Better Artificial Intelligence (AI) Stock: CoreWeave vs. Nebius

CoreWeave reported Q3 revenue of $1.36 billion, up 134% year‑over‑year, but trimmed full‑year guidance to a $5.1 billion midpoint from $5.25 billion due to a temporary third‑party data‑center delivery delay; it sits on a roughly $56 billion revenue backlog, has 590 MW active capacity and 2.9 GW contracted after adding 120 MW in Q3. Smaller rival Nebius posted Q3 revenue of $146 million, up 355% YoY, and is backed by multibillion dollar five‑year contracts (Microsoft $17.4–$19.4B, Meta $3B) yielding a >$20 billion backlog and upgraded capacity targets (2.5 GW contracted by 2026; 800–1,000 MW active by end‑next year). From a valuation and diversification standpoint the piece favors CoreWeave as the better buy given a cheaper price‑to‑sales multiple versus the tech sector average (8.4) and a broader customer base, while Nebius carries faster near‑term growth but higher customer concentration risk.

Analysis

Market structure: Winners are neocloud operators (CRWV, NBIS), GPU suppliers (NVDA exposure implicit) and hyperscalers (MSFT, META) that secure on-demand capacity; losers are legacy colo/managed hosting and regional markets with constrained power. CoreWeave's scale (contracted 2.9 GW, ~$56B backlog) gives it clearer pricing leverage vs Nebius (~$20B backlog, concentrated MSFT/META deals) but both signal demand > near-term supply for AI-grade power and GPUs, supporting premium valuations until capacity catches up.

Risk assessment: Tail risks include export controls on accelerators, a material renegotiation/cancellation by MSFT/META (>=20% of NBIS backlog) and funding stress if yields rise >100bp (increasing WACC and dilution risk). Time horizons: immediate (days) — headline-driven volatility on guidance; short-term (weeks/months) — capacity bring-online confirmations and GPU shipments; long-term (quarters/years) — backlog conversion and margin normalization. Hidden dependencies: third-party developer timelines, power hookups, NVIDIA supply cadence and customers’ model-optimization reducing compute per $.

Trade implications: Tactical trade — initiate a 9–12 month long CRWV position (1.5–2% of portfolio) via call spreads or LEAP calls to capture backlog conversion while limiting downside; hedge with a 0.75–1% short NBIS or buy 3–6 month NBIS put spreads (30–50% OTM) to protect vs customer-concentration renegotiation. Consider a pair trade long CRWV / short NBIS sized ~1:0.6 by notional to reflect backlog ratios (~56B:20B). Rotate 1–2% from legacy colo REITs into scaled neoclouds while keeping overall sector exposure capped if 10‑yr yields >4%.

Contrarian angles: Consensus overweights headline backlog sizes and underweights convertibility constraints — a 10–30% slippage in backlog realization is plausible if delivery delays repeat. CoreWeave’s >60% post-IPO drawdown looks potentially oversold on execution clarity and diversified customer base; Nebius may be overbought (231% YTD) with asymmetric downside if MSFT/META push for price or build in-house. Historical parallel: prior colo build cycles saw 30–60% re-rates when funding tightened or hyperscalers internalized capacity; guard for similar mechanics here.

Key Decisions for Investors

• Establish a 1.5–2.0% portfolio long in CRWV with 9–12 month horizon using either 2026 LEAP calls or a 25/50% OTM call spread to cap cost; add if quarterly filings confirm sequential active capacity growth ≥100 MW/qtr or backlog conversion guidance restored.
• Implement a hedge on NBIS: short 0.75–1.0% notional NBIS or buy a 3–6 month put spread 30–50% OTM sized to cover downside risk; trim/cover if NBIS diversifies beyond MSFT/META (third-party customers >30% of revenue) or if active capacity target beats 1.0 GW by >20% next quarter.
• Execute a relative-value pair trade: long CRWV and short NBIS in a 1:0.6 notional ratio (reflecting ~56B vs ~20B backlog) with quarterly rebalancing; unwind the short if NBIS free-float increases or MSFT/META share of backlog falls below 40%.
• Reallocate 1–2% from legacy colo REITs into scaled neocloud exposures (CRWV, selective peers) but reduce net exposure by 50% if 10‑yr UST >4.25% or if GPU supply reports show shipment delays >30 days vs vendor guidance within next 60 days.