
Last 4 earnings calls · through Q2 FY2026
The bull case is that demand for cloud and AI capacity runs ahead of what Amazon can build, so revenue grows as fast as data centers and chips come online.
Amazon designs its own Trainium and Graviton chips, which lowers what it costs to serve AI and lets it undercut rivals that buy processors from others.
Backlog keeps rising and management says it is not concentrated in one or two customers, which gives more visibility into future cloud revenue than Amazon has had before.
Operating margins are at records because advertising and cloud now carry the company, and management says AI margins follow the core cloud pattern.
The bear case is that capital spending keeps outrunning revenue, so free cash flow is negative and profit depends on capacity that takes years to fill.
AI demand comes mostly from a handful of large labs and a few runaway applications, so the cloud backlog depends on a small group continuing to spend.
Amazon sets its own growth rate through construction schedules and memory chip supply, so revenue can lag even while customers are waiting for capacity.
Open antitrust and consumer-protection cases, plus reliance on China-based sellers and suppliers, can force changes or raise costs outside management's control.
In short: Backlog and AWS growth are rising fast, but capacity limits and dependence on a few large AI buyers remain.
Cloud is now the profit center of the company, so its acceleration lifts both revenue growth and operating income even as the retail business grows more slowly.
A few quarters ago management pointed to a $200 billion backlog; now it reports a $496 billion backlog growing triple digits and says it is not concentrated in one or two customers.
Management says capacity will still fall short of demand in 2026 and 2027, so revenue timing depends on construction schedules and memory chip supply rather than on customer orders.
Jassy describes AI demand as barbelled, with the labs buying enormous amounts of compute; if those few buyers slow down, AWS growth would feel it quickly.
In short: Operating margins are at records, but free cash flow is negative while capital spending outruns revenue.
Higher-margin AWS and advertising now carry the company, so total profit grows faster than revenue even while retail prices stay low.
Earlier the CFO framed AI as a margin drag through depreciation on new data centers; now the CEO says AI margins are following the core cloud trajectory and slightly ahead of it.
Trailing twelve-month free cash flow was -$11.6 billion, so the build-out is being funded from cash reserves and new borrowing rather than from internally generated cash.
The 2026 cash capital spending plan went from about $200 billion to about $220 billion after memory prices climbed, and management has raised this figure before.
In short: Custom chips and cloud scale are advantages; open antitrust cases and China-based supply chains are the risks.
Amazon says the chips business runs above a $25 billion annual rate and that Trainium carries most Bedrock inference, giving it a cost edge over rivals who buy GPUs.
Ads are the highest-margin large business inside Amazon retail, and its shopping data lets brands target buyers in ways other ad platforms cannot match.
Amazon settled an FTC matter and still faces open investigations, which can force changes to how it runs Prime, its marketplace, or AWS.
A large share of seller services and advertising revenue depends on China-based sellers, and much of the goods sold is sourced there, so tariff changes hit both selection and cost.