The demand for computing power required by artificial intelligence continues to grow at a pace that challenges the industry’s ability to add capacity quickly. CoreWeave CEO Michael Intrator said the company expects global compute supply to remain “wildly overwhelmed” by demand, while the growing move by large enterprises to build dedicated AI infrastructure is reshaping the market.

AI Demand Continues to Outpace Compute Supply

Intrator said CoreWeave is seeing demand from a broad range of customers, including hyperscalers, AI labs and companies deploying agentic AI solutions. According to the CEO, customers are repeatedly requesting larger amounts of computing resources, with their requirements sometimes changing significantly within just a few months.

CoreWeave’s financial results illustrate the scale of that demand. Revenue reached $2.58 billion in the second quarter of 2026, while remaining performance obligations stood at approximately $104 billion at the end of June. The company added nearly 500MW of active power capacity during the quarter, bringing its total to 1.5GW. Total contracted power capacity reached approximately 3.7GW.

Caterpillar Signals a Shift in Enterprise AI Infrastructure

One of the more significant developments highlighted by Intrator is the growing involvement of major industrial companies in building AI infrastructure. Caterpillar, for example, uses CoreWeave’s services while also exploring the development of its own dedicated AI clusters to gain greater direct control over the computing resources required to train and run AI models.

The move reflects a broader shift in corporate technology strategy. After two decades in which companies increasingly moved computing infrastructure to the cloud, AI is making compute capacity a more strategic resource. For businesses relying on AI for critical operations, greater control over infrastructure could provide additional flexibility, performance and availability.

For CoreWeave, however, the trend does not necessarily represent a threat to its business model. The company can provide infrastructure and software capabilities to enterprises that choose to own their hardware, allowing it to participate in the AI infrastructure market even when customers do not rely entirely on public cloud capacity.

AI Infrastructure Spending Continues to Rise

CoreWeave raised its 2026 capital expenditure forecast in August to $35 billion to $39 billion, up from its previous estimate of $31 billion to $35 billion. Capital expenditures reached $9.4 billion in the second quarter alone, compared with $6.8 billion in the previous quarter.

The expansion is taking place in a market where data center construction requires enormous amounts of electricity, advanced chips, land and financing. At the same time, local opposition to new data centers and difficulties connecting new projects to power grids could slow the pace at which additional capacity becomes operational. As a result, companies announcing plans to build gigawatts of infrastructure still face a separate challenge: turning those plans into functioning compute capacity.

The remainder of 2026 will be an important test for CoreWeave and the broader AI infrastructure market. Investors will be watching the pace of capacity additions, new contracts, capital spending and the ability of companies to bring planned infrastructure online. The expansion of industrial companies such as Caterpillar into dedicated AI clusters could also signal a new phase for the market, in which demand for compute is no longer driven primarily by technology giants and AI labs but is becoming a core infrastructure requirement across a much broader range of industries.