Amazon Stock Could Soar Over the Next Few Years -- and It's Coming From an Unlikely Source

Amazon’s AWS business generated 59% of operating profit in Q1 despite only 21% of revenue, underscoring the profit engine behind the stock. Management is planning $200 billion in capex this year to expand data centers, supported by demand from AI workloads and already-lined-up customers. The article also highlights triple-digit growth in Amazon’s custom AI chip business and suggests this could accelerate AWS growth beyond the 28% delivered in Q1.

Analysis

The real setup is not “Amazon vs. commerce,” it’s AWS as a capital-intensive capacity race where the winner can temporarily look like the slowest grower and still compound the most value. The market is underestimating how much of Amazon’s incremental earnings power now comes from utilization of already-funded infrastructure and from shifting workloads onto proprietary silicon, which structurally lowers unit cost and raises switching friction. That creates a second-order effect: every dollar of AI demand routed through Amazon’s stack compounds both cloud share and silicon attach rate, while pressuring higher-cost GPU-only deployments.

The competitive read-through is more important than the headline optimism. If custom chips continue to displace general-purpose compute in inference and selected training workloads, the economic moat moves from “cloud scale” to “full-stack cost engineering,” which is harder for pure-play hyperscalers and chip vendors to defend against. That is modestly negative for NVDA at the margin because the risk is not immediate share loss, but a slower expansion of its addressable mix in workloads where customers become cost-sensitive and optimize for efficiency rather than peak performance.

The risk is execution, not demand. Over the next 1-3 quarters, capex intensity can suppress reported free cash flow and keep multiple expansion capped if investors focus on near-term margins instead of forward utilization. The key reversal would be either a slowdown in AI spend broadly or evidence that custom chips are not materially improving cost/performance versus incumbent alternatives; both would weaken the thesis and shift the market back toward viewing AWS as a lower-growth utility-like asset.

The contrarian angle is that the stock may still not be priced for the operating leverage embedded in AWS’s mix shift. Consensus tends to anchor on retail revenue scale, but the real upside is a higher-quality earnings stream with improving self-reinforcing economics if Trainium/Graviton adoption keeps rising. The move is therefore probably underdone if AI inference demand remains durable and Amazon can monetize it through lower-cost proprietary compute rather than surrendering economics to third-party chip suppliers.

Key Decisions for Investors

• Long AMZN on 6-12 month horizon; prefer adding on any pullback tied to capex/FCF concerns. Risk/reward favors a higher multiple if AWS utilization data stays strong and proprietary chip adoption accelerates.
• Pair trade: long AMZN / short NVDA for 3-6 months as a relative-value expression of AI spend migrating toward lower-cost custom compute. Stop if NVDA reaccelerates on training-led demand or Amazon commentary shows chip adoption stalling.
• Buy AMZN Jan-2027 call spreads to express upside from operating leverage without paying full delta for near-term capex volatility. Best if entry follows a post-earnings digestion period.
• Avoid chasing NVDA on cloud-customization headlines until there is evidence the workload mix is still GPU-dominant; the risk is multiple compression from slower marginal share gains rather than absolute revenue decline.