CoreWeave Posts $452 Million Loss as AI Infrastructure Costs Surge
Nvidia-backed cloud operator's fourth-quarter loss nearly ninefold larger than prior year, exposing capital intensity pressures as sector grapples with profitability questions.
CoreWeave Inc. reported a fourth-quarter net loss of $452 million, nearly nine times the $51 million loss posted a year earlier, as the Nvidia-backed AI data center operator confronts the mounting financial strain of rapid infrastructure expansion.
The loss, disclosed February 26, arrived alongside an adjusted loss of $0.56 per share—wider than the $0.50 consensus estimate—according to Benzinga. Revenue climbed 110% year-over-year to $1.57 billion, marginally beating the $1.55 billion Street forecast. But shares tumbled more than 8% in after-hours trading as investors weighed light first-quarter revenue guidance of $1.9 billion to $2 billion—well below the $2.29 billion consensus—against capital expenditure projections of $30 billion to $35 billion for 2026, more than double the $14.9 billion deployed in 2025.
Interest Costs Devour Margins
The widening loss stems primarily from surging interest expenses, which climbed to $388 million in the quarter from $149 million a year prior, according to Futunn. CoreWeave carries nearly $19 billion in debt accumulated to finance GPU purchases and data center leases, and its debt burden stood at around $14 billion as of September, reported Yahoo Finance. The company reported adjusted EBITDA of $898 million with a 57% margin, but adjusted operating income fell to $88 million from $121 million a year earlier—lower than expected due to infrastructure deployed ahead of schedule.
CEO Mike Intrator told CNBC that the company made the decision to build faster to meet client demand. Nvidia graphics chips remain in short supply, he said, and CoreWeave brought 260 megawatts online in the fourth quarter alone, orchestrating deployment across more than 100,000 GPUs. The company finished 2025 with 43 active Data Centers and more than 850 megawatts of active power capacity.
Revenue Backlog Swells to $66.8 Billion
Despite the losses, CoreWeave’s contracted revenue backlog surged to $66.8 billion at year-end from $55.6 billion in the third quarter, driven by multi-billion-dollar contracts with Meta, OpenAI, and other AI labs, according to Markets Daily. The average contract length extended to five years from four years at the end of 2024. Intrator characterized demand as relentless, adding that the company signed approximately twice as many reserved instance customers in the fourth quarter as in any prior quarter.
The backlog provides revenue visibility but depends on CoreWeave meeting delivery obligations and data centers coming online on time. The company guided full-year 2026 revenue to $12 billion to $13 billion, representing roughly 140% growth at the midpoint. CFO Nitin Agrawal said substantially all of the projected $30 billion to $35 billion in 2026 capital expenditure is tied to already-signed customer contracts.
CoreWeave went public on Nasdaq in March 2025 at $40 per share, below the expected $47 to $55 range, after cutting its IPO size from $2.7 billion to $1.5 billion. At the time, 77% of revenue came from just two customers—62% from Microsoft alone—raising questions about concentration risk. The stock has since gained 36% year-to-date through February 26, according to SiliconANGLE.
Sector Faces Profitability Reckoning
CoreWeave’s results underscore broader pressures facing specialized AI Infrastructure providers, which are spending heavily to secure GPU supply and power capacity while competing against hyperscalers with far deeper balance sheets. Tech giants including Microsoft, Amazon, and Google are expected to spend more than $600 billion this year on AI infrastructure, according to Yahoo Finance. These firms can self-fund expansion with free cash flow, while CoreWeave relies on debt and equity markets.
Analysts remain divided. D.A. Davidson analyst Alexander Platt told Reuters that CoreWeave is being punished for either having too little or too much capex, and the fact the company is successfully bringing capacity online is a positive signal. But concerns persist about whether revenue can grow fast enough to outpace debt obligations. CoreWeave is also working to lower its weighted average cost of capital, with management noting the company secured more than $18 billion in debt and equity in 2025 and reduced its weighted average interest rate by 300 basis points.
- Customer concentration: two customers historically accounted for 77% of revenue, though diversification efforts are underway
- Debt burden: nearly $19 billion accumulated to fund expansion, with interest expenses rising faster than operating income
- Execution risk: revenue depends on bringing contracted capacity online on schedule amid GPU supply constraints
- Competition: hyperscalers building proprietary infrastructure could reduce reliance on third-party providers
What to Watch
Investors should monitor CoreWeave’s ability to convert its $66.8 billion backlog into realized revenue while managing debt costs. The company’s first-quarter guidance implies a sequential revenue decline despite surging demand, attributed to the timing lag between infrastructure deployment and revenue recognition. Management said normalized operations could achieve long-term profit margins of 25% to 30%, but reaching that threshold requires scale and execution without stumbles.
Broader sector dynamics also matter. If AI model efficiency improves—reducing compute intensity per inference request—demand for raw GPU capacity could moderate faster than infrastructure operators anticipate. Conversely, if frontier model development continues to require massive scale, CoreWeave’s early infrastructure advantage could prove durable. The next two quarters will test whether the company can narrow losses while doubling capital spending, a high-wire act that will define whether specialized AI infrastructure providers can achieve sustainable unit economics or remain perpetually dependent on external capital markets.