Quick Facts

  • CoreWeave targets $12.4 to $13.2 billion in 2026 revenue after posting $5.13 billion in 2025, a 168% year-over-year gain.
  • The company holds a $104 billion revenue backlog as of Q2 2026, excluding more than $25 billion in new Q3 commitments.
  • New customer contracts signed in Q2 carry contribution margins 5 to 10 percentage points above recent quarters, after a roughly 25% price increase across SKUs in July.

CoreWeave built its business on GPU availability during a period of acute scarcity. Now analysts are asking whether the company can hold customers when supply tightens less sharply. A new Breaking Analysis from theCUBE Research, published ahead of CoreWeave's Fully Connected conference, draws on 13 in-depth interviews totaling more than seven hours to examine that question directly.

Revenue grew from $229 million in 2023 to $1.92 billion in 2024 to $5.13 billion in 2025. Management guides to $12.4 to $13.2 billion for full-year 2026, with adjusted operating income of $960 million to $1.15 billion.

Q1 2026 revenue came in at $2.08 billion, exceeding the $1.97 billion analyst consensus and more than doubling from $981.8 million in the same quarter a year earlier. CoreWeave ended Q1 with a $99.4 billion revenue backlog and raised $8.5 billion in new debt during the quarter, following deals with AI startups Cline and Perplexity.

By the end of Q2 2026, the company had secured more than $20 billion in debt and equity for the year and closed with nearly $25 billion in total debt. Active power capacity reached 1.5 gigawatts after adding nearly 500 megawatts during the quarter. Contracted power stood at 4.2 gigawatts as of the Q2 earnings call, with a target of more than 1.85 gigawatts in active power by year-end. Capital expenditures are projected at $35 to $39 billion annually.

Gross margins ran at approximately 74% in 2024. Adjusted EBITDA margin reached 62% in 2024 when GPU hardware ran at full utilization. The July price increase of roughly 25% across SKUs, combined with component cost pass-throughs, pushed new contract margins higher in Q2.

CEO Michael Intrator framed the company's position in Q2 results: "CoreWeave reached an important inflection point this quarter as our scale began to translate into expanding operating leverage." He added that the shift from training to inference strengthens CoreWeave's position. "As the market moves from training to inference, that distinction matters more than ever. CoreWeave was built for exactly this."

The inference market is growing fast without displacing training. Futurum projects agentic and reasoning inference growing 219% year over year in 2026, rising from roughly $36 billion last year to $546 billion by 2030. More than 40% of enterprises already run some form of AI agent in production. CoreWeave views expiring GPU contracts on older hardware as a source of inference capacity at higher margins.

Two major customer deals anchor the backlog. CoreWeave and Meta Platforms signed a long-term agreement worth approximately $21 billion running through December 2032, covering multiple locations and including early deployments of the NVIDIA Vera Rubin platform. A separate agreement with Jane Street is valued at $7 billion.

Intrator has set a long-term power target of more than 8 gigawatts by 2030. CFO Nitin Agrawal said Q2 results reflect "disciplined execution against the strategy we outlined at our IPO to develop one of the largest AI Cloud footprints in the world."

The theCUBE Research analysis identifies four indicators to watch at Fully Connected: proof that pilots convert to production, how customers respond to repricing and renewals, whether approved workloads expand beyond current use cases, and whether deployed capacity generates cash returns at the pace the financial model requires. Those factors, analysts argue, determine whether a GPU availability advantage becomes a business built to last.

Read more: CoreWeave's next test: From GPU scarcity to a durable AI cloud

This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.