CoreWeave (Nasdaq: CRWV) Q4 2025 earnings: AI cloud revenue hits $5bn as 2026 capital spend targets $35bn

CoreWeave, Inc. (Nasdaq: CRWV), the purpose-built AI cloud provider that went public in early 2025, reported full-year revenue of $5.1 billion for fiscal year 2025, a 168 percent increase from $1.9 billion in the prior year, making it the fastest cloud company in history to reach that revenue milestone. Fourth-quarter revenue reached $1.6 billion, up 110 percent year over year, accompanied by a revenue backlog that grew 342 percent to $66.8 billion. The results confirm that CoreWeave has moved beyond early-stage positioning to become a structurally significant player in enterprise AI infrastructure, though the scale of its capital deployment and interest burden now raise questions that growth alone cannot answer.

How did CoreWeave achieve $5 billion in annual AI cloud revenue in its first year as a public company?

The answer is relatively straightforward and simultaneously alarming in its implications for the rest of the cloud industry: CoreWeave built infrastructure faster than anyone else at the moment the largest buyers of compute in history needed it. With active power capacity exceeding 850 megawatts across 43 data centers by year-end, and contracted power commitments extending to more than 3.1 gigawatts, the company has assembled what may be the largest purpose-built AI compute footprint outside of the hyperscalers themselves.

Customers ranging from AI labs and frontier model developers to hyperscaler cloud providers and enterprise adopters have signed long-term committed contracts. The backlog composition is notable: 42 percent of the $66.8 billion is expected to convert within 24 months, 37 percent in the 25-to-48-month window, and 21 percent beyond 48 months. This is not speculative pipeline. These are contracted commitments, and the near-term portion alone represents more than the company’s entire fiscal year 2025 revenue.

New customer wins in the fourth quarter included Cognition, CrowdStrike, Cursor, Mercado Libre, Midjourney, and Runway, alongside expanded arrangements with existing hyperscaler cloud customers. The breadth of that list signals that CoreWeave is no longer a niche provider to frontier AI labs. Enterprise adoption is accelerating, and the company’s decision to invest in vertical integration across the technology stack — from bare-metal compute and high-speed interconnects to orchestration software, storage, and agent development tooling — is beginning to pay off in customer retention and deal expansion.

What does CoreWeave’s $66.8 billion revenue backlog reveal about the structural shift in AI infrastructure procurement?

The backlog figure deserves more than a passing glance. At $66.8 billion against $5.1 billion in realized revenue, CoreWeave is sitting on roughly 13 times its annual run rate in committed future business. That ratio is not typical for technology infrastructure companies and reflects something more significant than a good sales quarter. It reflects a fundamental shift in how the world’s largest AI developers are thinking about compute access.

Hyperscalers and AI labs are moving away from purely spot-market or short-cycle procurement toward multi-year committed infrastructure arrangements. The reasoning is straightforward: training and inference workloads at scale require predictable, low-latency access to specialized compute that general-purpose cloud architectures were not designed to deliver efficiently. CoreWeave’s purpose-built approach, validated by benchmark performance certifications from external industry analysts, gives buyers confidence that the infrastructure will perform as specified under demanding conditions.

The risk embedded in a backlog of this magnitude is equally worth acknowledging. Backlog figures of this type include both firm remaining performance obligations ($60.7 billion) and additional estimated future revenue under committed contracts ($6.1 billion). The company notes that delivery is subject to availability and service requirements. Should data center build-outs experience delays, GPU supply constraints emerge, or customer financial positions deteriorate, backlog figures can and do diverge from realized revenue. The 342 percent growth in backlog also reflects the practice of signing very large deals, which means concentration risk is inherent even if the customer list is diversifying.

Why is CoreWeave’s adjusted operating income margin declining even as revenue scales at 168 percent annually?

This is the question institutional investors will be most focused on, and the answer requires separating structural reality from temporary investment dynamics. Adjusted operating income for Q4 2025 was $88 million, representing a 6 percent margin — down from $121 million and a 16 percent margin in Q4 2024. For the full year, adjusted operating income was $666 million on a 13 percent margin, compared to $356 million at a 19 percent margin in 2024.

