Amazon Stock: most of last quarter’s profit was a markup, not a sale

Amazon reported $62.6 billion of net income for the June quarter, up 245% from a year earlier. Most of it was not a sale.

Inside the same filing, Amazon discloses that “upward adjustments relating to equity investments in private companies of $50.5 billion in Q2 2026… reflect observable changes in prices, primarily from our nonvoting preferred stock in Anthropic.” That is an unrealised markup on a stake in a private company, and it is 1.8 times larger than everything Amazon’s retail, advertising and cloud businesses earned put together.

This is not an accounting irregularity. It is required treatment, disclosed plainly, and the underlying gain is real in the sense that someone recently paid those prices. But anyone holding Amazon stock on the strength of a 245% earnings jump is holding something different from what the headline implies.

Article Brief

Key Takeaways

5 Points30s Read

1. 01The headline–Amazon reported $62.6 billion of Q2 net income, up 245%, and diluted EPS of $5.75.
2. 02What drove it–A $50.5 billion unrealised markup on nonvoting preferred stock in Anthropic — 62.4% of pre-tax income and 1.84x Amazon’s entire operating income.
3. 03Ex-markup–Strip it out and pre-tax income was about $30.4 billion against $20.9 billion a year earlier, growth of roughly 46% rather than 288%.
4. 04The position–Amazon invested $8.0 billion in Anthropic convertible notes from Q3 2023 to Q4 2025; reporting puts the whole stake near $190 billion, about $17.65 per Amazon share.
5. 05Why it matters now–Anthropic is in talks to list near a $2 trillion valuation, which would replace an internal estimate with a market price — in both directions.

The arithmetic, from the filing

Amazon’s second-quarter report lays the pieces out in one column.

Total net sales were $200.6 billion, up 19.6%. Operating income — what the actual businesses produced — was $27.5 billion, up from $19.2 billion. Then comes the line that does the work: other income, net of $53.4 billion, against $1.1 billion in the same quarter last year.

That single line pushed income before taxes to $80.9 billion from $20.9 billion. Of it, $50.5 billion is the Anthropic markup.

Set those numbers side by side and the proportions are hard to unsee. The Anthropic adjustment equals 62.4% of Amazon’s pre-tax income for the quarter and 1.84 times its entire operating income. At the quarter’s effective tax rate it works out to roughly $3.59 of the $5.75 in diluted earnings per share Amazon reported.

Strip the markup out and pre-tax income was about $30.4 billion against $20.9 billion a year earlier — growth of roughly 46%. That is a good quarter by any normal standard. It is not a 288% one.

What Amazon actually owns

The position itself is well documented in the filing. Amazon invested $8.0 billion in Anthropic convertible notes between the third quarter of 2023 and the fourth quarter of 2025. Portions of those notes converted into nonvoting preferred stock in the first quarters of 2025 and 2026, which is why gains keep moving between comprehensive income and the income statement.

Published reporting puts Amazon’s total committed investment considerably higher than the $8 billion of notes and its stake somewhere between 15% and 21%, with the carrying value of the whole position around $190 billion at the end of June. Amazon does not publish a single consolidated stake percentage, and the convertible notes are subject to an ownership cap the company can waive at its election.

Against a market capitalisation of about $2.77 trillion and 10.79 billion shares outstanding, a $190 billion position is roughly $17.65 per Amazon share, or about 6.9% of the company’s market value.

Why the IPO changes the question

Anthropic is now in talks to go public. TECHi reported on Friday that it is seeking as much as $100 billion at a valuation near $2 trillion, with Nvidia weighing up to $10 billion as an anchor investor.

A listing converts an estimate into a price. Right now Amazon values its position using, in the filing’s own words, “valuation methods based on information available, including the rights and obligations of the convertible notes, other outstanding classes of securities, observable transactions such as new securities offerings, estimates of expected time to and type of liquidity events… and discounts for lack of marketability.” That is a careful, defensible process, and it is still an estimate carrying a discount for the fact that nobody can sell the thing.

If Anthropic lists near the valuations being discussed, that discount narrows and the position gets marked against a screen rather than a model. Some analysts have put the stake’s potential value above $400 billion at those levels. Whether it lands there or not, the mechanics are the same: Amazon’s reported earnings become more sensitive to another company’s share price, in both directions.

