Don't Buy Amazon Stock Until You Read This

Amazon plans $200B of capex in 2026 (up from $131B in 2025), largely to build AWS and generative-AI data-center capacity. Free cash flow is expected to fall sharply, with consensus projecting about a $10B FCF loss in 2026 and only $1.2B FCF over the past 12 months (down 95% y/y). The spending is partly funded with external capital—Amazon has raised $80B+ in debt in 2026—raising the bar for faster returns on AI investment.

Analysis

This is a capital-intensity story, not a pure growth story. When a mega-cap platform moves from “self-funding expansion” to debt-assisted buildout, the market usually re-rates the equity on FCF yield and buyback capacity before it re-rates on eventual revenue upside. In the next 1-3 months, the risk is not demand collapse; it is multiple compression if investors conclude the AI spend is front-loaded while monetization is back-end loaded.

Second-order winners are the picks-and-shovels suppliers to hyperscaler buildouts, especially NVDA, where incremental capex still converts into unit demand even if the buyer’s equity suffers. The cleaner relative trade is that suppliers can compound on volume while AMZN absorbs depreciation, interest expense, and working-capital strain. Competitively, this also gives MSFT/GOOGL-style peers a talking point around capital discipline: if AWS is forced to fund growth externally, rivals can frame themselves as higher-quality cash compounding machines.

The key risk is that the market is underestimating duration: if AWS backlog is real and utilization ramps quickly, the FCF drawdown may be a temporary accounting bridge rather than a value leak. What would falsify a bearish AMZN view is any sign that capex growth slows, debt issuance moderates, or AWS margin expansion offsets depreciation within the next 2-4 quarters. Longer term, the thesis breaks if AI infrastructure becomes a structurally higher-ROIC layer and AMZN can sustain that spend without a persistent spread widening in its debt market.

Contrarian view: the move may be only moderately negative because the street already knows Amazon can absorb periods of negative FCF. The more underappreciated issue is opportunity cost — every extra dollar spent on data centers is a dollar not available for buybacks, and that matters more in a market that rewards cash yield again. If investors start demanding proof of returns rather than promises, AMZN could lag even if fundamentals remain strong.

Key Decisions for Investors

• Long NVDA / short AMZN as a 1-3 month relative-value pair: express the view that hyperscaler AI capex accrues to the supplier faster than it accrues to the spender. Risk/reward favors the pair if Amazon’s next update keeps FCF negative while chip demand remains firm; thesis fails if AMZN shows faster-than-expected monetization or NVDA guidance softens.
• Avoid initiating new outright AMZN longs until the market gets evidence of FCF stabilization over the next 1-2 earnings cycles. For existing holders, use strength to trim rather than add; the near-term hazard is valuation compression driven by external financing and slower buyback capacity.
• If willing to own AMZN downside, use 3-6 month put spreads rather than outright shorts. The trade works if bond investors begin pricing higher funding costs or if management signals another step-up in 2026 capex; it is invalidated by a clear capex slowdown or an upside surprise in AWS margins.
• Watch AMZN credit spreads and debt issuance cadence as the real-time catalyst. A meaningful widening over the next several weeks would confirm the market is starting to discount execution risk and longer amortization periods for the AI buildout.