Amazon built a bigger network, and now its profit swings harder

A good quarter for sales, a far better one for profit

Most people meet Amazon as a cardboard box on the doorstep. Fewer think of the company behind it as a giant fixed-cost machine: fulfilment centres, delivery vans, data centres and the people who run them. In the quarter from July to September 2026, Amazon.com, Inc. took USD 202bn in revenue, up 12.1% from USD 180bn a year earlier. The figures cover the whole group, under US accounting rules (US GAAP), in US dollars. That is USD 21.8bn of extra sales in a single quarter. Solid growth for a company this size, but not a revolution.

Operating profit tells a louder story. It is what remains after paying for the goods, the shipping, the servers, the staff and everything else it takes to run the business, before interest and tax. It rose from USD 17.4bn to USD 28.1bn, an increase of 61.2%. Profit grew 5.06× as fast as revenue.

The operating margin, the share of each dollar of sales that ends up as operating profit, widened from 9.7% to 13.9%, a gain of 4.2 pp. Put simply: the same customer spending the same dollar left Amazon noticeably more than a year ago. The question is why, and whether that should be read as a permanent upgrade.

A warehouse costs the same on a quiet Tuesday

The mechanism is old and unglamorous. Some of Amazon's costs move with every order: the goods it buys for resale, the fuel and labour for each parcel, payment fees. Many do not. A fulfilment centre costs roughly the same whether it ships a record number of parcels or a modest one. A data centre consumes power and depreciates whether its servers are busy or idle. Software teams, management and the delivery network are paid for in advance.

Once revenue covers that fixed base, most of every additional dollar falls through to profit. The quarter shows it plainly. Of the extra USD 21.8bn in sales, USD 10.7bn turned up as additional operating profit. For a retailer that typically keeps only cents on each dollar, that is an unusually large slice of the increment.

What makes this more interesting is that the fixed base itself got heavier. Depreciation and amortisation, the cost of buildings, servers and equipment spread over their useful lives, rose from USD 16.8bn to USD 20.9bn. Amazon is carrying a much larger stock of assets than a year ago: total assets went from USD 728bn to USD 1,243bn. Normally, a bigger cost base eats into profit. Here profit grew anyway, which means the extra capacity was filled fast enough to pay for itself within the quarter.

EBITDA, operating profit with depreciation added back, shows the same pattern from another angle. It rose from USD 34.2bn to USD 49.0bn, and its margin widened from 19.0% to 24.3%. Strip out the cost of the machines and the business underneath is earning more on each dollar too.

The group-level figures do not say which part of Amazon delivered the extra profit: the online shop, the cloud business or advertising. That matters, because advertising and cloud services carry far fewer variable costs per dollar than shipping a parcel. The segment results would settle it.

Below the operating line, a quieter picture

Net profit, the bottom line after interest, tax and everything else, did not follow. It came to USD 20.7bn, down 2.4% from USD 21.2bn. The net margin slipped from 11.8% to 10.2%. I would not read that as weakness. A year earlier, net profit sat above operating profit, which means gains from outside the core business lifted it. This year the bottom line is earned closer to the shops and servers, which is the healthier kind of profit even if the headline is smaller.

Cash went the other way from profit as well. It fell 16.6% to USD 55.8bn at the end of September 2026. For a company growing its asset base this quickly, money spent up front on warehouses and data centres is the usual suspect; the cash flow statement would show how much of it went into new capacity. Equity rose to USD 614bn, and the equity ratio, the share of assets funded by owners rather than creditors, stood at 49.3%. That is not a stretched balance sheet.

The lever points down as easily as up

For customers, this explains a lot about Amazon's behaviour. A network this expensive wants to be full. Faster delivery, Prime perks, more third-party sellers and more ads on product pages all push more volume through buildings and servers that are already paid for. Every extra order in a half-empty warehouse is close to pure profit, and the company knows it.

For the market, it widens the gap with smaller rivals. A competitor without that scale pays similar fixed costs on far fewer parcels, so it cannot match the same prices or delivery times without losing money. Scale here is not a slogan. It is a cost structure.

The flip side is the one worth keeping in mind. Fixed costs do not shrink when orders do. Depreciation on the new data centres will arrive every quarter whether demand holds up or not. The same lever that turned 12.1% of sales growth into 61.2% of profit growth would turn a modest drop in revenue into a sharp drop in profit. Investors already seemed nervous: Amazon shares fell at the end of October 2026 as part of a broad sell-off in US technology stocks. The quarter's accounts themselves show a business firing well; the share price reflects other worries.

For anyone running a business with a heavy fixed base, whether a warehouse, a gym or a hotel, the same check sits in your own income statement. Split your costs into those that move with sales and those that do not, then work out how far revenue could fall before operating profit reaches zero. The answer is usually less comfortable than a good year suggests.

Amazon filled its new capacity faster than it paid for it, and the third quarter of 2026 shows what that does to profit. It did not become a lower-risk business in the process. It became a bigger lever, and levers are indifferent to direction.