1 Artificial Intelligence (AI) Stock That Could Surprise Investors in 2026

Shares of CoreWeave are down 55% from their 2025 highs despite revenue surging from $229M in 2023 to $5.1B in 2025 and management forecasting at least a 235% jump in 2026. The company reported a TTM net loss of $1.2B, huge infrastructure spending (Q4 capex $8.2B; planned 2026 capex ≥ $30B) and a contracted revenue backlog of $66.8B, while Microsoft comprised 67% of 2025 revenue — presenting both demand validation and concentration risk.

Analysis

CoreWeave’s playbook — owning the stack from racks to scheduler — creates a two-way supply shock: it can corner near-term GPU inventory for large customers while also driving down the market price for on-demand GPU hours once utilization steps up. That dynamic gives CoreWeave pricing leverage with hyperscalers and a structural advantage over pure-software players, but it also turns the company into a capital-intensity arbiter of the AI compute market; whoever controls capacity will set spot rates and utilization pacing for years.

The key near-term hinge is utilization velocity, not headline growth. If utilization rises faster than fresh capacity is brought online, unit economics will improve sharply and cash burn can flip to meaningful free cash flow within 12–24 months. Conversely, if GPU supply outpaces demand (due to easing geopolitical constraints, slower model iteration, or hyperscalers internalizing capacity), then utilization falls and the asset base becomes a leverage amplifier — forcing either deep pricing cuts or equity/capital raises with meaningful dilution.

Second-order winners include GPU and memory suppliers who can monetize long-term purchase commitments and offer vendor financing; Nvidia stands to gain pricing power and financing revenue, while memory vendors win sustained BOM demand. Second-order losers are smaller GPU-cloud resellers and on-prem hardware refresh cycles, which face margin compression and accelerated consolidation as scale advantages and custom orchestration stacks (scheduler + data fabric) become entry barriers.

This is fundamentally a capital-allocation story masquerading as an AI revenue story: the market is pricing optionality on utilization improvement versus capital structure risk. Watch the cadence of contract conversions, customer mix shifts away from single large buyers, and any moves toward non-dilutive financing (sale-leasebacks, vendor-backed financing) — each will materially reprice risk premia over quarters, not days.

Key Decisions for Investors

• Small, asymmetric long on CRWV via defined-risk options: Buy Jan-2027 LEAPS call spread on CRWV (buy nearer-ATM, sell 2x-3x OTM) sized to 0.5–1.0% of portfolio. Rationale: captures multi-quarter utilization inflection with capped downside (max loss = premium). Target return: 2–4x if utilization and pricing normalize; downside limited to premium paid.
• Hedge the asymmetric long with NVDA tail protection: Purchase 3–6 month NVDA puts (moderate notional sized ~25–50% of CRWV position) to protect against a systemic AI demand shock that would depress both GPU hours and GPU-equipment vendors. This converts an idiosyncratic bet into a market-aware asymmetric stance.
• Relative-value overweight memory exposure (MU) versus Intel (INTC): Add MU (size 1–2% portfolio) on expectation of sustained BOM demand and pricing leverage from long-term commitments; keep INTC flat/underweight as its server traction remains uncertain. Time horizon 6–18 months; look to trim MU into any >25% outperformance versus peers.
• Event hedge / short if red flags appear: If management pursues large equity raises, materially extends customer concentration, or if public GPU supply disruptions reverse (leading to price collapse), initiate a short CRWV position with stop-loss at predetermined liquidity levels or buy puts to size tail-risk protection. Timeframe: reactive within days of the catalyst.