Amazon’s Andy Jassy Just Made a Startling Prediction.

AWS has reached a $169 billion annual revenue run rate, while its AI revenue run rate has surpassed $25 billion roughly three years into the generative-AI boom. Amazon raised planned 2026 capital expenditure to $220 billion from $200 billion as it builds infrastructure for demand it says it cannot fully meet this year or next. CEO Andy Jassy said AWS could ultimately generate at least $400 billion—and potentially $1 trillion—in annual revenue, supported by demand across both AI and non-AI cloud services.

Analysis

The investable issue is not whether AWS can sustain high growth, but whether incremental infrastructure dollars earn returns above Amazon's cost of capital before capacity is commoditized. A higher capex envelope raises near-term depreciation, power-contract, and networking expense faster than reported revenue; this can depress consolidated free cash flow and delay multiple expansion even if AWS bookings remain strong. The key quarterly proof points are AWS growth reacceleration, segment operating-margin stability, and management disclosure that backlog/conversion—not merely customer experimentation—is absorbing new capacity.

Second-order beneficiaries are the scarce inputs rather than the hyperscaler: NVDA retains pricing power while leading accelerator supply is constrained, but the more durable bottlenecks may be power generation, grid equipment, cooling, and optical networking. AMZN's internal Trainium/Inferentia strategy is strategically negative for NVDA only if third-party customers adopt it at scale; until utilization and external-chip revenue are disclosed, it is primarily a margin-defense option rather than a confirmed share-loss signal for NVDA.

Consensus may be underpricing the execution risk embedded in an aspirational long-run revenue endpoint. Enterprise AI spending can migrate from training to inference with lower compute intensity per unit of customer value, while Microsoft Azure, Google Cloud, and private-cloud deployments can pressure AWS pricing as capacity arrives. Over the next 1-3 months, capex guidance and free-cash-flow revisions should dominate the stock; over 6-18 months, the relevant question is whether AWS margin expands despite depreciation and power costs. Falsify the constructive AMZN view if AWS growth fails to improve for two consecutive quarters while capex rises, or if AWS operating margin contracts by more than 200bp without a corresponding backlog acceleration.

Key Decisions for Investors

• Maintain a measured long AMZN versus short QQQ through the next two earnings cycles, sized for a 8-12% relative-risk budget. The setup works if AWS growth and retail efficiency offset capex-driven FCF concerns; exit the relative long on two quarters of decelerating AWS growth or a material reduction in capital-return capacity.
• Do not chase AMZN solely on the long-duration cloud narrative before earnings. Add only following evidence of stable AWS margin and improved revenue growth; absent that evidence, the likely near-term outcome is multiple compression from elevated depreciation rather than a fundamental break.
• Retain NVDA exposure for the next 6-12 months but hedge concentration with a long AMZN / short NVDA relative-value overlay only if AWS reports meaningful external adoption of Trainium/Inferentia and NVIDIA GPU procurement growth slows. The missing confirmation is customer-level chip mix and utilization; without it, a structural NVDA short is premature.
• Monitor power and data-center supply-chain proxies—ETN, VRT, PWR and GLW—for earnings revisions rather than headline-driven entries. Initiate selectively after backlog or margin guidance validates that hyperscaler capex is translating into equipment demand; the principal risk is project deferral from grid interconnection delays and power-price inflation.