Key Takeaways

  • CoreWeave bond yields spiked to 15%, with credit default swap spreads pricing in a 55% probability of default over five years.
  • Chamath Palihapitiya calculated that CoreWeave must raise another $35 billion in debt to buy and install servers required by signed contracts.
  • Ben Horowitz explained that Andreessen Horowitz avoided neo-clouds because LoudCloud got crushed when bandwidth collapsed from $1,500 per megabit in 1999 to $50 by 2001.
  • David Sacks countered that market panic is ordinary volatility when investors underwrite rapid growth toward multi-trillion dollar outcomes.

The Disagreement

Chamath Palihapitiya argued that CoreWeave is running into a severe financing squeeze. The bond market is charging 15% interest, and credit default swap spreads imply a 55% chance of default over a five-year horizon. CoreWeave signed huge customer commitments, but buying the hardware requires immense cash. As Palihapitiya calculated: “What this basically says is at the midpoint, he has to raise another $35 billion of debt to get all of the contracts that he signed financed, right? All the boxes financed.”

Ben Horowitz backed Palihapitiya by citing his experience operating LoudCloud during the dot-com crash. “In 1999 bandwidth was $1,500 a megabit and you couldn't get it, was very hard to get it, and so you had to get a long-term contract three years,” Horowitz said. “By 2001 it was $50 a megabit. So if you were holding bandwidth and you borrowed money to get it, you're dead.”

David Sacks rejected the panic. He argued that compute demand is real, durable, and structurally distinct from past bubbles. Sacks explained: “I would make the argument that this is just volatility, which you're going to get a lot of. Whenever you're underwriting to multi-trillion dollar outcomes in the future based on very rapid rates of growth, exponential rates of growth, the market's going to be super sensitive to any data points we have in the present.”

Who's Right (and When They're Wrong)

Palihapitiya and Horowitz have the arithmetic on their side. Palihapitiya summarized the vulnerability: “Any business model that thrives on a financing spread gets cornhold. It will not matter when the 10 years at 6%.”

Neo-clouds borrow heavily to purchase specialized hardware, planning to rent that capacity out at rates that clear their borrowing costs. That business model requires two conditions to survive: low interest rates and persistent hardware scarcity. When risk-free rates rise, borrowing costs explode. At 15% yields, the margin vanishes. Meanwhile, if chip supply catches up to demand or next-generation processors render current chips obsolete, rental prices drop fast. The company remains stuck paying off peak-cycle debt with trough-cycle revenue.

Sacks is correct that secular demand for compute remains strong. But being right about long-term demand does not save an equity holder from debt distress. In 2002, internet traffic kept rising, yet the telecom operators collapsed under debt. Tech startups bought cheap bandwidth from bankruptcy auctions while the original builders were wiped out.

What to Do With This

Check your infrastructure dependencies before the end of the week. If your startup relies on cheap compute contracts with secondary cloud vendors, assume those prices will face pressure or terms will tighten. Limit your committed capacity agreements to 12 months, and test an automated fallback script that can switch your core inference workloads to a hyperscaler like AWS or Azure if your provider faces credit distress.