CoreWeave backlog nears $100B

CoreWeave is basically a very expensive middle layer in the AI boom. It rents out GPU-heavy cloud infrastructure to companies building and running large AI models. On May 7, the company showed just how strong that demand still is — first-quarter revenue hit $2.08 billion and revenue backlog reached $99.4 billion. But the same report also showed why investors still get nervous around this business. ### What is CoreWeave actually selling? CoreWeave sells AI compute — mostly access to clusters of high-end GPUs, networking, software, and the operational layer needed to train and run models at scale. That matters because AI customers do not just need chips. They need power, cooling, data-center space, orchestration software, and someone who can turn all of that into usable capacity fast. CoreWeave’s pitch is that it sits “between the models and the silicon,” which is a neat way of saying it tries to make raw hardware usable for AI companies and enterprises. (investors.coreweave.com) ### What changed this quarter? The headline change was size. Revenue rose from $982 million a year earlier to $2.078 billion in the March quarter. The company also said this was its strongest bookings quarter ever, pushing revenue backlog to $99.4 billion as of March 31, 2026. That backlog figure is the part everyone latched onto, because it suggests customers are still lining up for future capacity even after the first wave of AI spending. (investors.coreweave.com) ### Why does backlog matter so much? Backlog is not the same thing as cash in hand, but it is a strong signal of committed demand. For a company like CoreWeave, that matters more than usual because the whole business requires massive up-front spending. If customers are willing to sign big multiyear commitments, CoreWeave can justify building out more power and data-center capacity. The company said it has now surpassed 1 gigawatt of active power and believes it is on track for more than 8 gigawatts by 2030. (investors.coreweave.com) ### So why were investors uneasy? Because the income statement is still brutal. Operating expenses grew to $2.222 billion. Operating loss widened to $144 million. Net loss reached $740 million, and net interest expense alone was $536 million. Even on the adjusted numbers investors like to use for infrastructure businesses, margins slipped — adjusted EBITDA margin fell to 56% from 62%, and adjusted operating income margin fell to 1% from 17%. (investors.coreweave.com) That is the core tension here: demand looks real, but financing and scaling that demand is expensive. ### Why is interest expense such a big deal? Because this business eats capital. CoreWeave has to secure GPUs, build or lease data-center capacity, and lock in power before revenue fully arrives. That means debt matters a lot. A backlog near $100 billion sounds amazing — and it is — but if the company has to spend aggressively just to fulfill those contracts, investors will keep asking how much of that future revenue turns into durable profit. (investors.coreweave.com) The catch is simple: AI infrastructure can look like a gold mine and a balance-sheet stress test at the same time. ### Where does Nvidia fit in? Nvidia is still the center of gravity because its GPUs remain the most sought-after hardware for advanced AI workloads. That gives suppliers like Nvidia huge leverage, but it also creates concentration risk for cloud providers and customers. If Nvidia cannot expand supply fast enough, buyers will look harder at alternatives — especially AMD — and infrastructure companies like CoreWeave will have to stay flexible about what hardware they deploy. (investors.coreweave.com) The supplier stack is becoming part of the investment story, not just a technical detail. ### Is this a demand story or a margin story? It is both. The bullish read is that almost no company gets to a $99.4 billion backlog unless customers urgently want what it sells. The bearish read is that hyperscale-style infrastructure businesses can grow incredibly fast while still punishing shareholders if capital costs, depreciation, and interest keep piling up. CoreWeave’s quarter did not resolve that argument. It sharpened it. (investors.coreweave.com) ### Bottom line CoreWeave just proved that AI compute demand is still enormous. Now it has to prove something harder — that it can turn that demand into a business that scales without drowning in the cost of building it. (investors.coreweave.com)