CoreWeave Scales AI Infrastructure While Balancing Margin Expansion and Financing Costs

By the Numbers

• Revenue: $3.66 billion, up 168% year over year and $1.09 billion sequentially.
• Revenue backlog: $132.5 billion, up 138% year over year.
• Adjusted operating income: $379 million, with a 10% margin, up from 5% in Q2.
• Adjusted EBITDA: $2.26 billion, representing a 62% margin.
• Net loss: $709 million, including $846 million in net interest expense.
• Capital expenditures: $14.5 billion in Q3; full-year guidance raised to $37–39 billion.
• Power capacity: More than 4.6 GW contracted and 1.7 GW active, with more than 1.85 GW expected online by year-end.

Key Highlights

• CoreWeave's $132.5 billion backlog reinforces its position in specialized AI infrastructure, with deployment speed determining how quickly customer commitments generate revenue.
• Improving operating margins, stronger pricing, and GPU recontracting opportunities could improve the economics of CoreWeave's capital-intensive business.
• CoreWeave's software and managed services are gaining traction with enterprise customers, extending its role in training, inference, and production AI.

The News

CoreWeave reported third-quarter 2026 financial results on November 10, with revenue increasing 168% to $3.66 billion and revenue backlog reaching $132.5 billion. The company raised full-year revenue guidance to $13.0–13.3 billion and adjusted operating income guidance to $1.275–1.375 billion. Capital expenditures reached $14.5 billion as CoreWeave expanded infrastructure capacity to support additional customer commitments. For more information see the company’s official earnings announcement.

Analyst Take

CoreWeave continues to expand rapidly as demand for AI computing outpaces available infrastructure. Revenue more than doubled year over year, backlog reached $132.5 billion, and the company added more than 200 MW of active power during the quarter. Growth is translating into improved adjusted operating margins, although the cost of financing its infrastructure remains substantial.

Our earlier research examined CoreWeave's debt-intensive expansion and its evolution from specialized GPU infrastructure into an integrated AI cloud platform. Q3 adds evidence of improving contract economics, opportunities to extend GPU revenue lifecycles, and early commercial traction for its software and managed services.

We see that CoreWeave's aggressive operational inflection is reshaping how capital is raised and recycled across the AI sector, as demonstrated by its new $2.6 billion DDTL 5.5 facility specifically tailored to underwrite higher-yielding, shorter-duration compute contracts. By coupling this structured debt evolution with an upsized $4.2 billion low-coupon convertible note and an inaugural 35-million-share at-the-market equity program, the company is actively insulating itself from volatile, double-digit secondary junk bond markets while maintaining the capital velocity required to support $37B-$39B in annual CapEx.

Moreover, the operational debut of multi-rack NVIDIA Vera Rubin NVL72 architectures, delivering 4.8x higher token throughput in production, proves that CoreWeave's rapid cadence of next-generation hardware deployment and sovereign grid acquisitions (surpassing 1.3 GW contracted abroad) creates physical and performance moats that general-purpose hyperscalers and other rivals cannot easily emulate.

Backlog and Power Define the Deployment Opportunity

CoreWeave reported qualified customer demand through 2027 at ten times its available capacity, with demand for 2028 already several times greater than supply. These estimates extend beyond signed contracts but indicate continued pressure on AI infrastructure availability.

More than half of CoreWeave's $132.5 billion backlog is associated with contracts expected to begin service delivery by the end of 2026. Contracted power reached 4.6 GW, compared with 1.7 GW active, leaving a substantial pipeline of facilities and equipment still to deploy.

Securing power does not immediately translate into revenue. Facilities must be built, GPUs installed, and networking and cooling systems commissioned before customers can begin using the capacity. Active power also does not necessarily indicate fully utilized infrastructure. Deployment schedules and power availability will continue to influence revenue recognition and CoreWeave's ability to accept additional customer commitments.

Contract Pricing and GPU Reuse Improve the Economics

CoreWeave's adjusted operating margin reached 10%, up from 5% in Q2 and 1% in Q1, although below the 16% recorded a year earlier. New four- and five-year contracts are approaching $20–25 million in annual revenue per megawatt, while shorter-duration capacity can command approximately $40 million per megawatt annually.

