CoreWeave, an American cloud computing company that buys Nvidia graphics processing units (GPUs), installs them in data centres, and rents the capacity to artificial intelligence developers, reported second-quarter 2026 revenue of $2.58 billion on 11 August. The figure beat the analyst consensus estimate of $2.56 billion compiled by financial data group LSEG and more than doubled from $1.21 billion in the same period a year earlier, according to Reuters.
It was, by the Wall Street Journal's account, the company's fifth consecutive quarter of record revenue — a streak that captures both the pace of enterprise AI spending and the speed at which CoreWeave has scaled to meet it.
The company's contracted revenue backlog reached $104.2 billion as of 30 June, up from $99.4 billion at the end of the first quarter, according to Channel NewsAsia. That figure does not include more than $25 billion in net new customer commitments secured in early Q3. A backlog of that size is, in effect, a long queue of signed contracts for compute that has not yet been delivered — a measure of demand confidence, though subject to delivery and availability conditions that mean it may not be fully realised.
The revenue growth comes at a considerable and rising price. Technology and infrastructure expenses — the cost of running the GPU clusters that underpin everything — surged 125% to $1.51 billion in the quarter, up from $670 million a year earlier. Capital expenditure hit $9.4 billion in Q2, compared with $6.8 billion in Q1 2026 and $2.9 billion in Q2 2025, according to Channel NewsAsia. In response, CoreWeave raised its full-year 2026 capital expenditure forecast to between $35 billion and $39 billion, up from a prior range of $31 billion to $35 billion.
The losses are widening in step with the spending. CoreWeave posted a net loss of $626 million in Q2, more than double the $290 million loss recorded in the same quarter a year earlier. The mechanism is not subtle: net interest expense reached $640 million in the quarter, itself more than double the $267 million incurred in Q2 2025 — meaning the interest bill alone exceeded the reported net loss, according to Quartz. The company is borrowing heavily to build infrastructure, and the cost of that borrowing is compounding.
To fund the expansion, CoreWeave raised more than $10 billion through unsecured debt and convertible bonds during the quarter, secured a $3.1 billion term loan, and received a $1 billion strategic investment from trading firm Jane Street, Quartz reported. The company has relied on debt markets to finance a growth strategy premised on AI demand continuing at its current pace.
On adjusted measures, the quarter looked more comfortable. Adjusted earnings before interest, tax, depreciation and amortisation (EBITDA) reached $1.51 billion at a 59% margin, down from 62% a year earlier. The adjusted per-share loss of $1.03 came in better than the $1.20 Wall Street had expected. The generally accepted accounting principles (GAAP) loss per share of $1.14 also beat analyst estimates of $1.41, according to Yahoo Finance. Adjusted operating income of $128 million exceeded analyst expectations of $66 million.
The company's client roster continues to expand. CoreWeave signed a $21 billion agreement to supply AI cloud capacity to Meta through 2032, layering on a prior $14 billion commitment, and reached a separate multi-year compute agreement with AI safety company Anthropic for its Claude language models, Quartz reported. CoreWeave also counts Microsoft and OpenAI among its major customers.
On the infrastructure side, the company expanded its active power capacity by nearly 500 megawatts in the quarter, bringing the total to 1.5 gigawatts. It completed the first bring-up and validation of Nvidia's Vera Rubin NVL72 chip system — a next-generation GPU platform — and was added to the Nasdaq-100 Index during the period.
What remains genuinely uncertain is how the financial picture evolves as debt service and depreciation accumulate. The $104 billion backlog is an impressive number, but its realisation depends on CoreWeave's ability to build and operate capacity at the contracted scale. Margin trends will be shaped by energy and component costs that are not fixed. And should enterprise AI spending slow — even modestly — the arithmetic of heavy leverage against long-dated contracts becomes considerably less comfortable.
Comments (0)