Oracle Backs $50 billion CapEx Following Strong Earnings

Oracle reported Q3 FY2026 adjusted EPS of $1.79 vs $1.70 consensus and revenue of $17.19B vs $16.91B, a modest beat. Remaining performance obligations surged ~4x year-over-year to over $553B, and the company plans to raise $45–50B and spend up to $50B in FY2026 capex to support cloud/AI infrastructure. Oracle and OpenAI canceled a planned Texas AI data center expansion due to financing and changing needs, but management says it has capital to meet demand. Analyst coverage is positive (35 Buys of 45 analysts) with an average price target of $240, implying >60.5% upside as of March 10.

Analysis

Oracle’s push to own a sizable slice of AI infrastructure shifts the demand map away from pure hyperscale outsourcing toward vertically integrated, enterprise-centric buildouts. That reallocation benefits data‑center networking, power/cooling, and colo landlords because AI spend is front‑loaded into capacity and systems integration rather than pure software consumption. Conversely, third‑party cloud providers that monetize on broad multi‑tenant workloads could see pricing leverage erode on the most profitable, latency‑sensitive enterprise AI workloads.

Key tail risks are execution and technology substitution: large hardware programs are lumpy and capital‑intensive, so financing cadence and deployment pace (quarters → single years) will drive supplier revenue timing and margins. A faster move toward model compression, quantization, or specialized inference ASICs would materially reduce GPU intensity and shorten the reorder cycle; conversely, any stall in model training demand (OpenAI‑type partner churn or macro capex tightening) shuts the incremental margin waterfall quickly. Monitor free‑cash flow trends and incremental gross margins on infrastructure contracts as the earliest hard readouts over the next 3–12 months.

The market consensus treats Oracle’s infrastructure push as a mostly risk‑free re‑rating lever; that understates two second‑order effects. First, heavy capex increases the incentive to monetize via long‑dated contracts and colo partnerships, which diffuses upside across suppliers rather than into Oracle’s pure software margins. Second, if Oracle negotiates deeply on GPU/accelerator pricing to control unit economics, it could compress vendor ASPs — a positive for Oracle but a negative for pure‑play GPU names. These dynamics create opportunities for relative‑value trades between infrastructure suppliers, colo landlords, and chip vendors.

Key Decisions for Investors

• Buy ORCL equity on weakness (enter on >5% pullback from today) sized at 1–2% portfolio weight with a 12‑month target of +30–40% if enterprise AI contracts convert at healthy margins; hard stop at -15% to control capital intensity/execution risk.
• Buy ORCL 12–18 month LEAP call (delta ~0.30–0.45) as an asymmetric play—risk limited to premium—sell 25–33% of position on any >50% move to lock profits; use puts (6–9 month 10–15% OTM) as tactical insurance if financing spreads widen.
• Relative‑value pair: long ARISTA (ANET) 6–12 months (benefits from switch/router reorders and higher port density) versus short NVDA 3–6 month call spread to hedge market beta. Thesis: networking equipment will re‑rate if capex allocates more to interconnect and less to incremental GPU volume; target 20–30% relative outperformance, max drawdown capped by notional sizing.
• Tactical long on data‑center REITs (EQIX or DLR) with 6–18 month horizon to capture higher vacancy re‑pricing and longer lease terms from enterprise AI tenants. Size modestly (0.5–1% PF each), take profits on a 20–30% rally; watch for supply additions that would invalidate thesis.