The upside case is that CoreWeave remains a scarce AI deployment layer rather than a temporary GPU reseller.
Backlog, active-power growth,
inference demand, and higher-margin software attach can expand revenue sharply, while flexible pricing and third-party software distribution make the model more resilient. I assume some multiple compression from peak scarcity conditions, but not a collapse, because power access, operational speed, and workflow integration should still matter in 2031.