Down 21% This Year: Is Oracle the Most Undervalued AI Stock on the Market?

Oracle raised fiscal-year capex to $50 billion as it pivots to become an AI data-center hyperscaler, funding the push with $25B of debt and $25B of equity/convertible issuance. The investment surge has driven free cash flow to a roughly $44 billion deficit through the first three quarters of fiscal 2026 and pushed long-term debt to about $148 billion (nearly +50% vs end-2025), even as management guides revenue of $67B for fiscal 2026 and $90B for fiscal 2027. Backlog jumped 325% YoY to $553 billion, driven in part by OpenAI work; first-mover and founder-led execution are positives, but elevated leverage and negative FCF make the stock a high-risk, high-reward proposition.

Analysis

An incumbent refocusing its business toward capital-intensive AI infrastructure changes the competitive map beyond GPUs — it reweights bargaining power across real-estate landlords, power utilities, EPC contractors (chillers, PDUs), and specialist integrators. That creates a multi-year procurement cycle where hardware suppliers can extract premium pricing in tight windows, but operating leverage will only materialize once utilization exceeds high fixed-cost break-evens; mismatches here create idiosyncratic cashflow and credit volatility.

Regulatory and sovereignty deals introduce non-linear stickiness: custom on-prem installs and local-data mandates raise customer switching costs, but they also concentrate counterparty risk (large multi-year dependencies on a few government or enterprise customers). Key near-term catalysts that will move markets are milestone deliveries, utility permitting/power capacity notices, and GPU supply shocks; each can swing perceived creditworthiness inside 3-12 months.

The consensus fails to price two asymmetric outcomes. Upside is optionality captured if contracts morph into annuity-like managed services with high gross margins and low churn; downside is a classic asset-heavy mis-execution where scale chase leads to negative free cash flow for several years and true de-rating of equity but material gains for GPU/accelerator suppliers. This bifurcation argues for trading instruments that isolate compute-supply exposure from capital-structure/credit exposure rather than a straight long-or-short equity bet.

Key Decisions for Investors

• Pair trade (6–12 months): Long NVDA via a 3–6 month call spread (buy ATM call / sell 20–30% OTM call) sized to 1–2% portfolio risk; short ORCL via 6–12 month put (defined-risk) or modest equity short sized to 1% risk. Rationale: capture GPU pricing upside while hedging capital-structure/cashflow execution risk. Target 2:1 reward:risk if NVDA > +20% and ORCL falls >10%; stop-loss at 50% of premium paid.
• Credit hedge (12–36 months): Buy protection on senior ORCL bonds / CDS or purchase out-of-the-money 1–2 year ORCL credit default swaps sized to expected exposure from equity positions. Rationale: crystallizes downside if project execution strains liquidity. Cost = insurance premium; payoff asymmetric if credit spreads widen materially above current levels.
• Long NFLX (9–18 months): Buy NFLX stock or a 12–18 month call spread (moderate size). Rationale: second-order beneficiary — AI-driven production/personalization can compress content unit costs and lift incremental margins; target +30–50% upside if realized. Risk management: trim at 20% drawdown or if engagement metrics deteriorate materially.
• Optionality on INTC (18–36 months): Buy long-dated INTC LEAPS or deep ITM call spreads as a convex recovery punt (small allocation). Rationale: if competitive CPU/accelerator cadence improves, incumbents supplying custom silicon for hyperscale deployments rerate. Max loss = premium paid; target 3x upside on execution surprise.