CoreWeave, Inc. 10-Qquarter ended 2026-06-30

Executive Summary

CoreWeave sells GPU-based cloud computing to AI developers and enterprises, earning revenue primarily through committed service contracts. It is expanding its integrated AI platform and data-center capacity to serve larger workloads from existing customers. Q2 revenue rose 112%, but infrastructure costs and financing expenses widened losses as the company built capacity ahead of future revenue. Its $103.7B of remaining contractual obligations offers substantial growth visibility, but investors must weigh that opportunity against heavy borrowing, delivery risk and customer concentration.

What's Happening in the Business

• Customer expansion drove growth: Existing customers generated ≈93% of the revenue increase in both Q2 and H1. CoreWeave is combining GPU infrastructure, networking and storage with tools for model and agent development, seeking to serve more of customers’ AI workloads rather than computing capacity alone.
• Large commitments are broadening the customer base: These include Meta’s initial commitment of up to $21.0B and Jane Street’s ≈$6.0B commitment announced in April. Customer A’s Q2 revenue share fell from 71% to 36%, while Customer B reached 26%; dependence has shifted toward several large accounts rather than disappeared.
• Capacity expansion is accelerating: H1 capital spending supported GPUs, servers, networking and data centers, while net property and equipment rose 53% from year-end. Development joint ventures extend CoreWeave’s involvement into building facilities, increasing control over capacity but also construction and funding exposure.

Key Financial Highlights

• Cost pressure: Cost of revenue reached 34% of revenue from 26%, reflecting rent, power and depreciation; technology and infrastructure expense also grew faster than sales as capacity expanded.
• Cash conversion: H1 operating cash flow benefited from $2.54B of depreciation and amortization and a $1.37B increase in deferred revenue. Customer prepayments help fund construction, but do not eliminate the investment shortfall.
• Earnings quality: Unrealized strategic-investment gains contributed $109M in Q2 and $111M in H1, cushioning losses without representing cloud-service operating improvement.
• Financing and dilution: H1 private placements raised $2.98B net, while common shares outstanding increased 9.8% to 551M. Stock-based compensation was $318M; expansion is being financed partly through shareholder dilution.
• Liquidity: Cash, marketable securities and available facilities totaled $15.6B, up from $6.86B at year-end. Much of that capacity is borrowing availability rather than cash, tying continued expansion to access to financing.

Risk Assessment

• Financial-reporting controls: Unresolved material weaknesses cover financial-system controls, segregation of duties and staffing; no material misstatement resulted, but remediation still requires implementation and testing.
• Financing and fixed commitments: Debt principal payments total $35.6B, alongside $16.3B of operating lease liabilities and $35.5B of payments on leases not yet commenced, creating substantial obligations before new capacity earns revenue.
• Customer and AI-demand concentration: The three largest customers supplied ≈72% of Q2 revenue, so a default, spending slowdown or reduced need for computing could leave infrastructure obligations unsupported despite take-or-pay contracts.
• Delivery and power: Previously disclosed third-party data-center delays and a power shortfall that caused customer service credits demonstrate how construction or utility constraints can disrupt contracted deployments.
• Supplier and technology dependence: All deployed GPUs are NVIDIA products, while larger cloud competitors can bundle or underprice services; supply disruption, faster hardware obsolescence or pricing pressure could weaken returns on expensive infrastructure.

What to Watch

• Contract conversion: CoreWeave expects 41% of remaining performance obligations to become revenue within 24 months. Timely deployments and revenue recognition would validate the expansion; delays or contract changes would weaken that visibility.
• Customer breadth: Existing accounts supplied ≈93% of revenue growth despite a less concentrated largest-customer share. Further growth across Meta, Jane Street and other customers would broaden the base; renewed concentration would preserve dependence on a few buyers.
• Capacity delivery: Data-center expansion and development joint ventures are under way following earlier delivery delays. Facilities entering service without customer credits would support execution; further power or construction problems would undermine it.
• Funding economics: H1 FCF was deeply negative and interest expense rose sharply. Improving operating profitability and a narrower cash shortfall would support the investment case; continued borrowing and share issuance without those improvements would weaken it.