreWeave and other companies across the industry, you will repeatedly hear similar narratives.

This is the real challenge at the power delivery enclosure level. It's not a challenge of electricity itself, right? Currently, there is sufficient electricity, and we believe there will be ample supply in the coming years. However, the true challenge lies within the power delivery infrastructure.

Q: So, Michael, is this completely unrelated to Core Scientific, or is it entirely separate from what you have experienced?

Michael Intrator: So I wouldn’t specifically comment on any of our data center providers.

We are working closely with all data center vendors, doing everything possible to ensure they deliver the necessary infrastructure to us in the end. We have made remarkable progress in infrastructure delivery. As you can see, our capacity continues to grow, now reaching approximately 590 megawatts. Since the last earnings call, our capacity has increased by another 120 megawatts.

As we continue to scale up deliveries, we are seeing significant success. But I think it doesn’t matter which specific data center provider it is. This is a systemic issue that the entire industry must face in the foreseeable future. The key point — or at least for me, the key point — is that delays in infrastructure construction will not affect our backlog, nor impact our ability to fully capitalize on contracts set to be delivered.

Q2, Keith Weiss: Thank you all for addressing this question, and congratulations on another impressive quarter in clearing your backlog.

You’re absolutely right. We’ve never seen any cloud service provider scale as rapidly as you have. Mike, my question, one we get asked a lot, relates to the risk of overcapacity, but I think it’s more specific than that — people are concerned about potential overcapacity in AI labs signing up for capacity.

However, my question to you is, how should we think about your infrastructure and the infrastructure you build in terms of its actual substitutability? When you build for specific customers, are those data centers applicable to any customer? Can they be used for inference and training, or are you truly building custom solutions tailored to specific clients, which might limit your flexibility if a client underperforms or behaves unpredictably, thus reducing your degrees of freedom?

Michael Intrator: Yes, Keith, that’s a great question. Actually, as we advance relationships with all our customers, we spend a lot of time thinking about this very issue. In short, the infrastructure is substitutable.

It can transfer from one customer to another. The infrastructure is built to the most demanding specifications, so it can be used for both training and inference. We’ve thought deeply about ensuring that our infrastructure maintains as much optionality and flexibility as possible during construction. What I want to emphasize is that much of this flexibility and substitutability is due to the exceptional software suite we provide, which allows the infrastructure to be utilized so effectively, right? Just like when SemiAnalysis conducts their annual alternative review, CoreWeave is repeatedly singled out as the best solution in the world for this type of infrastructure, and there’s a reason for that.

This includes hyperscale cloud providers, emerging cloud companies, and all other players attempting to deliver such infrastructure. We simply excel at it, and we believe that by offering such a powerful software suite to deliver infrastructure, we are preserving significant value.

Q3 Kash Rangan: The backlog growth has been quite remarkable. Let me make two brief points. The first is, Mike, I think you mentioned earlier that you would diversify contractors in terms of data center operations. Perhaps you could give us a concrete update on how far we are from reaching a point where any disruptions unrelated to your business do not affect your revenue outlook? How far are we from that point?

Secondly, when you look at industry developments, I mean, no one anticipated — perhaps some did — the $250 billion contract between OpenAI and Microsoft, nor the $300 billion contract between OpenAI and Oracle. Suddenly, of course, CoreWeave has a unique value proposition, enabling the rapid and highly efficient construction of GPU clusters almost at the speed of thought. However, given that we are discussing multi-hundred-million-dollar contracts awarded to hyperscale cloud giants, what will keep you distinct three or four years from now when supply and demand reach equilibrium? What will be the shining value proposition that keeps CoreWeave relevant when we look back?

Michael Intrator: Yes, thank you. Let me break this down into two parts, if that’s okay? Your first question was about diversification and when it starts and stops, thereby affecting our quarterly performance metrics? I want to emphasize that as individual data center projects become smaller relative to our overall portfolio of operational data centers, delays of a few weeks will have an increasingly smaller impact on overall results, right? For example, when you deliver 590 megawatts, and there is an incremental 200 or 300 megawatts, that portion represents a substantial share if delivered in the next quarter, correct? As we scale up and begin constructing our full capacity of 2.9 gigawatts, a delay of one or two weeks for a 100-megawatt data center will not materially affect outcomes. As Nitin mentioned, we expect the vast majority of this 2.9 gigawatts to come online within the next 12 to 24 months. This way, you can clearly see that as scale increases, the curve becomes smoother, with each data center having a relatively smaller impact. That’s part one regarding scalability.

