CoreWeave vs. Applied Digital: Which AI Data Center Stock Will Make You Richer?

CoreWeave exited 2025 with a $66.8 billion contracted backlog and revenue that surged 168% to >$5.1B, with 850 MW of active power now and guidance to exceed 1.7 GW by end-2026 and >3.1 GW contracted (mostly online by 2027); analysts forecast revenues of $12.4B in 2026 (+142.5%) and $23.1B in 2027 (+86%), and the stock trades at ~8.5x sales. Applied Digital has contracted ~600 MW (~$16B prospective lease revenue) including a 400 MW campus leased to CoreWeave (~$11B over 15 years) and reported fiscal Q2 2026 revenue up 250% to $126.6M with a $31.2M net loss; analysts expect revenues of $346M in fiscal 2026 (+61%) and $535.6M in 2027, and it trades at ~27.3x sales. Overall, CoreWeave appears better positioned for higher returns given faster growth, stronger backlog and a lower sales multiple, but material execution and financing risks (heavy 2026 capex of $30–35B and near-term margin/debt pressure) remain for both companies.

Analysis

Infrastructure owners, hyperscalers, and the GPU supply chain are diverging winners. Land, power contractors, and electrical-equipment OEMs gain durable, predictable cashflows from long leases, while GPU vendors and specialized colo operators capture the upside of bursty compute demand — creating a two-speed market where compute-as-a-service margins compete with build-and-hold IRRs. Expect pressure on mid-tier cloud incumbents that neither own cheap capital nor deep OEM relationships, as they will be squeezed between hyperscalers’ scale and vertically integrated AI-native operators.

The primary macro/operational risks are financing cadence and supply-cycle normalization. Heavy upfront capital requirements make access to debt and structured lease financing the gating factor over the next few years; a meaningful repricing of credit or a moderation in enterprise AI IT spend would disproportionately hurt fast-scaling operators relative to landlords with investment-grade tenants. Separately, an acceleration in GPU supply or a move to more compute-efficient model architectures could compress utilization and pricing, reversing current pricing power within a single hardware cycle.

Actionable relative-value opportunities arise from disconnects between growth optionality and multiple compression. Positions that long operationally scalable, contract-backed cashflows while shorting execution-dependent rollouts capture convexity if projects slip or financing tightens. Hedged option structures (buying calls as convex exposure and selling nearer-term premium) buy time for execution without committing large capital while capping downside from a sudden financing shock.

The consensus underweights timing risk and overweights headline demand. Market narratives focus on secular AI demand but underplay how grid permitting, interconnection queues, and GPU replacement cycles create discrete cliff risks in 6–24 month windows. That mispricing favors trades that isolate execution versus secular demand — capture the secular with leases and hedge the execution with short-duration puts on the build-first players.

Key Decisions for Investors

• Long CRWV equity (2% NAV) with 12–24 month horizon; target ~2x on successful execution of capacity ramp and financing windows; hard stop-loss at 30% to limit exposure to a financing or demand shock.
• Buy 9–15 month CRWV calls (out-of-the-money) sized as 0.5% NAV to capture asymmetric upside while risking only premium; finance by selling near-term covered calls on a liquid tech name (e.g., NVDA) if option spreads are favorable.
• Pair trade: long CRWV / short APLD equal-dollar (1% NAV each leg) over 12 months — long the operator with variable pricing, short the high-multiple build-to-lease exposure; unwind if APLD funds projects with non-recourse lease financing or if CRWV misses capacity milestones.
• Protective hedge: buy 6–12 month APLD puts (or short APLD) sized to cover tail risk from project slippage when long CRWV; rationale is to guard against correlated execution/credit squeezes across the ecosystem.