Amazon Just Revealed This 1 Business Is Worth More Than 82% of Fortune 500 Companies (Hint: Not Cloud Computing)

Amazon's AI chip business has an annual revenue run rate >$20B and is growing at triple-digit % YoY; CEO Andy Jassy said it could be ~$50B as a standalone business (~7% of 2025 revenue of $717B). Amazon plans to spend $200B in capex in 2026 with the majority already tied to customer commitments to be monetized in 2027-28. Management emphasizes long-term prioritization of AI opportunity; the disclosures could materially re-rate investor expectations given the sizable revenue and growth implications and the company's valuation at ~33x earnings versus a three-year average of 67x.

Analysis

Amazon’s shift from buyer to seller of bespoke AI racks creates both a margin arbitrage and a channel conflict that few models price in today. By internalizing silicon design and system integration, Amazon can compress infrastructure unit economics for AWS while capturing system-level value that historically accrued to GPU vendors and OEMs; that could translate into a structural re-rating of its services margins over a 12–36 month window. Third-party demand for integrated racks faces a behavioral hurdle — enterprise customers trade off price for vendor neutrality and software ecosystem lock-in, so material third-party rack revenue will likely follow not just product readiness but credible third-party references.

Supply-chain dynamics will be the next battleground: foundry capacity, advanced packaging (substrates/interposers), and high-density power delivery become chokepoints that create optionality for firms that secure slots early. That gives Amazon leverage but also raises a capital-intensity and execution profile that increases variance — a slip in supply (or a sudden re-prioritization by TSMC) can flip multi-year forward economics quickly. Regulators pose a non-trivial tail risk as vertically integrated cloud+hardware offerings invite questions about preferential treatment of hosted customers, a litigation/regulatory timeline measured in quarters-to-years.

The common consensus underweights two second-order outcomes: (1) a rapid AWS margin inflection will compress competitive GPU multiples even if absolute GPU demand grows, and (2) successful sales of racks to third parties would create an OEM-like competitor that forces incumbents to accelerate software portability and pricing flexibility. Both outcomes create tradable dispersion across compute suppliers and chip incumbents; the market is set up for idiosyncratic moves rather than a single macro-driven rerating in the next 6–24 months.

Key Decisions for Investors

• Long AMZN (12–24 months): Build a 3–5% position on pullbacks or ahead of AWS earnings; thesis is >20% upside if AWS gross margins expand materially from verticalized silicon, with a 20% downside if execution or regulatory issues emerge. Use a staggered buy on 5–10% intraday retracements to manage entry.
• Pair trade — Long AMZN / Short NVDA (9–18 months): Express capture of system-level value by owning AMZN while shorting NVDA equity or buying NVDA 9–12 month OTM puts. Target a 2:1 reward:risk (e.g., 30–40% upside on AMZN vs 15–20% downside risk) — rationale: Amazon can take share in select inference/scale workloads even as NVIDIA retains high-margin niches; hedge NVDA’s software moat as the chief risk.
• Short INTC (6–18 months): Small-to-medium position via buying puts or outright short, targeting ~40–60% downside in a worst-case share-loss scenario if Intel fails to defend data-center CPU share against hyperscaler verticalization. Stop-loss tight: cover on any clear turnaround signs (major foundry win or design-win announcements).
• Tactical hedge / alpha — Long AAPL (12 months) as defensive hedge: Maintain a 1–2% position to offset macro/consumer cyclicity while AI capital reallocation plays out; expect low correlation to hyperscaler-capex-led moves and provide ballast if AI capex disappoints.