The economics of existing GPU infrastructure are also changing. More than 130,000 GPUs will come off contract in 2027, followed by approximately 170,000 in 2028. Recontracting the 2027 capacity at current prices could generate nearly $4 billion in incremental annualized revenue without purchasing replacement GPUs.

Demand for earlier-generation accelerators may extend their revenue-generating lifecycles, particularly for inference and workloads that do not require the newest hardware. The ability to balance multiyear customer commitments with shorter-duration contracts and managed services could improve utilization and returns on existing infrastructure.

Financing Costs Remain Substantial

CoreWeave spent $14.5 billion on capital expenditures in Q3, while net interest expense reached $846 million, exceeding its $207 million in GAAP operating income.

The company expects approximately $9 billion in customer prepayments and deposits over the next three months, including more than $3 billion associated with existing infrastructure being recontracted. Its weighted average cost of debt declined to 7.9%, down 250 basis points year over year.

Customer-backed financing and asset-backed lending can reduce funding pressure, but CoreWeave remains sensitive to deployment delays, financing costs, and customer concentration. Improving operating margins must eventually produce sufficient cash flow to support infrastructure expansion and debt obligations.

CoreWeave Omni introduces an additional approach to capital efficiency, allowing customers to operate workloads on the company's platform while retaining ownership of their hardware. Its first customer agreement provides an initial commercial reference, although the financial contribution has not been disclosed.

The AI Platform Strategy Gains Enterprise Traction

CoreWeave's transition from GPU infrastructure provider to integrated AI cloud platform accelerated with its acquisition of Weights & Biases in May 2025 and the September 2026 launch of Forge. The platform combines model development, training, inference, evaluation, and application improvement.

Seventy percent of new cloud customers added during Q3 were already Forge customers, indicating that software adoption is helping establish infrastructure relationships. Managed inference has exceeded $250 million in booked annual recurring revenue, providing an early measure of demand for services beyond dedicated GPU capacity.

CoreWeave now serves 10% of Fortune 500 companies, while financial services account for approximately 10% of backlog. More than half of new customers added during 2026 came from industries outside AI labs and technology.

The integration of software, data services, networking, and accelerated computing gives CoreWeave more ways to support production AI workloads. The opportunity is to deepen customer relationships and improve infrastructure utilization, although the contribution of its software platform to consolidated revenue and margins remains difficult to isolate.

What Was Announced

CoreWeave reported Q3 revenue of $3.66 billion, up 168% year over year, and revenue backlog of $132.5 billion. Adjusted operating income reached $379 million, representing a 10% margin, while adjusted EBITDA totaled $2.26 billion. The company raised full-year revenue guidance to $13.0–13.3 billion, projected $37–39 billion in capital expenditures, and increased its year-end annualized revenue run-rate forecast to $20–21 billion.

Management reported more than 4.6 GW of contracted power and 1.7 GW active, with additional capacity expected online by year-end. Management also highlighted stronger contract pricing, GPU recontracting opportunities, approximately $9 billion in anticipated customer prepayments and deposits, and continued enterprise adoption of Forge and managed inference services.

Looking Ahead

CoreWeave's $132.5 billion backlog and expanding customer relationships reinforce its position among specialized AI cloud providers. Its ability to serve AI developers, hyperscalers, and enterprises will depend on converting contracted demand into deployed infrastructure and recognized revenue. Customer diversification will also influence how effectively CoreWeave competes with established cloud platforms as enterprise AI adoption expands.

Power availability and deployment capacity remain central to that growth. With more than 4.6 GW contracted and 1.7 GW active, CoreWeave has considerable infrastructure still to bring online. The timing and efficiency of those deployments will affect revenue growth, customer commitments, and capital requirements. Stronger contract pricing, improving operating margins, and opportunities to recontract existing GPUs could improve returns, but financing costs and hardware lifecycles remain important constraints on profitability.

CoreWeave's investments in Forge, managed inference, and integrated data services also extend its role in production AI. Software adoption and new enterprise relationships provide opportunities to improve utilization and generate recurring service revenue. The combination of infrastructure scale, deployment execution, and integrated software capabilities will determine whether CoreWeave can sustain its growth and establish a profitable position in the AI cloud market.