The second part relates to a question that has been asked since our founding: Why is CoreWeave able to deliver GPUs faster? Why are we capable of delivering the GPUs NVIDIA uses to run its machine learning performance benchmarks? Why have we developed the defining software in this space? Each quarter, you’ll see us widening our lead because we tailor our cloud services to real-world use cases. As you’ve seen in previous analyses, this sets us apart.

We are actively expanding our product offerings. We are building or acquiring additional computing capacity to further reduce the commoditization of the computing services we provide. Thus, a company specifically established to deliver such computing services will have a competitive edge moving forward.

Q4: Tyler Radke: Regarding the delays you mentioned this quarter, could you elaborate further on their impact on 2026? I know, Nitin, you’ve provided some high-level analysis on capital expenditures. But based on the information you currently have, particularly regarding the 24-month portion of RPO (Recovery Point Objective), how should we view the implications of these changes on revenue? Do you anticipate that the delays will be fully resolved in the first quarter? Will we see higher growth rates next year compared to this year? Any insights would be helpful.

Michael Intrator: Let me start, and then I’ll hand it over to Nitin. I think it’s important to clarify that the capacity ramp we’re seeing now is related to infrastructure tied to a single vendor. Meanwhile, we are working with other vendors to execute parallel PAFs (Procurement, Authorization, and Financing) for other contracts. Therefore, there will be some short-term impacts on delivery.

Going forward, you will see that we have the ability to accelerate progress this year to compensate for prior delays. So, the majority of the delays you’ve observed should be resolved by the first quarter of next year. Nitin?

Nitin Agrawal: Yes, Tyler, you’re correct. Most of the capital expenditure delays we experienced in Q4 will be completed in Q1. As you would expect, we will gradually increase capacity throughout Q1. As Mike mentioned earlier, by adjusting delivery timelines with customers, we ensure they retain full capacity and the associated contract value, meaning total revenue related to that customer remains unaffected. In our next earnings report, we will share more details about our 2026 construction plans and related revenue projections. However, as we emphasized this quarter, given strong customer demand — reflected in the growth of our revenue backlog and sustained customer interest — we anticipate that 2026 capital expenditures will be more than double those of 2025.

Q5 Michael Turrin: I would like to connect some comments, as the growth in bookings has been quite significant, and there are many questions surrounding the booking sequence. It sounds like you're saying the supply chain impact you see is more specific to certain customers. What I want to better understand is whether this will affect your pace of signing new customers or if it's more about post-RAMP signings and a particular customer context? Additionally, a smaller question: Is the NVIDIA deal specifically reflected in the backlog metric? Given the slightly different contract structure with NVIDIA, I’d love to hear you elaborate on the implications of that deal. Thank you very much.

Michael Intrator: Certainly. This does not impact our ability to attract more customers. I think it’s important to understand that we are building infrastructure in parallel. There was an issue with one data center that affected us, but we have 32 data centers in our portfolio.

They are all progressing at varying stages. Thus, each project is being handled independently. As Nitin mentioned earlier, we have 2.9 gigawatts of contracted power coming online within the next 12 to 24 months. We will work to find customers who can utilize that power, which will significantly impact our future revenue.

This data center will complete its upgrade first, and then we will proceed from there. Would you like to talk about the NVIDIA deal?

Nitin Agrawal: Yes, Michael, regarding the NVIDIA deal, we are very excited about it. This contract allows for previously reserved capacity for NVIDIA to be interrupted and resold to other customers.

Therefore, the nature of this contract enables us to serve numerous smaller clients profitably, such as high-growth AI labs that prefer short-term, low prepayment terms, while eliminating any risk to us regarding capacity utilization. Given the flexibility in the contract to interrupt and resell capacity, we are very excited about it. Accounting standards require us to exclude from RPO the capacity expected to be resold to other customers. To clarify, if this capacity is not resold, it will remain with NVIDIA and will be recognized as revenue. Hence, you will see this NVIDIA contract in our revenue backlog, but largely it is not part of our RPO.

Michael Intrator: Let me add to that. As Nitin said, we are very excited because this contract will allow us to provide infrastructure to emerging companies, startups, and those that struggle to access the computational infrastructure needed to build their businesses.

Thus, this interoperability is an extremely powerful tool for enhancing system resilience and creating opportunities for new companies to join CoreWeave’s broader product portfolio. We are very excited about this. We believe it is an excellent architecture. This agreement with NVIDIA ensures they will bear the full economic cost since we will be selling them computing resources.

I would like to emphasize that this indeed represents an extremely disciplined approach to computational financing, enabling us to reach market segments that were previously inaccessible to us, or frankly, to anyone.

Brent Thill: I just want to confirm that your capital expenditure has been cut by 40% this year. I’m confirming that this is feedback from just one client, correct? You are not assuming widespread delays across other clients, right?

