.png){kind=logo}
Michael Intrator pivoted CoreWeave from a crypto-mining venture into an AI-native cloud company that says it now serves nine of the ten leading foundation model providers
Q2 2026 revenue rose 112 % year over year to USD 2.58 billion, with backlog reaching roughly USD 104 billion, alongside a USD 626 million net loss driven mainly by interest expense
Growing resistance to new data center construction, shown by Texas halting new grid connections, is turning power availability into a second constraint alongside GPU supply
CoreWeave has a rare problem since the demand is higher than its available capacity. Yet it still lost $626 million in one quarter. That gap sits at the center of its AI cloud strategy. The man behind it is Michael Intrator, a former natural gas trader who saw GPUs turning scarce early and built a company around owning that scarcity.
Intrator did not arrive at cloud computing through Silicon Valley. He co-founded Hudson Ridge Asset Management, a natural gas hedge fund. Before that, he spent years trading commodities at Natsource Asset Management.
In 2017, he co-founded what became CoreWeave under a different name, Atlantic Crypto, alongside Brian Venturo, Brannin McBee, and Peter Salanki. The original business mined cryptocurrency using GPUs. When crypto margins compressed in 2019, the company renamed itself and pivoted toward renting spare GPU capacity for general computing.
That background offers a useful lens for what happened next. CoreWeave began treating computing capacity the way a trader treats a scarce commodity. GPUs were acquired ahead of demand and contracted against future workloads, rather than simply purchased as equipment.
Also Read: How Modern COOs Build Autonomous Operations with AI Agents
That lens shaped Intrator's next move. Instead of competing with hyperscalers across general-purpose cloud computing, CoreWeave focused on a narrower problem: delivering large volumes of accelerated computing for AI workloads. That focus let the company optimize around GPU density, networking, and orchestration rather than serving every category of enterprise computing.
The result is what the industry now calls the neocloud model. These are infrastructure providers built around accelerated computing rather than the broad mix of workloads hyperscalers handle.
CoreWeave says nine of the ten leading foundation-model providers now run on its platform. That figure followed an April 2026 agreement to support Anthropic's Claude workloads, adding to existing customers including Microsoft, Meta, and OpenAI.
CoreWeave's public listing in March 2025 put that model under closer scrutiny. Investors were not just evaluating whether the company could win AI customers. They were weighing whether contracted demand could generate returns large enough to justify the capital needed to build the infrastructure behind it.
Also Read: CEO 2027 Checklist: 10 Strategic Priorities for Leading an AI-First Business
The second quarter of 2026 sharpened that question. Revenue rose 112 % year over year to USD 2.58 billion, ahead of analyst estimates. Revenue backlog reached roughly USD 104 billion. Net interest expense of USD 640 million drove most of the quarter's USD 626 million net loss.
CoreWeave's problem is not demand. It is the cost of turning demand into capacity. On the August 2026 earnings call, Intrator described the company's near-term capacity as effectively sold out. He framed scarcity as proof the model works, not a flaw to explain away.
Intrator has also pushed CoreWeave into AI inference, running trained models in production rather than only training them. Training creates sharp bursts of demand. Inference creates steady computing needs as models serve users day to day.
For CoreWeave, that means higher use of infrastructure it has already paid to build and a business that reaches beyond training alone. That model is now running into a different limit. The bottleneck is not only how many GPUs CoreWeave can obtain. It is whether the company and the wider power grid can supply the electricity needed to run them.
Texas recently halted new data center grid connections, a sign the constraint is shifting from chip supply to power supply. Intrator has kept framing AI as a lasting shift in computing demand, not a passing cycle.
This makes CoreWeave a reference case for the emerging neocloud category. If its economics hold, specialized AI infrastructure becomes a durable layer of the cloud market. If leverage and softer demand expose cracks in the model, CoreWeave could instead show the limits of infrastructure-first AI growth.
CoreWeave's next phase will test whether sold-out capacity and a USD 104 billion backlog can outrun the interest payments funding them. The model works at an extraordinary scale today. What comes next is not a question about whether AI needs more compute. It is a question about whether the economics of supplying that compute can keep up with how fast the demand is growing.
Michael Intrator is the co-founder, chairman, and CEO of CoreWeave. Before building the AI cloud company, he worked in commodities and co-founded Hudson Ridge Asset Management, bringing a finance and infrastructure-oriented perspective to the business.
Michael Intrator helped transform CoreWeave from a cryptocurrency-mining operation into a specialized AI cloud provider. The company leveraged its early experience with GPUs and built infrastructure specifically for demanding AI workloads.
CoreWeave provides cloud infrastructure designed primarily for AI workloads. Its platform combines GPU computing with networking, storage and software orchestration to support applications such as AI model training and inference.
CoreWeave is considered a neocloud since it focuses heavily on accelerated computing and AI infrastructure rather than trying to serve the broad range of enterprise workloads supported by traditional hyperscale cloud providers.
CoreWeave's biggest challenge is balancing rapid AI demand with the enormous capital required to build and operate GPU infrastructure. Rising GPU, data-center, power and financing costs could determine how sustainably the company can scale.
