CoreWeave's 2026 CapEx Spend to Double: High Risk or High Reward?

CoreWeave is accelerating AI-driven infrastructure investment, guiding 2025 capex of $12–$14 billion (revised down from $20–$23B earlier) but warning 2026 capex will be in excess of double 2025 as delayed powered-shell capacity shifts spend into early 2026. The company exited Q3 with a $55.6 billion revenue backlog and 2.9 GW contracted power (with >1 GW expected online in 12–24 months and ~1 GW unsold), but rising interest expense (Q3 interest $311M vs. $104M YoY; 2025 interest guide $1.21–1.25B) and near-term margin pressure from large fourth-quarter deployments raise financing and execution risk amid intense competition from Amazon and Nebius.

Analysis

Market structure: Accelerated capex from CRWV, NBIS and hyperscalers (AMZN, MSFT) implies a near-term build-out arms race where winners are scale players with diversified demand (AMZN, MSFT) and GPU supply control (NVDA suppliers), while smaller pure-plays without locked power/land will be price-takers. CoreWeave’s $55.6B backlog and 2.9GW contracted capacity signal strong demand, but aggregate supply additions (AMZN +3.8GW YTD, Nebius $5B plan) point to potential margin pressure and downward pressure on spot GPU rental rates 6–18 months out.

Risk assessment: Immediate risk (days-weeks) is earnings/guide shock as Q4 capex recognition shifts to Q1 2026 and markets reprice leverage; short-term (3–12 months) risk is interest-rate driven spread widening — CRWV interest expense guided $1.21–1.25B for 2025 — raising default/coverage concerns if revenue timing slips. Long-term (12–36 months) outcomes bifurcate: execution success yields outsized growth; failure or an enterprise AI demand slowdown creates high tail-risk loss given >8x P/B and heavy leverage. Hidden dependency: GPU supply (NVIDIA cadence), power permitting and third-party financing capacity; catalysts include NVIDIA supply cadence, hyperscaler pricing actions, and Q1 2026 reporting.

Trade implications: Tactical trades should favor high-quality hyperscalers and select infrastructure suppliers over leveraged pure-plays. Preferably: size long AMZN or MSFT (2–4% portfolio each) for 6–18 months to capture secular AI share gains; consider a relative-value pair long MSFT / short CRWV (weight 1:0.5) to express margin/financing divergence. Use options to express skew: buy CRWV 3–6 month put spreads (cap loss to 8–12%) and sell covered calls on AMZN/MSFT to finance premium. For credit, buy 1-year CDS or underweight CRWV paper if available; re-evaluate after Q1 2026 capex recognition.

Key Decisions for Investors

• Establish a tactical 2–4% long position in AMZN for 6–18 months to play scale, diversified revenue and ability to outspend peers; finance by selling 1–3 month covered calls to reduce cost basis (strike ~3–5% OTM).
• Initiate a relative-value pair: long 1%–2% MSFT and short 1% CRWV (size short at 50–75% of MSFT notional) to exploit CRWV’s higher leverage/interest expense; tighten stop-loss on short leg if CRWV falls >25% in 60 days or MSFT falls >12%.
• Buy CRWV 3–6 month put spread (e.g., buy 1–2% notional downside protection, cap loss to 8–12%) to hedge downside from margin/interest risk ahead of Q1 2026 reporting; target breakeven if CRWV drops 20–35%.
• Reduce direct exposure to small-cap pure-play AI infra equities by 30–50% and redeploy to large-cap cloud (AMZN, MSFT) and GPU supply chain (NVDA suppliers) over the next 90 days; reassess after Q1 2026 capex recognition and NVIDIA shipment updates.
• If active in credit, buy 1-year CDS protection on CRWV or underweight its corporate bond allocation if spread >600bps over Treasuries, and revisit after liquidity/event windows around Q1 2026 results.