Ackman's Amazon Bet: AWS Growth vs. the Capex Time Bomb

You'll find a lot of headlines right now saying Bill Ackman's Pershing Square "boosted" its Amazon stake by 8.6 million shares. The number is real. The verb is not.

Eight point six million shares is Pershing Square's total Amazon position as of the end of June — not a recent addition. The fund actually trimmed roughly a quarter of its stake last quarter. What it did do, though, was hold on to a concentrated bet: 8.6 million shares, worth over $2 billion, representing about 10.5% of a $19.5 billion portfolio. That is not the behavior of someone losing conviction. It's the behavior of someone who sees a specific payoff path and is willing to wait for it.

So the real question isn't what the headline says. It's what Ackman is betting on inside Amazon, and whether that bet still holds at today's price.

Amazon trades around $257 now, roughly 7% off the $272 level it hit in late July after reporting second-quarter earnings. The stock is up 11% year-to-date. Ackman started buying in April 2025 around $198. His fund has been up approximately 25% on the position. Not bad. But the story inside the numbers has changed significantly since he began accumulating.

The engine is Amazon Web Services. When Ackman started buying, AWS revenue growth was running around 17% year-over-year — solid, but not electrifying. In the most recent quarter, AWS revenue grew 37%, the fastest pace in 18 quarters. That puts AWS on a $169 billion annualized revenue run rate. AWS operating income jumped 64% to $16.6 billion, with an operating margin of 39%.

More than the growth rate, it's the backlog that tells the story of what's coming. Remaining performance obligations — the revenue Amazon has contracted but not yet earned — hit $496 billion as of June 30, up from $195 billion a year ago. The weighted-average remaining contract term is 6.4 years. Two deals alone account for much of that: OpenAI expanded its AWS commitment by $100 billion over eight years, and Anthropic committed to spend more than $100 billion on AWS over a decade. Ackman also expects AI inference workloads to fill the new capacity as it comes online, earning attractive returns over time.

This is where the picture gets complicated. To handle all this demand, Amazon is spending money at an extraordinary rate. The company raised its 2026 capital expenditure plan to approximately $220 billion, up from an earlier estimate of $200 billion. The increase reflects higher memory costs, not additional capacity. Cash capex in the second quarter alone was $53 billion.

The consequence: free cash flow turned negative. For the trailing twelve months, Amazon burned through $7.6 billion, compared to generating $18.2 billion a year ago. That's not a blip — it's a structural feature of the investment cycle. Amazon is building data centers that cost billions and take roughly two years before they can host paying customers. Servers and networking equipment purchased months before deployment have a payback period of less than three years, with a useful life of five to six years.

This is the tension at the center of Ackman's bet. He expects Amazon to compound earnings at more than 20% annually, and the backlog suggests the demand is there. But between now and when those data centers come online and generate the cash that pays for them, free cash flow looks ugly. Ackman's answer is that the market is looking at the wrong denominator. Operating cash flow for the trailing twelve months was $161 billion — strong and growing 33% year-over-year. Free cash flow is negative only because capex exceeds operating cash flow during this construction phase, not because the underlying business is generating less cash.

Amazon also nearly doubled its long-term debt in the first half of the year, from $65.6 billion to $128.9 billion, drawing $67 billion in proceeds across four debt issuances. The balance sheet remains manageable — total debt sits at roughly $544 billion against $78 billion in cash and $552 billion in equity, with a debt-to-equity ratio of 0.23. But the leverage move signals that Amazon is financing this cycle through both debt and its own operating cash, and the payoff timeline is measured in years, not quarters.

Then there's the retail business, which gets less attention than AWS but matters for the total company. North America revenue grew 16% to $116.2 billion, with operating income of $9.1 billion. Advertising — a high-margin business running at a $76 billion annual revenue rate — grew 26%. But a metric called the shipping spread flipped for the first time in 15 quarters: shipping costs grew 19% while paid unit growth was only 17%. That inversion, driven by fuel inflation and line-haul rate increases, ended a long streak of logistics efficiency gains. Whether it's a one-quarter blip or the start of margin pressure depends on whether Amazon can pass costs through or find automation savings fast enough.

Against peers, Amazon's valuation tells an interesting story. The stock trades at about 20 times trailing earnings, below Microsoft at roughly 27 times and Meta at 23 times. On a forward earnings basis, though, Amazon sits at about 39 times — the highest in the group. The forward multiple is elevated because the market expects the earnings spike from the Anthropic investment gain to fade and because capex is currently depressing free cash flow. An EV/EBITDA multiple of 16.4 is in line with the peer average. On a price-to-sales basis at 3.6, Amazon is the cheapest of the group by a wide margin.

Here's how to frame the decision. Ackman's bet rests on three linked propositions: AWS growth acceleration continues into at least 2027 and 2028 (the backlog suggests it will); the $220 billion in capex builds capacity that AI demand fills before the servers lose economic value; and the retail business maintains its margin trajectory despite the recent shipping spread inversion. If those propositions hold, Amazon compounds at the rate Ackman projects, and the current price — down from its post-earnings high — offers a reasonable entry.

The bear case is also concrete. Capex of $220 billion in a single year is enormous. If AI demand growth slows before the data centers are fully deployed, that capital sits idle while the debt service continues. Free cash flow negative for another year or two would keep the stock under pressure, particularly for investors who price using free cash flow rather than operating cash flow. The shipping spread inversion, if persistent, chips away at the retail margin improvement that has been a steady source of earnings growth. And a forward P/E of 39 means the stock is already pricing in significant earnings growth — a miss on AWS growth or a capex overrun would compress the multiple.

Amazon reports third-quarter earnings in late October or early November. That's the next proof window. The market will want to see whether AWS growth held above 30%, whether capex spending is tracking the $220 billion plan, and whether the shipping spread returned to efficient territory. Ackman has already shown he's willing to trim the position for portfolio reasons while keeping conviction intact. The 8.6 million shares he still holds — and the 10.5% of his portfolio they represent — say he believes the payoff comes well before the risk matures.

Whether that belief turns out to be right depends on one mechanical question: does AWS growth justify the construction spending in the two to three years before the data centers turn cash-positive? The backlog says yes. The free cash flow says wait and see. Ackman has clearly chosen a side.