Amazon Q2 2026 Earnings: AWS Growth and AI Capex Surge

The headline number was impossible to miss: net income more than tripled to $62.6 billion. But roughly $53.4 billion of that pre-tax profit came from a single non-operating line - a markup on Amazon's investment in Anthropic - not from selling more cloud capacity or ads ^[1]. That distinction matters because it changes what the quarter actually says about Amazon's core business versus its balance sheet. It is also exactly what a widely-upvoted community read flagged before the market's excitement had fully settled: the composition of the profit beat, not just its size, deserves scrutiny, and once the Anthropic markup is set aside, the underlying cash picture looks considerably weaker - trailing-12-month free cash flow swung to a $7.6 billion outflow from an $18.2 billion inflow a year earlier, as trailing-12-month capex hit $169 billion, up 64% year over year ^[3]. A counter-argument in that same discussion held that Amazon's adjusted results, stripped of one-time gains, would still have beaten estimates on their own, and that near-term cash burn is a reasonable price for demand growth of this magnitude. Both readings are defensible - which is itself the point: a single quarter's headline number is doing a lot of interpretive work here, and investors cheering the 7-10% after-hours pop were reacting to the top-line growth story more than the accounting behind the profit figure ^[3].

Amazon's stock jumped after the report, but the more telling comparison is what happened to companies making similar bets without Amazon's business model. Meta Platforms spent heavily on AI infrastructure too, yet its shares fell about 8% after its own report because investors could not draw as direct a line from that spending to revenue ^[2]. Amazon, by contrast, gets to point at AWS's 37% revenue growth and a 36.8% operating margin - ahead of Google Cloud's 35.6% ^[3]- as concrete proof the capex is converting into paying customers ^[2]. As one industry commentator put it bluntly, Amazon's hosting revenue is someone else's AI bill: AWS profits regardless of which company's model wins, because it is renting out the infrastructure underneath the race rather than betting on a single outcome ^[2].

Andy Jassy used the earnings call to make an explicit strategic argument: AWS does not need to own a frontier model to win the AI infrastructure race. "AWS and Amazon Bedrock can have a wildly successful business without its own frontier model, and the reason is that there's not going to be a single model to rule them all," he said ^[2]. The numbers back the bet so far - AWS's AI business and its custom chips business each exceeded $25 billion in annualized revenue run rate, more than doubling year over year, up from an AI run rate of just over $15 billion as recently as Q1 2026 ^[5][6][9]. Jassy went further, suggesting the AI opportunity could eventually become a trillion-dollar annual revenue business for AWS with strong free cash flow and return on invested capital ^[7]. It is a deliberately different wager than the one frontier-model labs are making - Amazon is betting on being the neutral pipes and chips underneath whichever models win, rather than fielding the winning model itself, and this quarter's growth numbers are the first hard evidence that customers are actually buying into that pitch at scale.

Amazon raised its 2026 capex guidance from $200 billion to $220 billion, explicitly citing higher memory chip costs ^[4]. That number was not a surprise in direction - Alphabet had already lifted its own 2026 guidance from $180-190 billion to $195-205 billion before Amazon reported, and BofA's Justin Post had predicted Amazon would land near $210 billion on rising memory prices, just under the actual figure ^[3][8]. What stands out is Jassy's own admission about what that spending buys: even at $220 billion, "we will still not have enough capacity to meet all the demand we have in 2026, and I believe this dynamic will also be true in 2027 too" ^[3]. That statement reframes the entire capex story - this is not discretionary growth spending Amazon could dial back if demand softened, but a company openly rationing infrastructure against a demand curve it cannot fully satisfy. It also lines up with the growth trajectory that got AWS here: 20% growth in Q3 2025, 24% in Q4 2025, 28% in Q1 2026, and now 37% in Q2 2026 - the fastest pace since the quarter ending December 2021 ^{[9][10][11][12]}. Five straight quarters of acceleration is the evidence Jassy is using to justify spending Amazon cannot yet turn into positive free cash flow, and Cowen's Derrick Wood had warned before the print that another capex raise, following Google's precedent, could still spook the market regardless of the growth numbers behind it ^[3].