Amazon

On this evidence Amazon sits outside your circle of competence, and the price is above base-case value.

Amazon (AMZN) earns money from its online stores, fees from third-party sellers, the AWS cloud business, advertising and the Prime subscription. Over time it has earned high returns on equity, averaging 20%, and it carries little debt, with net cash of $53.7B. Quality is less clean than those averages suggest: there was 1 loss year, operating margin has moved by 3.1% around its average, the latest return on invested capital fell to 18%, and reported profit includes large investment gains. The engine places Amazon outside your circle of competence because the cash the business returns depends on whether heavy new spending earns a return, and the filing admits that is uncertain: "we may not be successful enough in these newer activities to recoup our investments in them." At $250.00 the share price sits above the base-case value of $105.67, with a bear case of $62.28 and a bull case of $167.08.

Amazon sells products it buys and resells, and it earns fees when third-party sellers use its stores and fulfilment services. It also sells cloud computing through AWS, sells advertising, and sells subscriptions such as Prime. The filing says product sales are recorded gross, while seller sales are recorded as Amazon's net share, which affects how revenue looks from one year to the next.

• Unit sales growth in North America and International stores
• Third-party seller fees and fulfilment services
• AWS customer usage, partly offset by pricing set in long-term contracts
• Advertising sales across stores, video and devices
• Subscriptions such as Prime, and the cost of shipping offers that support them

Inside the circle? Amazon's retail, seller and advertising model is easy to follow, and it produced positive owner earnings in most years, with 1 loss year across the period. But capital spending has risen far above depreciation, and the filing says the returns on new bets may not be recovered. Reported net income also includes large gains on investments, so a ten-year estimate of owner cash is hard to make with confidence.

Amazon's protection appears to come from scale across its marketplace, fulfilment network and cloud business, and from the number of products and sellers on its sites. The filing describes hundreds of millions of products sold across dozens of categories, which gives buyers and sellers a reason to stay. Returns on capital are a mixed cross-check: 36% on average over ten years, but 18% in the latest year and a low of 5.4%, so the advantage has not always produced high returns on new spending. The company says it chooses low prices, so the filing does not show the pricing power that Buffett treats as the most important test. The filing also says AI and online comparison make it easier for new and smaller rivals to enter, so the moat looks narrow rather than durable, and the filing gives no evidence on how long it will last.

Management has kept most of its cash inside the business rather than returning it to owners. The dividend payout average is 0.0%, and the share count has still grown by 1.3% a year, so owners have been diluted rather than helped by buybacks. Capital spending runs at 1.3× depreciation, and the filing expects technology spending to keep rising. Reported profit is also lifted by investment gains, which makes owner cash harder to read than net income suggests. The filing is detailed about risks and one-off charges, but it relies on phrases like 'we believe' and says little about what returns the new bets will earn.

• Share count change averages 1.3% a year, so equity pay and share issuance are adding to the number of shares owners hold a claim on. The filing says it manages dilution, but the count has not fallen.
• Estimated maintenance capital spending is $110.7B, against normalised owner earnings of $38.7B. Most of the cash the business produces is going back into the business.
• The filing ties most of the other income to a gain on the Anthropic investment, so part of reported net income ($77.7B) comes from an investment rather than from retail or cloud operations.
• The balance sheet gives management room to keep investing: $53.7B of net cash and 0.2× debt to equity. Financing is not the constraint on the spending plan.

• FailsRational capital allocationFails

  The diluted share count has changed by 1.3% a year. Shrinking means buybacks are returning value; growing means owners are being diluted.

  Why this test
• FailsMargin of safetyFails

  At $250.00, the price is above the base-case value of $105.67 (the bear case is $62.28), so there is no margin of safety (-137%). The standard lens asks for 25%.

Buffett's first test is whether you can understand a business's economics and estimate its future, and the evidence here does not give you that confidence for the next ten years. Consistent earning power, which Berkshire's acquisition criteria ask for, fails on the loss year and the swings in margin. Pricing power, which Buffett called the most important test, is unclear: the filing says "We believe that offering low prices to our customers is fundamental to our future success," which shows a company that uses low prices to win customers and does not show a record of raising them. Owner earnings, the cash an owner could take out after capital spending, are harder to read: capital spending runs at 1.3× depreciation, and owner earnings were $32.7B in the latest year against a normalised figure of $38.7B. The one dollar retained test passes, with 9.6× of market value created for each dollar kept, and the low debt tests pass. The rates test fails: the owner earnings yield of 1.4% is well below the 5.1% 10-year Treasury yield. Capital allocation also fails, because the share count has grown by 1.3% a year, so owners have been diluted rather than helped by buybacks.

• Capital spending compared with depreciation, currently 1.3×, in the next annual report

  A ratio that stays high means more of the cash goes back into the business before any owner can take it out.

• Return on invested capital in the next year, compared with the latest 18%

  The filing says returns on new spending may not be recovered, and a further fall would show that risk arriving.

• Gains from investments, such as the Anthropic stake, inside other income

  Reported net income of $77.7B includes these gains, so owner cash is a better guide to earning power.

• Change in the share count, currently 1.3% a year

  Buybacks would show management returning cash to owners, which is the behaviour Buffett looks for when the price is below value.

• One-off charges, such as the Federal Trade Commission settlement and severance costs booked in the latest year

  Repeated one-offs would weaken the case for consistent earning power, and the filing gives no sign yet of how often they recur.

Questions the filings couldn't answer

• What return will the spending on AI infrastructure and the Amazon Leo satellite network earn, and when? The excerpts give no figures for this.
• The excerpts do not give segment profit history in numbers, so AWS and retail margins cannot be separated with the evidence here.
• How long can the moat last if AI lowers the cost of comparison shopping and entry for smaller rivals? The filing names this risk but gives no measure of it.