That's correct. It relates to a single data center provider partner and the associated delays. As we mentioned in our prepared remarks, the majority of issues, the vast majority, will be resolved in the first quarter. And by the fourth quarter, you will see material impacts on construction and progress.

Q7 Raimo Lenschow: When thinking about next year’s capital expenditures, could you also talk about funding sources? We’ve seen many other companies increasingly adopting leasing models. You’ve talked about capex being what you need to do. How do you view the path forward, and how can different funding mechanisms give the company more flexibility? Thank you.

Michael Intrator: So as you see, we have driven innovation both in technology and in financing, right? Our perspective is that we look at all possible ways to finance and expand, and then we choose the most cost-effective way to scale and serve our customers better. If leasing is the optimal solution, we’ll go for leasing. But we have evaluated many different financing structures.

Over the past three years, we’ve built multiple financing structures to secure ample funding. Looking ahead, we will continue to explore and fully leverage these structures. Before finalizing any funding source, we don’t sign contracts with customers lightly. We evaluate each transaction individually to ensure that all elements — data centers, power, GPUs, and financing — are in place, thereby guaranteeing successful delivery of computational services to our clients.

Q8 Amit Daryanani: I’d like you to talk about the impact on your capex and time to market for power products as you transition from relying on third-party data center providers to building more of your own data centers. We’d love to understand what you think the optimal mix looks like and what the capex implications might be when you’re leaning more towards building your own data centers versus relying on third-party providers.

Michael Intrator: Let me clarify. We’re not saying we are going to build our own data centers and stop using third-party data center providers. What we mean is that building our own data centers is one way to mitigate risks in delivering the entire project portfolio. Therefore, we will continue to work with partners who provide data center capacity, allowing us to use the facilities they build for us. All of this remains valid. We need that capacity to continue operating at our current pace and scale.

We view self-building as an integral part of the overall system. It allows us to get closer to the physical infrastructure and become more deeply integrated into the global supply chain, gaining firsthand insights. We believe that to reduce risks in a complex supply chain environment as efficiently as possible, it is essential to maintain both capabilities simultaneously.

Q9 Brad Zelnik: Currently, 2.9 gigawatts of power have been committed for delivery, with more than 1 gigawatt of power still uncontracted with customers. Meanwhile, we continue to see announcements of some other large deals across the industry. Given the strong demand, how do you see the progress of signing up the remaining capacity, and how would you frame this issue for us?

Michael Intrator: The completion of other deals strongly validates the supply and demand dynamics we’ve been describing for years, right? No single company has the ability to provide infrastructure globally to meet the growing demands of the world’s largest tech firms, the biggest AI labs, government agencies, enterprises, and so on. Thus, the closing of other deals reflects the enormous demand pressure faced by us — and hyperscale data centers, AI labs, and data centers in general. This further reinforces and validates the themes we have consistently emphasized. We believe that ultimately, our offering — a full-stack solution encompassing everything from hardware to software — represents the most valuable infrastructure that can be brought to market.

We still believe this will significantly increase demand for our infrastructure. Regarding the remaining capacity, we are carefully considering how to continue advancing the diversification of cloud services. We remain focused on various applications that will play crucial roles in shaping the future functioning of the world. We aim to allocate this infrastructure to relevant parties as soon as possible to ensure their successful product launches and business operations.

Nitin Agrawal: There are a few key points to note, as we mentioned in our prepared remarks. Currently, there is hardly any client whose orders account for more than 35% of our total revenue, a significant decrease from 85% at the beginning of the year. Additionally, 60% of our orders come from investment-grade clients. Therefore, we will be very cautious about these factors when managing our existing sellable capacity.

Q10 Brad Sills: I would like to ask a question regarding the concept of PowerShell being a bottleneck. Does CoreWeave utilize any intellectual property for data center construction? Are there any lessons learned from this delay that can be applied to other data center projects you are overseeing? I’m just trying to understand what you can do to address the bottlenecks encountered in resolving this specific contract. Thank you.

Michael Intrator: Yes, what I would say, Brad, is that I don't think our lessons are solely derived from this one delay. We have been operating in a globally systemically supply-constrained market for three years now. We understand how challenging it is. With each new wave of demand, the market becomes increasingly tight.

So, when you ask us what measures we will take in the future to position ourselves, what I really want to encourage you to think about is that we have built an entire organization within CoreWeave that helps us enhance our capacity through self-construction. This allows you to delve deep into the supply chain. You know where the power is and how it is distributed. You also understand what is required to build power shelves because you construct them yourself.

Beyond that, the partnerships you have with other third-party suppliers will enable us to achieve as much success as possible in an environment that will remain challenging for quite some time.

Thank you. That concludes today’s Q&A session.

Editor/jayden