The upside case does not require a pure-software
rerating. It requires CoreWeave to keep capacity scarce, convert
backlog into live revenue on time, and make higher-margin
inference, storage, networking, and control surfaces material enough that customers buy outcomes and reliability, not just GPU hours. If that happens, balance-sheet drag should matter less by 2031 and the equity can compound without perfection.