Amazon Q2 2026 earnings: AWS revenue grew 37% year-over-year to $42.2 billion, its fastest pace since 2021, driving a 9-14% stock surge even as capex guidance rose to $220 billion and free cash flow turned negative.

Amazon's headline Q2 2026 numbers look almost too good: net income of $62.6 billion, or $5.75 per diluted share, blew past the $1.82 consensus estimate^[1]. But the scale of that beat has less to do with retail or cloud operations than with a single balance-sheet event - a pre-tax gain of more than $53 billion tied to the mark-to-market value of Amazon's equity stake in Anthropic^[2]. Strip that gain out and the underlying earnings picture is considerably less dramatic, a detail largely absent from the celebratory framing that followed the print. Investor discussion circulating alongside the earnings reaction estimated that, excluding the Anthropic gain, net income was closer to roughly $9 billion for the quarter, with adjusted EPS separately estimated by other investors as missing prior guidance and landing close to flat year-over-year - two independent back-of-envelope community estimates using different methodologies, not a single reconciled figure. None of this makes AWS's operating performance fake: AWS operating income genuinely rose to $16.6 billion at a 39.4% margin^[1]. But it does mean that much of the enthusiasm behind Amazon's 9-14% post-earnings stock surge^[3]was reacting, in significant part, to a one-time investment gain rather than a repeatable jump in core profitability.

A tougher question than the accounting quirk is a structural one raised in investor discussion around the print: how much of AWS's headline 37% growth^[1]reflects external enterprise demand versus Amazon's own capital circulating back to itself. Amazon holds a large and growing equity stake in Anthropic^[2], one of AWS's biggest AI compute customers, and community financial analysis flagged that Amazon's quarterly purchases of non-marketable securities - largely tied to Anthropic - run into the tens of billions of dollars, raising the question of how cleanly an 'AI revenue run-rate' can be separated from investment dollars flowing back through the same balance sheet. Nothing in Amazon's public disclosures separates organic AWS demand from demand effectively pre-funded by Amazon's own investment activity, and that ambiguity matters more as AWS's contracted backlog and AI/Chips run-rates get cited as evidence of durable, market-driven growth^[4]. It doesn't invalidate the acceleration - AWS's AI and Chips businesses each independently topped $25 billion in annualized run rate^[1]- but it complicates any clean read of AWS's growth as pure external validation of enterprise AI adoption. A widely-shared rebuttal in the same discussion pushed back on parts of the adjusted-EPS math, but even that rebuttal conceded the core point: the filing doesn't let anyone cleanly separate organic AWS demand from Amazon-funded demand.

The clearest, least disputed cost of AWS's growth spurt shows up in Amazon's cash flow statement. Trailing-twelve-month free cash flow swung from an $18.2 billion inflow a year earlier to a $7.6 billion outflow^[5], driven entirely by capex outpacing gains in operating cash flow. Amazon simultaneously raised its full-year 2026 capex guidance to about $220 billion, up roughly 10% from the prior $200 billion figure, attributing the increase specifically to rising memory chip costs rather than added capacity volume^[6]. Jassy's own framing on the earnings call underscores why the spending isn't slowing: even at $220 billion, AWS still will not have enough capacity to meet 2026 demand, and he described already-visible 2028 demand as striking, implying the capacity shortfall - and the capex needed to close it - persists well into 2027^[4]. Community skeptics also flagged how that spending is increasingly being financed: long-term debt reportedly rose sharply quarter-over-quarter to roughly $129 billion, a recent bond offering was only modestly oversubscribed versus Amazon's historical norm, and datacenter buildouts are leaning on off-balance-sheet financing structures involving private-credit partners. Taken together, analysts frame this as a genuine inflection point for AI capex economics - one where hyperscalers' ability to keep converting spending into growth is being tested in real time^[7].

The starkest single-week contrast in this earnings cycle wasn't between Amazon and its cloud peers - it was between Amazon and Apple. Both companies reported the same week, and both were shaped by the same AI-driven chip and memory supply squeeze, but in opposite directions^[3]. Amazon converted surging AI infrastructure demand into a 37% AWS growth beat and rallied as much as 12% the next session, while Apple fell roughly 8% on weak guidance tied to the same constrained chip supply chain^[8]. Apple CEO Tim Cook explicitly flagged the 'increasing impact' of the memory shortage on the company's outlook^[9]- the identical dynamic that let AWS raise prices and expand margins was squeezing Apple's hardware costs and availability. That divergence illustrates a broader pattern worth tracking through the rest of this AI infrastructure cycle: the same underlying scarcity (AI accelerators, high-bandwidth memory) can simultaneously reward cloud providers who can pass costs through to enterprise customers and punish device makers who compete for the same components at fixed consumer price points. AWS's own growth trajectory - 24% in Q4 2025, 28% in Q1 2026, then 37% in Q2 2026^[10]- shows this is an accelerating story, not a one-quarter blip, which raises the stakes for how long the chip/memory bottleneck keeps cutting both ways.