CoreWeave Just Landed a Deal With Perplexity. Why This Neocloud Stock Could Be the Comeback Play of 2026.

CoreWeave's revenue has grown in the triple digits and it claims the title of fastest cloud to reach $5B in annual revenue; the stock initially rose >300% post-IPO but is down ~50% from its high. The company secured a multi-year deal to run Perplexity's inference on Nvidia GB200 NVL72 clusters, and benefits from close ties to Nvidia (article states CoreWeave comprises 27% of Nvidia's stock portfolio). Key risks include rising debt for infrastructure buildout, potential slowdown in AI spending, and geopolitical headwinds (e.g., war in Iran).

Analysis

Core-capacity specialists create a demand bifurcation: model makers and inference-heavy apps outsource to specialized clouds rather than owning boxes, concentrating unit economics around GPU suppliers and a handful of rental platforms. That concentration tightens upstream pricing power for Nvidia-class silicon but simultaneously magnifies platform-level operating leverage — a 5ppt drop in cluster utilization can swing adjusted EBITDA by double-digit percentage points within a single reporting quarter.

Debt-fueled expansion converts seasonal or lumpy GPU demand into existential equity risk: if utilization falls below ~70% for multiple quarters, liquidity windows compress and management faces either dilutive raises or deferred deployment that contracts long-run margins. Export controls, regional power constraints, and hyperscaler countermoves (price-locks/subsidies) are three distinct catalysts that can reverberate through revenue and margin curves with 3–12 month lead times.

Consensus underprices the commoditization channel: as more providers standardize on the same high-end Nvidia stack, a spot market for inference capacity will emerge faster than most models assume, compressing gross margins by 400–800bps over 12–24 months unless platforms convert customers into sticky, value-add services (ops tooling, fine-tuning pipelines, data residency). That sets up a binary outcome — durable software-driven monetization and outsized returns, or a margin race that favors the deepest-pocketed hyperscalers — and creates clear tradeable asymmetries for the next 6–18 months.

Key Decisions for Investors

• Long NVDA via 3-month call spread (buy 1 ATM, sell 1 20% OTM) sized to 2% NAV: entry before next earnings; rationale — captures upside from continued GPU ASP/take-rate strength while capping implied vol cost; target 2.5x premium, max loss = premium paid.
• Selective long CoreWeave (CRWV) equity, 1–2% NAV, only on >25% pullback from recent levels; set hard stop at -30% and target +100% over 12–18 months. Hedge systemic GPU-cycle reversal by buying NVDA 3-month puts equal to 25% notional of the CRWV position (reduces tail risk at modest cost).
• Tactical short/put on INTC (buy 9–12 month puts, 1% NAV): thesis — secular CPU share pressure in AI datacenter and slower acceleration in accelerator roadmap vs incumbents; risk/reward asymmetric if Intel disappoints on tapeouts, expect 1.5–3x payoff if competitive momentum stalls.
• If worried about industry commoditization, initiate a market-neutral pair: long CRWV (1% NAV) / short broad cloud heavyweight (size to dollar-neutral) for 6–12 months — isolates platform execution vs hyperscaler pricing response. Exit or rebalance if utilization trends improve sustainably over two consecutive quarters.