CoreWeave posted Q1 2026 revenue of $2.08 billion, more than doubling year-over-year from $981.8 million and beating LSEG consensus of $1.97 billion. The result is the clearest public-market confirmation yet that enterprise demand for GPU-backed inference infrastructure has entered a sustained, large-scale capex cycle — not a speculative build-out.
CoreWeave leases Nvidia GPU capacity from its own data centers to AI model developers, charging for compute-hours. In Q1, technology and infrastructure costs jumped 127% to $1.27 billion, reflecting buildout pace. Sales and marketing spend rose more than sixfold to $69 million as the company broadens its client base. CEO Mike Intrator said 10 clients are now committed to spending at least $1 billion on CoreWeave products — up from a roster dominated by Microsoft, which accounted for 62% of 2024 revenue. Intrator told analysts the company has "reached hyperscale."
CoreWeave's balance sheet reflects the same aggression. The company raised $8.5 billion in new debt during the quarter, ending with roughly $25 billion in total debt. It has secured more than $20 billion in combined debt and equity in 2026 to date. Nvidia, a key supplier and strategic backer, purchased an additional $2 billion in CoreWeave stock. To support expansion, management revised its 2026 capex forecast to $31–$35 billion, citing component price pressures — though Intrator said the company has the supply-chain relationships to absorb the cost. CoreWeave reiterated its plan to bring 1.7 gigawatts of power online by year-end, out of 3.5 gigawatts of contracted power.
The $99.4 billion revenue backlog signals that CoreWeave's near-term supply is effectively pre-committed, with limited spot availability. Teams not already in queue for multi-year agreements will face growing lead times, and the window to negotiate competitive per-GPU-hour rates against a capacity-constrained provider is narrowing. The 10 clients with billion-dollar commitments include hyperscaler-scale AI labs; mid-market enterprises are competing for a smaller share of available capacity.
Net loss widened to $740 million in Q1, compared with $315 million a year earlier. Adjusted EPS came in at a loss of $1.12, versus analyst expectations of a $0.90 loss. Second-quarter revenue guidance of $2.45–$2.6 billion missed the $2.69 billion consensus, sending shares down as much as 10% in after-hours trading despite a nearly 80% year-to-date gain heading into the print. CoreWeave carries the operating leverage of a capital-intensive infrastructure business layered on top of early-stage revenue concentration risk.
S&P upgraded CoreWeave's credit outlook to positive from stable during the quarter, signaling that rating agencies view the backlog as credible revenue coverage for the debt load. Management reiterated that annualized revenue should exceed $30 billion by year-end 2027. The trajectory depends on inference demand remaining structurally elevated — on OpenAI, Anthropic, and the broader generative AI ecosystem continuing to consume compute faster than hyperscalers can provision it through their own builds.
CoreWeave's Q1 report sets a concrete benchmark: GPU cloud as a standalone, publicly traded infrastructure category can sustain triple-digit revenue growth at multi-billion-dollar scale. For CTOs weighing build-versus-buy on inference capacity, the implicit message is that the "buy" side of that equation is becoming structurally more expensive every quarter.
Written and edited by AI agents · Methodology