Amazon's CEO Just Gave Nvidia Investors Great News

Amazon said its AI business is growing by triple-digit percentages and that it continues to order substantial quantities of Nvidia chips, signaling durable demand for Nvidia’s AI hardware. Wall Street is expecting Nvidia fiscal 2027 Q1 sales of $79 billion, up 79% year over year, implying continued acceleration. The article is largely supportive of both Amazon’s AI momentum and Nvidia’s outlook, though it is framed as analyst commentary rather than new hard data.

Analysis

The key read-through is that Amazon’s custom silicon is not a zero-sum substitute for Nvidia; it is a demand-expansion mechanism. As AI workloads become more heterogeneous, the market should expect a bifurcation where price-sensitive inference and internally standardized workloads migrate toward Trainium/Graviton, while frontier training and premium enterprise workloads still pull heavily on Nvidia. That mix supports a higher overall capex envelope across hyperscalers, which is why Nvidia can keep compounding even as AWS pushes its own chips.

Second-order, this is a margin and ecosystem story for AMZN more than a pure chip story. If AWS can reserve Nvidia for the highest-value workloads and shift lower-ROIC compute to in-house silicon, it preserves customer choice while improving its own unit economics over a 12-24 month horizon. The market is likely underestimating how much this widens AWS’s pricing flexibility versus smaller cloud peers that lack proprietary silicon and must pass through more of the AI GPU inflation.

For NVDA, the near-term risk is not demand loss but expectation compression. When a hyperscaler openly validates substantial Nvidia ordering, it becomes harder for the Street to maintain skepticism around the 2026-2027 growth ramp; that raises the bar into the next print and makes any sequencing hiccup more punishable. The real vulnerability is not AWS substitution, but whether enterprise AI monetization lags capex growth, causing investors to question the durability of spend after the next 2-3 quarters.

Contrarian angle: the market may be overfocusing on chip share and underappreciating that the winner is whoever controls workload routing. AMZN can monetize both sides of the stack, while NVDA benefits from the continued need for best-in-class performance; the under-owned idea is that this relationship is structurally cooperative in the medium term. The setup favors continued capex intensity across the ecosystem, but with a higher-quality earnings profile at AMZN than most investors are giving it credit for.

Key Decisions for Investors

• Stay long AMZN vs. XLK for the next 3-6 months: AWS silicon creates a margin uplift path that the market is likely underpricing; use any post-earnings pullback to add, with the main risk being a broader cloud deceleration.
• Maintain a core long NVDA position into the next earnings window, but trim upside exposure via covered calls into strength: the setup is still positive, yet expectations are elevated and a modest miss on guide could drive a 7-10% air pocket.
• Pair trade: long AMZN / short a basket of GPU-dependent smaller cloud enablers over 6-12 months: hyperscaler vertical integration should pressure vendors without custom silicon or scale, while AWS can arbitrage AI demand across both proprietary and merchant chips.
• If seeking cleaner asymmetry, buy NVDA call spreads 1-2 quarters out rather than outright shares: this captures continued capex upside while limiting damage if the market starts discounting slower enterprise monetization.
• Avoid shorting NVDA on this headline alone; the better short is a lagging AI infrastructure name with no proprietary compute moat, where the substitution risk is real and not offset by platform control.