CoreWeave Is My Top IPO Pick Right Now: Over 75% of Its Revenue Is Locked In Through Contracts

CoreWeave's contracted revenue backlog rose 246% year over year to $104 billion at the end of Q2 2026, with management expecting 75% to convert into revenue within four years and more than 75% of its 2027 annualized revenue target above $30 billion already under contract. The company generated $5.1 billion of 2025 revenue, up 168%, and $3.1 billion of adjusted EBITDA, up 154%, while its GAAP net loss widened to $1.17 billion; analysts project 2026 revenue of $12.9 billion and adjusted EBITDA of $7.5 billion. The bullish backlog and AI-infrastructure growth case is offset by a 14.3 debt-to-equity ratio, ongoing GAAP losses expected to reach $2.9 billion in 2026, dilution risk, and potential interest-rate sensitivity.

Analysis

CRWV’s valuation debate is less about demand visibility than the quality and financing of that demand. Its customers are predominantly the same hyperscalers and AI labs with capacity alternatives; as their owned data-center buildouts come online, they can use third-party GPU cloud as a bridge rather than a permanent workload destination. The key diligence item is contract enforceability: take-or-pay provisions, termination rights, customer concentration, and whether commitments are contingent on power delivery or GPU deployment milestones—not headline backlog.

The second-order beneficiary is NVDA: CRWV’s expansion remains a levered channel for incremental accelerator procurement, while NVDA’s equity ownership creates an alignment that may facilitate supply access. However, that relationship also raises circularity risk: if GPU economics weaken, CRWV faces both lower utilization/renewal pricing and declining residual values on the collateral supporting its debt. The equity is therefore a long-duration call on GPU rental pricing financed with fixed obligations, not a conventional software-like recurring-revenue story.

Near term, shares can rebound on evidence that delayed sites are energizing and revenue converts without another capital raise. Over 1-3 months, quarterly deployed-GPU growth, utilization, capex per incremental revenue dollar, and interest expense are more important than backlog additions. Over 6-18 months, the decisive question is whether EBITDA converts to free cash flow after maintenance/replacement GPU capex; a widening GAAP loss alone is not dispositive, but persistent negative post-capex cash flow would justify multiple compression.

Consensus may be too willing to annualize adjusted EBITDA while ignoring refinancing risk and dilution. Conversely, a short solely on leverage is vulnerable if rates ease and deployment catches up, because contracted capacity can create a sharp operating-leverage inflection. The thesis turns constructive only if CRWV demonstrates stable utilization, no material customer cancellation, and declining net-debt-to-EBITDA despite continued buildout.

Key Decisions for Investors

• Maintain CRWV as a tactical watchlist long rather than a core position; initiate only after the next earnings release confirms site-delivery milestones and no downward revision to 2027 revenue conversion. Use a 5-7% portfolio-risk-budget position with a stop/reassessment on a renewed guidance cut or evidence of customer deferrals.
• Prefer a 6-12 month pair of long NVDA / short CRWV for AI-infrastructure exposure: NVDA captures hardware demand with materially less project-finance and utilization risk, while CRWV is exposed to leverage, dilution, and cloud-pricing compression. Cover the CRWV leg if quarterly utilization and post-capex cash flow improve simultaneously for two quarters.
• For upside participation without open-ended equity risk, consider CRWV call spreads 6-12 months out only after earnings, financed by selling a higher strike; avoid naked calls ahead of capex, debt, or guidance disclosures. The trade requires implied volatility to price a lower move than the post-IPO earnings range.
• Monitor MSFT, META, and AMZN capital-expenditure guidance and owned-capacity commissioning dates as leading indicators. Accelerating internal capacity is negative for CRWV renewal pricing over 6-18 months; continued third-party capacity shortages would invalidate the short side of the pair.
• Set a balance-sheet alert: any equity issuance, materially higher funding cost, or net-debt-to-adjusted-EBITDA deterioration despite backlog conversion should trigger a reduction or short bias. A demonstrated decline in leverage alongside free-cash-flow improvement is the clearest falsifier of the bearish credit thesis.