Amazon Is Two Businesses Wearing One Stock Price

Amazon’s headline last quarter said net income of $62.6 billion, nearly triple a year earlier. Read the release and a different number shows up: real operating income of $27.5 billion, up 43%. The gap between those two figures is $53.4 billion, and almost all of it is a non-cash gain from marking up Amazon’s stake in Anthropic. One of those numbers tells you how the business is doing. The other tells you what an accountant did with a private stock position. Confusing the two is the easiest mistake to make with this stock right now.

Amazon trades around $253.71 as I write this, a market cap near $2,736.6 billion. My thesis is that this is really two businesses sharing one ticker: a retail operation that still produces most of the revenue and a cloud business that produces most of the profit growth, and the accounting noise from the Anthropic stake this quarter made that split harder to see, not easier.

Retail carries the revenue, AWS carries the growth

Break down where the money actually comes from. In the quarter ended June 29, North America retail brought in $116.2 billion, or 57.9% of total revenue. International retail added another $42.2 billion, or 21.0%. Amazon Web Services brought in the same $42.2 billion, 21.1% of revenue, essentially tied with the entire international retail operation. Total revenue for the quarter was $200.6 billion, up 20% from a year earlier and 11% sequentially.

Here is the part that matters more than the split itself. AWS revenue grew 37% year over year, its fastest pace in 18 quarters, while total company revenue grew roughly 20%. A segment doing a fifth of the revenue is growing at nearly double the company’s overall rate. That is not a rounding error. It is the reason AWS keeps taking a larger share of Amazon’s operating profit even while staying a minority of revenue, and it is the main reason I think about this as two separate businesses rather than one diversified retailer.

Retail is not standing still either, and I do not want to undersell it. Gross margin across the whole company came in at 50.3%, up from 48.9% a year earlier, and full-year revenue reached $716.9 billion, up 12% from $638.0 billion in 2024. Some of that gross-margin gain comes from advertising and third-party seller services, both of which sit inside the retail segments and carry far better margins than shipping boxes does. Retail funds the experiments. It is just no longer the part of the business setting the pace.

Funding the buildout with debt, not just cash flow

One detail that got less attention than the Anthropic gain: Amazon priced its first-ever sterling-denominated bond this month, issued in four tranches with spreads of 70 to 110 basis points over benchmarks for maturities ranging from three to 19 years. A company sitting on the cash flow Amazon generates does not need to tap a new currency and a new investor base unless it wants to diversify how it funds a capital program of this size, or wants to lock in financing before rates move further. Either reading is fine by me. What I would not want to see is a debt load growing faster than operating cash flow for several quarters running, because that is the pattern that turns an aggressive capex plan into a balance-sheet problem instead of a growth story.

The $53 billion number that is not profit

Now the part that got most of the headlines for the wrong reason. Amazon’s reported net income of $62.6 billion for the quarter included a $53.4 billion non-cash, pre-tax gain from remeasuring its stake in Anthropic, after the company added another $5 billion of nonvoting preferred stock to that position during the quarter. That gain runs through the income statement under current accounting rules for equity investments with observable price changes, which is why a stake in a private AI lab can single-handedly triple a quarter’s reported earnings.

Operating income, the number that actually reflects running the business, was $80.0 billion for the trailing year and $27.5 billion for the June quarter alone, up 43% from $19.2 billion a year earlier. That 43% is a real, healthy number on its own. It simply is not the number sitting in the headline, and treating trailing earnings per share of $12.43 as a clean run rate would badly overstate what this company earns from operating a retail and cloud business, which is what most of its valuation has to be based on going forward.

AWS margins keep climbing while revenue reaccelerates

AWS operating income came to roughly $16.6 billion in the quarter, up from about $10.2 billion a year earlier, for an operating margin near 39.4%, up from 32.9%. A cloud segment growing revenue 37% while also expanding margin by more than six points is doing the opposite of what happens when a business chases growth by cutting prices. Microsoft’s cloud business tells a version of the same story at a different scale, and I read AWS’s acceleration as more evidence that enterprise AI workloads are still being built out faster than anyone forecast a year ago, not slower.

