Amazon posted record Q2 revenue of $200.6 billion and 43% operating income growth, supported by strong adoption of in-house Trainium and Graviton AI silicon.

Amazon closed Thursday at $265.13, down $2.15 or 0.80%, inside a session range of $264.71 to $269.58 after opening at $267.24. The after-hours print ticked to $265.28, up $0.15. Market cap sits at roughly $2.86 trillion, down 5.71% across the past week.

That figure matters because on August 3, the number carried a three in front of it. Amazon closed that Monday at $284.02, up 4.58%, after touching an all-time high of $287.20 and crossing $3 trillion in market capitalization for the first time — joining Apple, Microsoft, Nvidia and Alphabet. The stock had climbed roughly 20.6% across two sessions following the July 30 earnings release and was up 25.40% year to date.

Eight trading days later, the entire breakout is gone. From $287.20 to $265.13 is a 7.7% drawdown, and the stock now trades $22.07 below its record with the trailing 30-day range spanning $226.16 to $287.20 and an average of $254.78.

The tape underneath tells the story. The stock has fallen in 7 of the last 10 sessions. Volume on Thursday ran well below the 30.4 million daily average. A pivot-top sell signal fired on August 3 — the exact day of the record — and price has not established a new bottom pivot since.

Against the 52-week range of $196.00 to $287.20, current price sits in the upper third. Against the year, the stock is up 14.78%. Against the April low, it remains roughly 30% higher. This is a drawdown inside an uptrend, not a broken chart.

What makes it worth dissecting is that nothing about the fundamental picture changed between $287.20 and $265.13. Amazon reported the best quarter in its corporate history on July 30 — the first $200 billion revenue quarter, AWS growth at an 18-quarter high, record operating margin — and the stock is now trading 7.7% below where it stood the day after. The market repriced the same numbers within eight sessions. That happens when the initial reaction prices the revenue line and the second reaction prices the cash flow line.

Second-quarter revenue reached $200.6 billion, up 20% year over year and clearing consensus at $196.16 billion per Amazon's Q2 2026 results. It was the first time the company crossed $200 billion in a single quarter.

Operating income came in at $27.5 billion, up 43% year over year. Operating margin expanded to 13.7% from 11.4% — a record for the company and a 230 basis point improvement. Revenue growth of 19.6% outpaced an 18.0% increase in operating expenses, which is the cleanest possible demonstration that scale is still producing leverage at half a trillion dollars of annualized revenue.

The North America segment generated $116.2 billion, up 16%. International contributed the balance, running roughly 22% of consolidated revenue led by Germany, the United Kingdom and Japan. Retail-related revenue accounts for approximately 74% of the total, AWS 17%, and advertising services 9%.

Beneath the headline, the consumer business showed genuine momentum rather than mix-shift. Perishable customer growth has risen 50% since the start of the year. Same-day delivery now reaches 2,300 cities. Grocery and everyday essentials — the categories Amazon has spent a decade failing to crack at scale — are finally compounding.

A $600 million tariff-related refund flowed through as a one-time benefit. Management stated it largely absorbed tariff costs rather than passing them to customers, which supports volume at the expense of margin and is a decision that shows up in future quarters rather than this one.

Cost pressures were flagged explicitly: higher memory chip prices, fuel inflation and rising linehaul rates. Memory in particular is not a transient issue — it is the same shortage that forced the capex revision and that has driven Sandisk up sixfold this year.

The operating quarter stands entirely on its own. Strip out every accounting item and Amazon delivered 20% consolidated growth that accelerated sequentially, record margins, and 43% operating income expansion at a company generating $173.58 billion in trailing EBITDA at a 20.97% margin.

The stock is down 7.7% from the day it printed those numbers.

The segment that carries the entire thesis delivered its best quarter since 2021. AWS revenue hit $42.2 billion, up 36.7% year over year against consensus at 31% — a six percentage point beat on the single most watched line in the report.

The acceleration sequence is what makes it credible. AWS grew 16.9% in Q1 2025, 17.5% in Q2, 20.2% in Q3, 24% in Q4, 28% in Q1 2026 and now 36.7%. Five consecutive quarters of accelerating growth in a business at $169 billion annualized run rate is close to unprecedented at that scale.

Margins expanded harder than revenue. AWS operating income reached $16.6 billion against $10.2 billion a year earlier, producing a 39.4% operating margin, up approximately 650 basis points year over year.

That headline needs its footnote. Roughly $600 million of the margin came from mark-to-market gains on energy derivative contracts — a first-time material item. Excluding it, margin expansion was about 520 basis points. Q3 guidance explicitly assumes no impact from energy derivative remeasurements going forward. The underlying expansion is real; the printed number overstates it by roughly 130 basis points.

The mechanics behind the margin are worth understanding because they determine durability. Software and process improvements optimize server capacity. Custom silicon and custom network gear lower the cost base per unit of compute. Capacity optimization reduces idle infrastructure. None of those are cyclical — they compound as the fleet grows.

The risk sits on the depreciation line. Net property and equipment reached $446.046 billion, with AWS receiving 76% of Q2 additions. If AWS growth decelerates before newly built capacity fills, depreciation pressures that 39.4% margin even while cash spending stays elevated. That is the specific mechanism by which a growth story becomes a margin story becomes a valuation problem.

For now the sequence is running the right direction: capital is landing where demand and margins are strongest, and the revenue acceleration arrived in the same period as the spending.

Remaining performance obligations — contracted AWS revenue not yet recognized — jumped to $496 billion from $364 billion the prior quarter. That is a $132 billion sequential increase in a single three-month window, against a figure that stood at $244 billion at the end of 2025.

At the current AWS revenue pace, $496 billion represents roughly three years of forward coverage. It is the reason the market tolerated a $20 billion capex increase: the demand is pre-sold rather than speculative.

The concentration is the problem. Approximately $100 billion of the $132 billion sequential increase is attributable to a single counterparty — Anthropic's ten-year Trainium commitment signed in April 2026. Add OpenAI's multi-gigawatt commitment, layered on top of a $50 billion Amazon investment in the company with Trainium usage projected to consume 2 gigawatts of capacity, and the two leading AI labs account for the dominant share of the backlog.

That creates a specific structural exposure. AWS's forward revenue is now substantially a derivative of whether two private, cash-burning research companies can fund multi-year infrastructure commitments. Both are financed by capital markets that have shown they can close abruptly.

Beyond the labs, the customer list disclosed includes Uber and Pinterest plus AI startups like Neura Robotics and Odyssey. Real, but not $100 billion real.

The counterargument is that contracted backlog with named counterparties is materially better than the alternative, which is building capacity against forecast demand. Alphabet raised capex guidance and its stock fell. Amazon raised capex by more and its [text cuts off]