CoreWeave CEO on Nvidia, AI, and Building the Cloud | At Barron's

CoreWeave’s business is built around accelerated computing: combining NVIDIA silicon, CoreWeave software, and large-scale infrastructure to serve training and inference workloads for AI companies. Mike Intrator says the company is not simply renting GPUs; it is delivering a full stack that can run massive parallelized computing jobs better than legacy cloud models. He argues that the shift from sequential cloud computing to parallelized computing is the foundation of the AI era, and that CoreWeave exists because that transition created a new market opportunity.

Intrator also frames CoreWeave as a physical, capital-intensive business rather than a pure software story. He says the company uses a mix of self-built data centers and third-party facilities across the US, Canada, Western Europe and the Nordics, and that tighter energy markets make more control over data-center bricks and mortar strategically important.

On financing, he says CoreWeave structures GPU purchases through debt, with contracts, power agreements and data-center arrangements placed into an SPV “box” so the cash flow pays down the obligation before returning equity value. He rejects bubble concerns, arguing there is a systemic shortage of compute and that the AI economy will keep accelerating as token costs fall and new use cases emerge.