I want to flag where I am inferring rather than citing a hard number. Amazon does not break out exactly how much of AWS’s growth is generative-AI infrastructure demand versus ordinary cloud migration. My own read, based on the size of the reacceleration after several quarters near 17-19% growth, is that AI workloads are now a meaningful driver of the segment, not just a talking point on the earnings call. That is an inference, and a reasonable person could weight it differently.

Capex is rising to fund both businesses at once

Amazon raised its 2026 capital spending plan to around $220 billion, citing higher memory chip costs among other inputs. That is an enormous number, larger than the total revenue of most companies in the S&P 500, and it funds data centers, chips and warehouse automation at the same time. Some of that spending buys the GPUs and custom silicon that companies like Nvidia and AMD sell into the same AI buildout, which is worth knowing if you also hold either of those names, since Amazon is simultaneously a huge customer and, through AWS’s own Trainium chips, a long-term competitor to both.

Heavy capex cycles are where cloud economics can turn ugly if demand does not show up on schedule. Salesforce’s struggle to prove that AI features translate into revenue per customer is the cautionary version of this story: spending ahead of monetization works only as long as the monetization eventually shows up. AWS’s margin expansion this quarter argues that, for now, Amazon’s spending and its returns are still moving in the same direction. A few quarters of capex growing faster than AWS revenue, without margin holding up, is the pattern that would make me reconsider that.

Two P/E numbers that cannot both be read simply

Trailing earnings put Amazon at 20.4 times earnings, far below its own five-year average of 39.9 times. Forward earnings put it at 26.9 times, well above the trailing figure. Normally a stock getting cheaper on a forward basis signals slowing growth. Here it is the opposite of that signal: the trailing multiple looks artificially low because the Anthropic gain inflated the denominator, and the forward multiple is the more honest read on what the operating business is actually expected to earn. Consensus estimates imply forward earnings per share of $9.42, which on paper is a decline of -24% from trailing EPS of $12.43. That decline is almost entirely the comparison against an inflated base, not a forecast that the business will earn less next year.

Price-to-sales tells a cleaner story since revenue is not distorted by the equity gain: 3.6 times sales now against a five-year average of 2.6. That is a real premium versus Amazon’s own history, and it is being paid for a company growing revenue near 20% with a cloud segment growing at almost double that.

Metric	Amazon now	Five-year average
P/E (trailing)	20.4	39.9
P/E (forward)	26.9	n/a
P/S	3.6	2.6
P/B	5.0	6.5
Quarterly revenue growth	20%	n/a
AWS revenue growth (Q2 2026)	37%	n/a

What would prove the split wrong

The stock jumped +15.3% on earnings day, against an average earnings-day move of 7.8%, so the market clearly read through the accounting noise to the operating numbers rather than reacting to the headline profit figure. That is a reasonable reaction. Analyst targets average $335, ranging from $250 to $400 across 39 estimates, with 97% rating it a buy, which suggests Wall Street has largely made the same adjustment I am describing here.

My one specific counter-case, and the main risk to this thesis: if AWS revenue growth slows back toward the high teens over the next two quarters while capex keeps climbing toward that $220 billion run rate, the margin story reverses and the multiple gets harder to defend, because the market is currently paying a premium price-to-sales ratio for an acceleration that would no longer be happening. Watch the next quarterly AWS growth figure against this quarter’s 37% more closely than the headline net income line, which will keep being noisy as long as the Anthropic stake sits on the balance sheet at a fluctuating value.

Short interest is just 0.9% of float, so there is almost no bearish positioning to unwind if the next quarter disappoints, which cuts both ways. It means less of a short-covering tailwind on good news, and less of a cushion if the market decides the capex number is bigger than the growth story can support. Amazon’s quant rating has held at a middling C through this entire earnings cycle, unchanged from before the report, which tells me the market’s own models are not yet convinced the AWS acceleration changes the multiple-year picture, even after a 15% single-day jump in the stock.

Analysis and opinion only, not investment advice. Figures come from Amazon’s quarterly earnings report and its filings on SEC EDGAR; valuation multiples are approximate and were checked on September 18, 2026.