Amazon closed at $230.86 on July 28, down 17% from its May high of $278.56, with second-quarter earnings due the following afternoon. The headline numbers still look strong: AWS grew 28% in the first quarter, its fastest pace in fifteen quarters, and advertising compounded 24% to a $72 billion annual run-rate. The rating is Watchlist, and the reason has little to do with doubting the business.

The first issue is what the reported earnings actually contain. Trailing twelve-month EPS of $8.36 puts the stock at a seemingly cheap 27.6 times earnings, the lowest multiple in five years. But roughly a quarter of that net income is a non-cash mark-to-market gain on Amazon’s stake in Anthropic, an $8 billion investment now carried above $70 billion on private-round reference prices. Strip that gain out and core EPS falls to about $6.22, putting the real multiple at 37.1 times, the most expensive of the three major hyperscalers against Microsoft near 21 times and Alphabet near 24 times. Much of the bullish commentary citing Amazon’s “cheap” valuation is anchored on the wrong number.

The second issue is capital expenditure. Spending of roughly $330 billion across 2025 and 2026 will add an estimated $55 billion of annual depreciation on six-year server lives, an expense that is contractually certain regardless of demand. Trailing free cash flow has already turned negative for the first time since 2022, down from $32.9 billion two years ago. The arithmetic works if AWS keeps growing near 30%; at 20% growth, the shortfall costs roughly $13 billion of operating income, about 14% of the consolidated total. The first 140 basis points of that margin pressure already showed up in the first quarter.

None of this touches the moat. Prime, marketplace density, and custom Trainium silicon remain genuinely difficult to replicate, and a $364 billion AWS backlog, excluding a separate Anthropic agreement worth over $100 billion, underwrites years of growth even if new bookings slow. A probability-weighted fair value near $235, built from a 25% bull case at $304, a 50% base case at $246, and a 25% bear case at $143, sits barely above the current price. That is not the asymmetry a buyer wants heading into a binary print, with the stock trading below every major moving average and a bearish MACD crossover in place.

The path to a better entry is specific: a pullback into $196-210 would price in most of the bear case, or a Q2 print showing AWS growth at 30% or better with margin defended above 36% would resolve the question in the bulls’ favor.

The complete scenario model and segment breakdown are available in the full Amazon (AMZN) report on our Reports page.