Amazon reported second-quarter results Thursday afternoon, and the headline profit number is close to useless. Net income: $62.6 billion, against $18.2 billion a year ago. Earnings per share: $5.75, against $1.68 last year and a street looking for $1.82. That is not a beat. That is a revaluation wearing a beat’s clothing.
Sitting inside the number is $53.4 billion of pre-tax non-operating income, primarily a gain on Amazon’s investment in Anthropic. It is a mark on a stake in a private company, not cash collected from customers, and marks that big move in both directions. Take it out and the operating business earned $27.5 billion, up from $19.2 billion a year ago — an excellent quarter that nobody would have described as a quadrupling.
Revenue is the clean line. Net sales came in at $200.6 billion, up 20% from $167.7 billion, roughly $4 billion past what analysts modeled and the first $200 billion quarter in the company’s history.
The number that actually moved the stock
AWS did $42.2 billion, up 37% year over year against a street near $40.5 billion — the cloud unit’s fastest growth since 2021. Its operating income was $16.6 billion, up from $10.2 billion a year ago and about $3 billion past consensus, which means the margin widened while the capital spending ran hot. Amazon also said the AWS artificial intelligence business and its in-house chips business have each passed a $25 billion annualized revenue run rate, both more than doubling from last year. Advertising grew 26%.
Shares rose more than 9% in extended trading. Not for the $62.6 billion. For the 37%.
Our take: This week the market wrote down its rule for the AI buildout, in three verdicts. Microsoft got paid because Azure accelerated to 43%. Meta got punished because it spent like Microsoft and had no acceleration to show for it. Amazon just got paid because AWS grew 37% and the cloud margin went up while the capex went up. The buildout is fundable — on one condition, which is that the cloud line grows faster this quarter than it did last quarter. Everyone else gets asked to justify the invoice. And when a headline EPS quadruples on an investment mark, the number worth reading is always two lines down.
The part they will argue about Friday
Guidance is the soft spot. Amazon told investors to expect third-quarter revenue of $197.0 billion to $202.0 billion — growth of 9% to 12%, against a street near $204 billion and a long step down from the 20% just posted. Operating income guidance of $22.5 billion to $26.5 billion compares with $17.4 billion in the same quarter last year, so the profit engine is still climbing even as the top-line guide comes in light.
That gap is the whole argument. Bulls will say Amazon guides conservatively and the AWS backlog says otherwise. Bears will say a company growing 20% does not guide to 9% unless something in the retail business is cooling. Both sides are reading the same page.
What to watch
- Whether 37% holds. Acceleration is the entire thesis. One quarter of deceleration and the stock gets graded like Meta was.
- The Q3 revenue guide versus the actual. If Amazon lands above the high end, the conservatism story wins and the light guide is forgotten.
- The Anthropic mark. A gain that size becomes a comparison problem next year, and a headline risk if private AI valuations cool.
- The capital line. The market has now shown it will fund enormous spending — but only while the cloud revenue attached to it is speeding up.
Amazon spent the quarter building the most expensive infrastructure in corporate history and closed the books with a profit number that mostly reflects someone else’s valuation. The real result was underneath: the cloud is growing faster than it has in five years, and for now, that is what buys the data centers.