That is the part worth holding onto. A markup this size is not a one-way benefit. The same accounting that delivered $50.5 billion of gains in one quarter will deliver markdowns if Anthropic’s valuation falls, and a public listing makes those moves faster and more visible.

The commercial side is the better story

Underneath the mark, the operating business did something more durable.

AWS revenue reached $42.2 billion in the quarter, up 36.8%, and AWS operating income reached $16.6 billion, up 63.6%. That single segment produced more operating profit than North America and International retail combined. Amazon also disclosed that it delivered on capacity commitments tied to its own silicon.

The filing quantifies two of those relationships. In the first quarter of 2026, AWS and OpenAI expanded an existing $38.0 billion multi-year commitment by a further $100.0 billion over eight years. In the second quarter, AWS and Anthropic expanded their arrangement by more than $100.0 billion over ten years. Both, the filing notes, include “contractual obligations related to the performance of AWS chips.”

Read those two sentences together and the shape of Amazon’s AI position becomes clear. It has locked in enormous long-dated cloud commitments from the two leading model developers, and both commitments are tied to how well Amazon’s own Trainium silicon performs. TECHi covered the chip side of that arrangement when the supply pace first became visible.

The awkward symmetry is that Amazon is an investor in one of those customers. Money goes out as investment and comes back as committed cloud spend — the same circular structure running through Oracle’s $664 billion backlog, where roughly half the contracted work traces to a single AI lab.

Amazon has also kept adding. In April it agreed to invest up to a further $25 billion as part of an AI infrastructure arrangement, and Fortune has noted that a listing is what will finally settle how much Amazon and Google actually have riding on Anthropic. Until then, the size of the position is an estimate that Amazon itself produces.

Where the stock actually sits

Amazon closed Friday at $256.78, up 1.9% on the day. It is up 13.4% for the year and 7.7% since the end of June.

That quarter-to-date figure is the interesting one. In TECHi’s ranking of 41 AI stocks this quarter, Amazon placed thirteenth — well behind the enterprise software names that led, and well ahead of the semiconductor and neocloud names that fell. For a company that just booked a $50 billion gain on the hottest private asset in technology, a middling quarter is itself a signal: the market appears to be discounting the markup rather than capitalising it.

That is arguably rational. Unrealised gains on an illiquid private holding are worth less than operating cash flow, and investors have generally declined to pay a full multiple for them.

The valuation tell

There is one number that gives the whole thing away, and it takes a second to spot.

Amazon trades on a trailing price-to-earnings multiple of about 20.3 and a forward multiple of about 23.6. That ordering is backwards. For a company whose earnings are expected to grow, the forward multiple should be the lower of the two, because next year’s earnings are bigger. When trailing is cheaper than forward, it usually means the trailing number is carrying something that is not expected to repeat.

Here, it is carrying the markups. Amazon disclosed $62.8 billion of upward adjustments for the first six months of 2026 — $50.5 billion in the second quarter and roughly $12.3 billion in the first. After tax, that is about $4.47 per share of the $12.66 in trailing twelve-month earnings.

Take those two quarters out and trailing earnings fall to roughly $8.19 a share, which puts Amazon on about 31.3 times earnings rather than 20.3. That is a materially different-looking stock, and it is a conservative version of the adjustment: any markups recorded in the second half of 2025 would push the adjusted figure lower still.

None of this makes the reported multiple wrong. It makes it a poor basis for comparison with companies whose earnings come entirely from selling things. Wall Street appears to understand this — the consensus target sits near $328 against a $256.78 close, with 59 of 61 covering analysts at buy or strong buy and none at sell — but the screen-level multiple does not show it.

What to watch, and what not to

The wrong lesson here is that Amazon’s earnings are somehow fake. They are not. The markup follows observable transaction prices, the disclosure is explicit, and the underlying asset is one of the most sought-after private companies in the world.

The right lesson is narrower. When you read Amazon’s headline earnings growth, separate the two engines. One is a retail, advertising and cloud business compounding at a healthy but ordinary rate — pre-tax income up roughly 46% excluding the mark, with AWS growing 36.8% and its margins widening. The other is a stake whose value is set by private financing rounds and, soon, possibly by a public market.

Three things will tell you how this develops: whether Anthropic’s listing actually happens and at what price, whether AWS keeps converting those hundred-billion-dollar commitments into delivered revenue, and whether Amazon’s reported earnings start swinging with another company’s valuation in quarters when that valuation goes the other way.

The first two are upside. The third is the one nobody has had to price yet.