CoreWeave, Inc. 10-Qquarter ended 2026-09-30

Executive Summary

CoreWeave sells GPU-based cloud computing and AI development tools to AI developers and enterprises, earning revenue primarily through committed capacity contracts. It is expanding data centers and infrastructure to serve larger customer workloads, backed by $132B of remaining performance obligations. Q3 revenue grew 168%, faster than the nine-month pace, but infrastructure costs and borrowing expenses left substantial losses. The opportunity is converting contracted demand into productive capacity; the investor question is whether those contracts can support the investment and financing burden without excessive dilution.

What's Happening in the Business

Growth came mainly from customers expanding their use of CoreWeave, rather than new customer wins: existing customers supplied ≈90% of Q3’s revenue increase and 92% over nine months. The platform combines NVIDIA GPUs, networking, storage and orchestration with model- and agent-development tools.

Customer concentration is shifting rather than disappearing. Customer A’s Q3 revenue share fell to 24% from 67%, while Customer B reached 30%. Named commitments include up to ≈$21.0B from Meta through 12/20/2032 and ≈$6.0B from Jane Street, announced in April. These contracts broaden the demand base, but a few customers still determine much of CoreWeave’s growth.

Management expects infrastructure and go-to-market investment to remain elevated and increase relative to 2025. Expansion includes leased capacity and development through joint ventures, adding construction responsibilities alongside operating responsibilities. More equipment is producing revenue, but also depreciation, power bills and rent before financing costs.

Contract status warrants attention: $1.9B of beginning deferred revenue was reclassified as customer liabilities after contracts became cancellable. This does not establish that customers cancelled, but it changes the certainty attached to those advance payments.

Key Financial Highlights

• Infrastructure economics: Q3 gross margin fell to 68% from 73%. Rent, utilities and power, and power-system depreciation increased by ≈$488M, $146M and $119M, respectively. Infrastructure depreciation within technology expense reached ≈$1.7B vs $605M as equipment entered service.
• Cash generation versus earnings: Nine-month OCF reached $5.64B, +276%, despite the common-stockholder loss above. Depreciation reduced earnings without a current cash outflow, while customer receipt timing—including a $3.49B increase from deferred revenue—supported cash flow; receivables absorbed $1.13B.
• Investment and funding: Nine-month equipment and software purchases were $23.6B, +277%, leaving FCF of $(17.9)B vs $(4.75)B. Debt carrying value reached $46.9B, +119% since year-end; cash was $4.12B, with $5.65B of available secured borrowing capacity.
• Future commitments: Uncommenced leases entail estimated undiscounted payments of $47.4B—more than eight times nine-month OCF—beyond existing leases. Contracted future revenue provides support, with 36% expected within 24 months and another 38% in months 25–48, but is neither cash nor profit.
• Dilution: Private placements raised $2.98B net; shares outstanding increased 10% since year-end to 554M. An unused authorization permits another 35M shares, leaving equity funding available at shareholders’ expense.

Risk Assessment

• Financial-reporting controls: Unresolved material weaknesses cover technology controls, segregation of duties and staffing; management reported no resulting material misstatement, but remediation remains unproven.
• Funding and refinancing: Principal payments of $11.0B fall due in 2027, while investment substantially exceeds internally generated cash; tighter credit could force costlier refinancing, dilution or delayed expansion.
• Customer and contract exposure: Concentrated revenue, cancellable contracts and receivables concentrated among four customers expose CoreWeave to payment delays or reduced demand while leases and financing remain payable.
• Capacity execution and suppliers: All deployed GPUs are NVIDIA products, and prior third-party data-center delays and power shortfalls show how supply or construction problems can postpone revenue and trigger customer credits.
• Demand and pricing: More efficient AI models, customers building their own infrastructure and larger cloud competitors could weaken utilization or pricing against long-lived capacity commitments.

What to Watch

• Customer expansion: Existing customers generated most growth. Continued workload expansion and broader customer participation would support the model; cancellations or shrinking commitments would weaken contracted visibility.
• Capacity delivery: Third-party delivery delays and power constraints have already affected execution. New capacity entering service and generating revenue would validate expansion; further delays or credits would expose commitments without matching revenue.
• Funding economics: FCF remains deeply negative and borrowing costs exceed operating income. Improving operating profit and a narrower investment funding gap would strengthen the thesis; further debt and share issuance without improvement would weaken it.