Applied Digital vs. CoreWeave: Better Stock to Own in 2026?

CoreWeave and Applied Digital are positioned for material growth in 2026 as AI cloud demand outstrips capacity: CoreWeave rents GPUs and related services, plans to more than double capex next year, has reportedly turned away an estimated $40–50 billion from Microsoft, and Citigroup checks indicate 2026 capacity is sold out with bookings into early 2027. Applied Digital supplies specialized high-power data-center shells and colocation, holds an $11 billion services contract with CoreWeave, delivered the first 100 MW of a planned 400 MW Polaris Forge 1 deployment, and signed a $5 billion lease for Polaris 2 while partnering with Macquarie on campus expansion. The analyst view favors Applied Digital due to the scarcity and value of cheap, stable power, though both firms are described as set for “hypergrowth.”

Analysis

Market structure: CoreWeave (CRWV) and Applied Digital (APLD) are capturing the two scarcities in AI scale — high-end GPU cycles and affordable, expandable power/real-estate — which gives both pricing power in 2026 as capacity is reportedly sold out. Winners include neocloud operators, GPU suppliers (NVDA) and data‑center infrastructure providers; losers are undifferentiated colo and smaller cloud players who cannot secure GPUs or power and may face margin compression. Expect utilization-driven revenue acceleration (high‑teens to +50% YoY scenarios) and upward pressure on regional power prices and copper/transformer demand where rapid campus builds occur.

Risk assessment: Key tail risks are Nvidia allocation shifts (NVDA reprioritizing hyperscalers), construction/permitting delays at APLD campuses, and a concentrated counterparty risk (CRWV/APLD revenue tied to a few large customers including an $11bn CRWV‑APLD relationship). In the near term (days–weeks) expect headline-driven volatility around capacity announcements and NVDA product/guide events; medium term (3–12 months) execution and financing (Macquarie slots, project debt) dominate; long term (2–5 years) grid constraints/regulatory limits and technology shifts (more efficient chips) could compress returns. Hidden dependencies: both companies’ growth is effectively levered to NVDA GPU supply and local grid upgrades — monitor GPU allocation data and interconnection queue progress as second‑order controls on revenue timing.

Trade implications: Favor APLD for asymmetric, infrastructure‑moat exposure and CRWV only as a growth play with execution risk — implement size and risk controls. Use pair trades to isolate execution risk (long APLD, short CRWV notional 50%) and use NVDA exposure via LEAPS to play sustained GPU tightness instead of relying on single‑name CRWV upside. Cross‑asset: expect elevated implied volatility in both names; prefer buying time (LEAPS or 9–15 month calls) to shorting weekly options; bonds: increased issuance by both could pressure high‑yield spreads in the space if capex accelerates.

Contrarian angles: Consensus assumes perpetual capacity shortage; that underprices the scenario where Nvidia scales supply, hyperscalers self‑deploy, or regulators slow new builds — any of which would reprice growth expectations dramatically. CRWV may be more vulnerable than APLD because the latter owns power and land (stickier cash flow), so CRWV’s valuation could be more sensitive to GPU allocation changes. Historically, infrastructure owners (landlords) outlast boom‑era operators; small execution slippages can flip high‑growth narratives quickly, creating 30–70% downside windows in stressed selloffs.

Key Decisions for Investors

• Establish a 2–3% long position in Applied Digital (APLD) over the next 4–8 weeks, scale in on any >8% pullback; target +40–60% by end‑2026, set an initial hard stop at −20% and reassess if APLD fails to add ≥100MW of contracted power within 9 months.
• Implement a pair trade: long APLD (2% portfolio) and short CoreWeave (CRWV) at 1% notional (50% hedge ratio) to capture APLD’s power moat versus CRWV’s execution/GPU allocation risk; rebalance if CRWV misses capacity guidance or if APLD signs additional ≥200MW leases.
• Buy NVDA directional convexity with limited downside: purchase Jan 2027 LEAPS ~20% OTM (or nearest available) equal to 1% portfolio exposure to capture sustained GPU tightness; hedge 20–30% of notional by selling 3–6 month calls if implied vol spikes above 80% to offset premium.
• If third‑party datapoints confirm >90% sold‑out 2026 capacity for CoreWeave or APLD announces an incremental ≥100MW power/lease deal, add up to +1–2% incremental long APLD exposure within 30 days; conversely reduce CRWV exposure by 50% if NVDA signals meaningful reallocation to hyperscalers or if APLD/CRWV project completion dates slip >90 days.