Down 53%, Should You Still Buy CoreWeave's (CRWV) Stock?

CoreWeave shares have fallen about 53% from their June 2025 high of $183.58 to roughly $86, despite 2025 revenue rising 168% to $5.1B and adjusted EBITDA increasing 154% to $3.1B. Analysts forecast 2026 revenue growth of 151% to $12.9B and adjusted EBITDA growth of 144% to $7.5B, supported by a $104B contracted backlog, up 246% year over year. The investment case remains high risk: CoreWeave is GAAP-unprofitable, has repeatedly raised debt and equity, and reported a 14.3 debt-to-equity ratio, leaving it vulnerable to higher rates or slower AI infrastructure demand.

Analysis

The central valuation error is treating adjusted EBITDA as a cash-flow proxy. CRWV’s economics are dominated by GPU depreciation, financing costs and recurring capacity build-outs; a low headline EBITDA multiple is not meaningful unless contracted revenue converts into operating cash flow after maintenance and growth capex. The key diligence item is backlog quality: cancellation provisions, customer concentration, pricing escalators and the share tied to take-or-pay commitments rather than framework agreements. The claimed performance advantage versus hyperscalers is marketing, not a durable moat, absent independent benchmarks and evidence that customers cannot replicate capacity through AWS, Azure, Google Cloud or direct GPU leases.

NVDA is a second-order beneficiary, but its ownership and supply relationship create circularity risk: incremental CRWV financing can support GPU demand while increasing the probability that end-customer AI spending is being pulled forward. AMZN and MSFT face limited near-term displacement because they monetize AI through broader cloud, software and enterprise bundles; however, a sustained CRWV price/performance advantage could pressure their GPU-cloud margins before it materially affects revenue. The more acute risk sits with CRWV equity holders: a modest utilization shortfall or delayed customer deployment can impair debt service coverage given fixed infrastructure costs.

Near term, the stock can rally on backlog additions, financing availability and evidence of higher utilization; over 1-3 months, quarterly capex, free-cash-flow burn and debt refinancing terms matter more than adjusted EBITDA. Over 6-18 months, hyperscaler self-build, custom silicon adoption and lower GPU rental pricing could compress returns on newly deployed clusters. Consensus appears too focused on revenue visibility and insufficiently focused on whether the company can fund the next capacity cycle without incremental dilution or materially higher borrowing costs.

Key Decisions for Investors

• Do not chase CRWV on the valuation headline; establish a watch position only after the next earnings release confirms operating cash-flow conversion, utilization and no deterioration in debt pricing. Add only if management demonstrates contracted revenue is translating into cash receipts rather than further working-capital and capex consumption.
• For a defined-risk bullish expression, consider a 3-6 month CRWV call spread rather than common equity, sized small. The thesis requires a utilization/backlog catalyst; exit if guidance implies another material equity raise or if interest expense rises faster than adjusted EBITDA.
• Use a relative-value basket: long NVDA versus short a small CRWV position if evidence emerges that GPU demand remains strong but specialized-cloud returns are compressing. This isolates continued AI hardware demand from CRWV’s financing, concentration and execution risk.
• Monitor MSFT and AMZN cloud disclosures for GPU capacity pricing and AI infrastructure margins. A material reduction in GPU rental pricing or expanded first-party capacity would falsify the standalone-specialist premium and would be a negative catalyst for CRWV within 1-2 quarters.