The compression is almost entirely attributable to the explosive growth in stock-based compensation following the IPO. Stock-based compensation totaled $630 million for fiscal year 2025, compared to $32 million in 2024. That single line item accounts for the bulk of the margin divergence when moving from GAAP operating loss of $46 million to adjusted operating income of $666 million. The adjusted EBITDA figure tells a more operationally instructive story: $3.1 billion for the full year at a 60 percent margin, declining modestly from 64 percent in 2024 as depreciation scales with the asset base.

The deeper issue is that CoreWeave’s capital intensity is increasing faster than its margins can absorb. Capital expenditures for fiscal year 2025 totaled approximately $14.9 billion, with $8.2 billion deployed in Q4 alone. For fiscal year 2026, the company is guiding for $30 to $35 billion in capital expenditures. That is a staggering commitment for a company with $5.1 billion in trailing revenue, and it implies continued reliance on external financing at scale.

How is CoreWeave financing $30 to $35 billion in capital expenditures planned for fiscal year 2026?

CoreWeave has developed what management describes as a systematic approach to financing at scale, and the capital structure assembled over the past year reflects that methodology. The company closed a $2.6 billion convertible senior note offering in Q4 2025 and expanded its revolving credit facility to $2.5 billion, providing near-term liquidity buffers. However, net interest expense reached $1.2 billion for fiscal year 2025 and is expected to run at $510 to $590 million in Q1 2026 alone, implying a full-year 2026 interest burden potentially exceeding $2 billion.

The asset-backed nature of CoreWeave’s infrastructure — long-lived GPU clusters under multi-year customer contracts — provides a financing template that is more akin to project finance or sale-leaseback structures than traditional technology company debt. The contracted backlog effectively serves as collateral, and lenders can model cash flow with reasonable precision. The risk is that any meaningful disruption to backlog realization, or a shift in the financing environment, would compress CoreWeave’s financial flexibility at precisely the moment it needs to deploy capital most aggressively.

The conversion path to profitability runs through scale efficiencies that have not yet materialized at current levels of capital deployment. Management targets a long-term adjusted operating income margin of 25 to 30 percent, a figure that implies substantial operating leverage as the asset base matures and depreciation as a percentage of revenue stabilizes. Whether that path is achievable depends heavily on whether AI workload demand remains as concentrated and long-term in nature as current backlog suggests.

What does CoreWeave’s fiscal year 2026 guidance of $12 to $13 billion in revenue imply for competitive dynamics in AI cloud infrastructure?

Revenue guidance of $12 to $13 billion for fiscal year 2026 represents a projected growth rate of roughly 140 to 155 percent on a full-year basis. The company targets an annualized exit run rate of $17 to $19 billion by year-end 2026 and greater than $30 billion exiting 2027. These are not projections built on analyst modeling assumptions. They are informed by a committed backlog that already covers a substantial portion of the guided range.

For competitors, the numbers are clarifying. Amazon Web Services, Microsoft Azure, and Google Cloud have each announced aggressive AI infrastructure investment programs measured in the tens of billions annually, but their general-purpose architectures and multi-tenant customer bases mean they are not direct substitutes for CoreWeave’s specialized positioning. The more interesting competitive dynamic is between CoreWeave and emerging specialized AI cloud providers, none of which have achieved anything close to comparable scale or backlog depth.

The hyperscaler relationship is also more collaborative than competitive at present. The fact that existing hyperscaler cloud customers are expanding their arrangements with CoreWeave suggests that general-purpose cloud providers are themselves purchasing specialized AI compute capacity rather than building it all internally. This dynamic, if it persists, would validate CoreWeave’s positioning as a wholesale infrastructure layer rather than a direct end-customer competitor to the major cloud platforms.

What are the key execution risks CoreWeave faces as it scales toward $12 to $13 billion in revenue in 2026?

