Bufetico
In its second quarter of 2026, Amazon reported earnings of $5.75 a share. The consensus estimate was $1.81.
Put that number under the price and the price-to-earnings ratio drops a lot from the week before. That’s the ratio you’d reach for first, and here it’s the one that will mislead you fastest. The numbers that decide the question are further down in the release.
What the earnings figure contains
Amazon’s net income for the quarter was $62.6 billion. Of that, roughly $53.4 billion came from non-operating other income, mostly gains on its investments in Anthropic.
That’s a real gain. It didn’t come from selling anything on the marketplace or from renting servers. It came from a stake going up in value on paper.
So the earnings figure that made the quarter look so good is mostly not the business. Divide the price by that number and what you’ve measured is the price against a one-off paper gain on a private holding.
Check this for any company: before you use an earnings number, find out how much of it came from operations. A single line in the release usually tells you.
So go to the cash
This is why I start from cash. Cash is harder to shape with accounting choices, and a gain on an investment you haven’t sold doesn’t show up in it.
And this is where Amazon gets interesting. Free cash flow over the twelve months to June 2026 was minus $7.6 billion.
Negative. For a company that produced $38.2 billion of it in the twelve months to the end of 2024, by Amazon’s own definition.
A screener that runs on cash will either fail Amazon outright or leave the square gray, depending on how it’s built. Either way, the standard method has just stopped giving you an answer, and you’re back to reading.
Why the cash went negative
Two numbers explain it, and they point in opposite directions.
Operating cash flow over the same twelve months was $161.4 billion, up 33% year over year. The business made more cash than it ever has.
Purchases of property and equipment rose by $66.1 billion. That’s the gap. The cash arrived and got spent, on data centers and AI infrastructure.
So the negative figure comes from a decision to spend everything the business makes, and then some, on building capacity. Whether that decision pays off is a separate bet, and the cash flow statement says nothing about it.
The trap from the other direction
“The cash is negative because of capital spending” is where a lot of analysis stops, and stopping there is its own mistake.
Spending isn’t automatically investment. Money poured into capacity that doesn’t earn a return is money gone, and on the cash flow statement it looks identical to money poured into capacity that does. The line item can’t tell them apart, and neither can any ratio built on it.
What eventually tells them apart is whether the spending shows up in returns years later. That evidence doesn’t exist yet.
Amazon has done this before. In 2021 and 2022 its free cash flow went negative, minus $14.7 billion and then minus $16.9 billion, while it doubled its warehouses, and in the two years after that it came back to $32.2 billion and $32.9 billion. I wrote about how to read a series like that.
It worked out that time. That tells you the pattern isn’t automatically a warning, and nothing more.
What I’d check
Four things, in this order, and none of them produce a verdict on their own.
How much of the earnings came from operations. If most of it didn’t, any ratio built on earnings is describing something else.
Whether operating cash flow is still growing. This separates a company that’s spending from a company that’s shrinking. Amazon’s grew 33% over the year.
Whether the shape of the series is falling or interrupted. Three consecutive years of declining cash with no spending story behind it is a different situation from one sharp drop with a capital program attached. The shape matters, and the average alone won’t show it.
What the spending has to earn back for the price to make sense. That’s the valuation question, and no filing settles it for you.
Where this breaks
Reading the numbers the way I’ve laid out tells you what the company did with its cash. Whether the capital program pays is a different bet, and a filing can’t settle that one either.
There’s also a risk in the framing itself. “Negative cash flow is fine when it’s investment” has excused a lot of money being burned over the years, and it sounds just as reasonable when it’s true as when it’s not. The only defense is to insist on the operating cash line: a company that’s really deteriorating shows it there, and no amount of capital spending explains that away.
And my own cash measure had the same blind spot, which I didn’t see when I wrote this. It counted every dollar Amazon put into a data center as a dollar lost, the same as if it had vanished. That’s the first-to-last mistake again, wearing a different hat. Two days after this went up I found it and changed the measure.
A method that leaves the square gray is frustrating. It’s still better than a number that pretends it was all settled.
The short version
The earnings figure is mostly an investment gain. The cash figure is negative because the company is spending more than it makes on infrastructure. Both are one line each in the release, and either one on its own will walk you confidently in the wrong direction.
What the spending buys is a judgment you have to make and own. A fair value with two decimals already made that call for you, and hid it in the math.