LIVINGSTON, N.J., May 7, 2026, 17:01 EDT
CoreWeave Inc came in ahead of quarterly revenue forecasts Thursday, reporting first-quarter sales of $2.08 billion—buoyed by rising demand for AI computing power in its cloud unit. The Nvidia-backed firm disclosed a revenue backlog of $99.4 billion. Net loss increased, with higher costs weighing on results.
The report dropped following the U.S. market close, hitting just as CoreWeave—one of the hottest AI infrastructure names—hangs in the spotlight. Earlier Thursday, Bloomberg said CoreWeave shares had been ripping higher in 2026, ramping up pressure on the company to show it can actually deliver on its buildout goals.
The reason: CoreWeave operates in a niche slice of cloud computing—offering rentals of powerful chips and infrastructure for AI model training and inference. According to Reuters, demand has jumped for these “neoclouds”—smaller players like CoreWeave and Nebius, focused squarely on AI workloads—as businesses hurry to lock in access to scarce hardware. Reuters
Revenue landed close to the $1.97 billion analysts had penciled in, LSEG numbers cited by Reuters show. Operating expenses surged, jumping past double to $2.22 billion. Net loss deepened, coming in at $740 million, up from $315 million the year before.
Michael Intrator, CoreWeave’s chief executive, described it as the “strongest bookings quarter” the company has seen. “We sit between the models and the silicon,” he said. CoreWeave has now surpassed 1 gigawatt of active power and is aiming for over 8 gigawatts by 2030. CoreWeave
The company’s revenue backlog — its projection for future revenue locked in by customer contracts — climbed above the $66.8 billion seen at December’s close. Among the latest: a $21 billion boost in cloud capacity for Meta Platforms, a $6 billion agreement with Jane Street, and a fresh deal with Anthropic.
CoreWeave reported it’s bumped up its total contracted power by an additional 400 megawatts, now surpassing 3.5 gigawatts. For AI data center operators, active power matters—it’s the real measurement of how much electrical load is currently switched on and ready to drive compute tasks.
The price tag of the race is still steep. Capital expenditures hit $6.8 billion for the quarter—off from $8.2 billion last quarter, but miles ahead of the $1.9 billion logged a year ago. That’s money poured into long-term assets like data-center gear and infrastructure.
Adjusted EBITDA jumped to $1.16 billion, up from $606 million a year ago—this figure excludes interest, taxes, depreciation, amortization and certain other expenses. Adjusted operating income, however, slid to $21 million compared with $163 million. The adjusted operating margin also tightened, dropping to 1% from the previous 17%.
Shares barely budged in choppy after-hours action following the results, Reuters said. Investors seemed caught between CoreWeave’s surging demand and a cost structure that continues to expand just as quickly.
Still, the setup isn’t without its pitfalls. Interest costs shot up, hitting $536 million. Debt, both current and non-current, stacked up to $24.9 billion at the end of March, according to the company’s balance sheet. Delays in customer rollouts, softer AI demand, or a bump in financing expenses could stall how quickly the backlog turns into actual cash.
CoreWeave plans to share forward-looking guidance during its earnings call. Now comes the real question: out of the $99.4 billion backlog, how much can actually turn into revenue—without squeezing margins further, increasing debt, or straining the power supply.