Three execution risks warrant close attention. First, the power and physical infrastructure buildout required to support $30 to $35 billion in annual capital expenditure is operationally demanding at a level most technology companies have never attempted. CoreWeave added 11 data centers in 2025 and expanded active power capacity by 260 megawatts in Q4 alone. Maintaining that deployment velocity while managing quality, security, and operational reliability across a rapidly expanding global footprint introduces meaningful execution risk.

Second, the GPU supply chain remains a potential constraint. CoreWeave’s infrastructure is built around NVIDIA compute, and the company’s status as an NVIDIA Exemplar Cloud for GB200 NVL72 training workloads reflects a deep and strategically important relationship. However, NVIDIA’s own production capacity and allocation decisions will directly affect CoreWeave’s ability to convert capital commitments into deployable infrastructure on the timelines customers require. Any supply-side friction at NVIDIA translates almost immediately into CoreWeave’s delivery timeline risk.

Third, customer concentration is a structural feature rather than a temporary condition. The nature of AI infrastructure contracts means that a small number of very large customers generate a disproportionate share of revenue. If any of CoreWeave’s largest relationships change materially, whether through model architecture shifts that reduce compute intensity, internal infrastructure buildouts at major AI labs, or customer financial distress, the impact on revenue could be outsized. The diversification signaled by new enterprise customer wins is encouraging, but the enterprise segment is at an early stage of CoreWeave’s overall mix.

The acquisitions of Monolith and Marimo during Q4 2025, extending CoreWeave’s platform into industrial AI applications and Python-native AI development workflows respectively, signal an intent to deepen platform stickiness and broaden the addressable market beyond pure infrastructure rental. Whether these moves translate into meaningful revenue diversification within the planning horizon is an open question.

Key takeaways: what CoreWeave’s Q4 2025 earnings mean for the company, its competitors, and the AI infrastructure industry

• CoreWeave’s $66.8 billion revenue backlog, up 342 percent year over year, is the most significant data point in the release. It implies structural demand shift toward long-term committed AI compute procurement rather than spot-market purchasing.
• Full-year capital expenditure guidance of $30 to $35 billion for 2026 makes CoreWeave one of the largest infrastructure investors globally, a position that requires the financing architecture and execution discipline of a capital-intensive industrial company, not a software business.
• Adjusted operating margin compression from 19 percent to 13 percent year over year is primarily an IPO-related stock compensation artifact, not a deterioration in core unit economics. The 60 percent adjusted EBITDA margin is structurally robust for an infrastructure business at this scale.
• Interest expense exceeding $1.2 billion in fiscal year 2025, with Q1 2026 guidance of $510 to $590 million, means debt servicing alone will consume a substantial portion of operating cash generation through the next growth cycle.
• The hyperscaler expansion signal — existing hyperscaler cloud customers deepening their CoreWeave relationships — suggests that even the major cloud providers are purchasing specialized AI compute rather than building all capacity internally, validating CoreWeave’s wholesale infrastructure positioning.
• GPU supply chain dependency on NVIDIA is both a strength and a concentration risk. CoreWeave’s Exemplar Cloud designation provides strategic priority, but any production or allocation disruption at NVIDIA directly affects CoreWeave’s ability to deliver on contracted commitments.
• Revenue guidance of $12 to $13 billion for fiscal year 2026 is anchored in contracted backlog visibility, which is materially different from demand-dependent forecasts issued by most technology companies. The guidance is credible but not risk-free.
• Enterprise diversification is progressing but not yet a financial offset to concentration risk. Customers such as CrowdStrike, Mercado Libre, and Cursor represent the beginning of a broader enterprise base, but the revenue mix remains weighted toward AI labs and hyperscalers.
• The long-term adjusted operating margin target of 25 to 30 percent implies substantial operating leverage that has not yet materialized. Investors pricing for that outcome should monitor depreciation trends, interest expense trajectory, and the pace at which the contracted backlog converts to recognized revenue.
• CoreWeave’s emergence as a standalone AI infrastructure company at this scale is a structural challenge for existing cloud providers, who must now explain to investors why they did not build equivalent capacity sooner, or partner with companies that did.