SpaceX vs Amazon: Which Is the Better Stock to Own Over the Next 10 Years

SpaceX has a substantial space-infrastructure lead, with more than 11,000 operational Starlink satellites versus roughly 400 for Amazon, while Amazon plans commercial LEO broadband service by year-end. SpaceX reported $7.8B in Q2 2026 revenue and projects $100B of ARR by year-end, but its expansion requires heavy capex and profitability/free-cash-flow timing remains uncertain. The analysis favors Amazon as the more durable long-term compounder, citing $200B of quarterly revenue, long-duration AI cloud contracts and improving economics from proprietary chips, while characterizing SpaceX as a higher-risk growth investment.

Analysis

The relevant valuation divergence is durability of returns on invested capital, not relative AI exposure. AMZN can monetize compute through a broad software, database and enterprise-sales stack; incremental adoption of Trainium/Inferentia would improve gross margin and reduce NVDA supplier concentration. SPCX's apparent near-term compute economics should be discounted as scarcity rent: as hyperscaler capacity and power availability normalize over the next 12-24 months, short-duration pricing and utilization are likely to mean-revert before its fixed infrastructure cost base is fully absorbed.

SPCX's satellite advantage creates a potentially credible distribution moat, but its market multiple will be governed by launch cadence, spectrum rights, terminal costs, and churn rather than constellation size. The key 1-3 month catalyst is independently verifiable commercial traction—paid enterprise backlog, net adds, ARPU, and capex per incremental subscriber—not management ARR targets. A miss on any of these metrics would be especially damaging given the likely valuation embeds both broadband scale and an unproven orbital-compute option.

The overlooked second-order effect is that AMZN's satellite effort has strategic value even if it is financially subscale: it can defend AWS workloads at edge locations and improve enterprise bundle retention. T gains a low-capital enterprise-connectivity option, but the earnings impact is unlikely to matter until hardware availability, wholesale pricing, and customer activation are disclosed. GSAT/AAPL implications remain non-actionable without acquisition consideration, spectrum integration terms, and regulatory approvals.

Consensus may underprice AMZN's capex/depreciation drag over the next several quarters while overpricing SPCX's scarcity economics. That argues for a relative, not outright, expression: AMZN's multiple can compress if AI revenue conversion lags capex, but SPCX carries greater downside from execution slippage and a less proven free-cash-flow bridge.

Key Decisions for Investors

• Initiate a 6-12 month pair: long AMZN / short SPCX, dollar-neutral, after confirming SPCX liquidity and borrow. Target 15-20% relative outperformance; stop if SPCX reports two consecutive quarters of accelerating paid subscriber growth with improving capex intensity, or if AMZN AWS growth decelerates materially despite rising AI capacity.
• Add AMZN only on post-results weakness tied to capex or depreciation commentary, rather than chase headline AI enthusiasm. Underwrite upside from AWS margin stabilization and custom-silicon mix over 12-18 months; invalidate if AWS backlog conversion and operating-margin guidance both deteriorate.
• Treat NVDA as a read-through watch, not a direct trade: SPCX GPU procurement is unlikely to move consolidated NVDA estimates absent disclosed multi-year capacity commitments. A disclosed material order or revised hyperscaler capex guidance would be the trigger to reassess.
• Do not underwrite T or GSAT earnings upside from satellite connectivity yet. Set alerts for commercial pricing, activated enterprise endpoints, terminal subsidies, and regulatory/spectrum milestones; without those data, the revenue share and margin contribution cannot be estimated.