CoreWeave closed at $71.88 on July 24, down 48% from its May high and 53% from its 52-week peak, even as the underlying demand for its GPU rental business kept growing. First-quarter revenue reached $2.08 billion, up 112% year over year, and the contracted revenue backlog hit $99.4 billion after the strongest bookings quarter in the company’s history. Active power crossed 1 gigawatt across 49 facilities, with total contracted power above 3.5 gigawatts. Very few businesses in any industry have scaled this fast.

The demand story is not the problem. Microsoft, OpenAI, Meta, and Anthropic have signed contracts that add up to one of the best demand signals private markets produce, and operating cash flow turned strongly positive at $2.98 billion in a single quarter. The problem sits one line lower on the income statement. Depreciation consumed 55.2% of first-quarter revenue and net interest another 25.8%, against an adjusted EBITDA margin of 55.7%. Netting the two real costs of the business model against that cash margin leaves negative $526 million, and adjusted operating margin has collapsed from 17% to just 1% in a single year as depreciation on the 2024-25 buildout catches up with the revenue it generates.

That arithmetic is the entire investment case in miniature. CoreWeave is, in economic substance, a leasing business: it converts capital and electrical power into contracted cash flows, and its returns depend on the spread between the rental yield on its GPU fleet and the cost of the debt that financed it. Total debt including leases stands at $35.15 billion against an Altman Z-score of 0.36, deep in the distress zone, and the 2026 capital spending plan of $31-35 billion requires roughly $25 billion more of external funding. The bull case, that contracts renewed after the first depreciation cycle carry near-100% incremental margin, is plausible but entirely unproven as of today.

A probability-weighted fair value near $89 sits about 24% above the current price, but that figure is earned by accepting a 30% chance of a bear case that loses 61%, with a bull case of $170 on the other side. That is too wide a distribution, at too unfavorable a moment, to justify buying ahead of second-quarter results on August 11. A sequential improvement in adjusted operating margin, evidence that renewals are pricing well on depreciated assets, or a pullback into the $50s that already discounts the bear case would each change the calculus. None of those conditions is met today.

The full financial breakdown, five-driver framework, and scenario analysis are available in the complete CoreWeave (CRWV) report on our Reports page.