- CoreWeave reported Q3 2026 revenue of $3.66 billion, up 168% year-over-year, alongside a $709 million net loss.
- The company expects to collect roughly $9 billion in customer prepayments and deposits over the next three months.
- CEO Michael Intrator said CoreWeave will not use the unsecured debt market to fund the business at current yields.
- CoreWeave’s backlog stands at $132.5 billion, and the company raised its full-year revenue guidance to $13-$13.3 billion.
What Happened
CoreWeave reported third-quarter 2026 revenue of $3.66 billion, up 168% from a year earlier, alongside a $709 million net loss and $846 million in quarterly interest expense, according to TechCrunch’s report on the November 11, 2026 earnings call. CEO Michael Intrator said, “We will not use the unsecured debt market to fund our business at current yields, nor do we need to.”
Why It Matters
CoreWeave is rewriting how it finances AI infrastructure buildout after years of relying on debt markets, following prior moves like its $8.5 billion GPU loan backed by a Meta deal and its $21 billion Meta GPU capacity expansion. The shift toward prepayments reflects pressure on neoclouds to prove they can fund growth without over-leveraging balance sheets.
Technical Details
CoreWeave expects to collect roughly $9 billion in customer prepayments and deposits over the next three months, though more than $3 billion of that figure covers existing infrastructure that has already been fully paid off. The company raised its full-year 2026 revenue guidance to $13 billion-$13.3 billion, and its backlog now stands at $132.5 billion.
Who’s Affected
Customers renting CoreWeave’s GPU capacity may face higher prices or shorter contract terms as the company shifts toward prepayment- and deposit-based financing instead of unsecured debt. The strategy depends on current AI compute demand and pricing power holding steady.
What’s Next
CoreWeave plans to rely on loans backed by specific customer contracts, drawing funds as needed to build out the infrastructure behind its backlog. The approach only holds if today’s demand for AI compute persists at current levels.
