AI infrastructure has to survive its borrowing costs as well as deliver better models. In the AI-finance segment of this week’s Hard Fork, beginning around 23:54, finance reporter Rob Copeland describes CoreWeave as an intermediary that buys processors, installs them in data centers, and rents computing capacity to AI companies. He cites roughly $70 billion in liabilities across a business operating 51 data centers, then describes a falling share price and growing debt-servicing pressure. Those are the episode’s reported figures, not an independently checked balance sheet; liabilities are broader than debt. Copeland sees the company as a possible early warning for a buildout financed on the assumption that growth keeps going. Technology reporter Erin Griffith connects that exposure to Anthropic’s public-market prospects and the limits of customers’ willingness to pay: infrastructure suppliers depend on demand turning into enough revenue to support the capacity being built. Their concern is a chain of commitments, from paying customers to model companies to the businesses financing the machines. Copeland expects the financial story to move faster than the technological one, but that remains a forecast, not evidence of an imminent default. The episode’s accompanying reading on AI borrowing gives the discussion its wider context.
“For my seat, it’s a financing story as much as anything else.” — Rob Copeland, around 25:19
“The financial aspect of this story, we’re going to see play out, in my opinion, faster than the technological side.” — Rob Copeland, around 25:48