CoreWeave Q2 Revenue Doubles, But $6.4B Capex and Net Loss Reveal Complex AI Cloud Reality

Woofun AI reports that CoreWeave filed an 8-K document with the SEC on August 11, unveiling second-quarter earnings that highlight a stark divergence between top-line growth and operational complexity. The release, covering the period ending June 30, presents a narrative where rapid revenue expansion is counterbalanced by significant structural challenges in capital deployment and profitability.

The headline figures underscore a dramatic surge in demand, with Q2 revenue reaching $2.575 billion, marking a 112.5% increase year-over-year. This growth is supported by a substantial pipeline of future business, including approximately $104 billion in backlog orders. Simultaneously, the company reported 1.5 gigawatts of active power capacity, serving as the current baseline for its operational footprint.

However, the financial reality extends beyond these aggregate totals, requiring a simultaneous analysis of the income statement, cash flow statement, and power resources. Revenue recognition, construction cost incurrence, and power resource availability occur at different stages of the value chain. These temporal mismatches create a complex picture where contracted capacity often exceeds current active utilization, obscuring the true pace of monetization.

Despite the revenue jump, the GAAP operating result deteriorated significantly, shifting from a profit of $19 million in the prior year to a loss of $49 million. This metric, which reflects core business operations more accurately than net income, indicates that the company's revenue growth has not yet translated into operating profitability. The widening gap between top-line growth and operating results highlights the heavy cost structure associated with scaling AI infrastructure.

Compounding the operational loss is a sharp rise in financing costs, with net interest expense for the second quarter reaching $640 million. This represents an increase of $373 million compared to the previous year, driven by the substantial debt required to fund expansion. The elevated interest burden is a primary contributor to the widening net loss, illustrating the financial leverage inherent in the company's growth model.

To provide a clearer view of operational efficiency, CoreWeave reported an adjusted EBITDA of $1.51 billion, yielding a margin of 59%. This non-GAAP metric is designed to assess performance by excluding certain non-cash and non-recurring items.

However, it cannot replace the GAAP operating result, and relying solely on adjusted metrics risks misinterpreting revenue growth as evidence that profits are keeping pace with construction costs.

Cash flow dynamics further illustrate the intensity of capital deployment, with operating activities generating a net inflow of $679 million. In contrast, expenditures on property, equipment, and capitalized software totaled $6.422 billion, roughly 9.5 times the operating cash flow. This disparity underscores the upfront nature of AI cloud infrastructure, where massive capital outlays precede revenue generation from delivered services.

Woofun AI data shows that financing activities provided the primary liquidity buffer, with a net inflow of $10.071 billion offsetting a $7.166 billion net outflow from investing activities. These financing proceeds included debt issuance, private equity offerings, and the purchase of capped call options for hedging. The reliance on external capital is evident, as internal operating cash flows are insufficient to cover the scale of ongoing investments.

Looking ahead, management expects capital expenditures to range from $11.5 billion to $13.5 billion in the third quarter, signaling continued aggressive expansion. At the end of Q2, contracted power capacity stood at 3.7 gigawatts, approximately 2.5 times the 1.5 gigawatts of active capacity. This gap highlights the challenge of converting committed resources into billable services, as contracted capacity does not equate to immediate delivery readiness.

The $104 billion backlog includes remaining performance obligations and estimated future recognitions, but it does not include additional net customer commitments of over $25 billion at the start of Q3. These figures cannot be treated as locked-in quarterly revenue or converted into fixed timelines based on current performance. The distinction between contracted resources and active GPUs remains critical, as the timeline for delivering these commitments depends on infrastructure readiness rather than mere contractual obligation.