Ran Amazon through my valuation model. It flagged it as a compounder and then refused to buy it.

Quick context first for anyone who needs it. Amazon is really three businesses stapled together: a retail marketplace, a logistics network, and AWS, the cloud infrastructure arm where most of the profit actually comes from. Advertising has quietly become a fourth. Retail runs on thin margins and volume, AWS runs on high margins, and the consolidated statements blend all of it, which is part of why single ratios are awkward with this company.

I run a model that pulls straight from SEC filings and puts a company through three gates before it can be called a buy. No forensic vetoes, business quality above 70, and a positive adjusted margin of safety. All three have to open.

Amazon is the strangest result I've had.

THE QUALITY SIDE

On quality it does almost everything right.

Three year average ROIC of 21.7%, against an 8% hurdle. Gross margin of 50.3%. Operating cash flow running 1.8x net income, so the profit is genuinely showing up as cash. Accruals at -8.6% of assets, which is negative, and negative accruals are the opposite of the pattern you see when someone is inflating earnings. Beneish M-Score of -2.61, well inside clean territory.

Net cash of $54.2bn excluding lease liabilities. Including the $89.3bn of leases it's net debt, and my solvency sheet deliberately keeps both definitions because they answer different questions.

My compounder flag came back true. That requires persistent high ROIC, low leverage and strong cash conversion all at once, and it doesn't fire often.

Business quality score: 69. Threshold is 70.

THE MOAT

The model scores it 75 out of 100, and almost all of that is quantitative rather than my opinion.

The spread between ROIC and my hurdle is 11.3 points, so capital going in earns well above what I'd demand of it. The cash conversion cycle is minus 51 days, meaning Amazon collects from customers roughly seven weeks before it pays suppliers, so growth funds itself rather than consuming cash. Gross margin has a standard deviation of 2.8% over four years, which is stable for a business this mixed. And the three year minimum ROIC is 18.8%, so the returns aren't a one year artifact.

The qualitative half I left at neutral across all six inputs: network effects, switching costs, brand, scale, cost advantage, intangibles. I score those by hand and I'd rather report a 3 than pretend I've done work I haven't. So the 75 is the demonstrated moat, not the argued one.

THE VALUATION GATE

Weighted intrinsic value came out at $57.62 a share against a price of $226. Bear case $23, base $55, bull $124.

Even the bull case sits 45% below where it trades.

That gap is wide enough that it needs explaining rather than just reporting, because the answer is the same thing I've been writing about all week.

Amazon's free cash flow went from $32.9bn in FY2024 to $7.7bn in FY2025. Capex was $131.8bn. A discounted cash flow model built on free cash flow is going to produce a low number when the free cash flow base has collapsed, and mine has no way of knowing whether that collapse is temporary or structural.

To its credit the model noticed. It classified Amazon as TRANSFORMACION, which is my regime for a business spending far above its own historical capex while still growing, and it downgraded DCF applicability from full to adjusted. In plain terms, it flagged that its own valuation is less trustworthy here than usual.

RISKS

Two of these came out of my own scoring rather than from a narrative.

Share count is up 4.8% over three years. That's the weakest item in capital allocation and it scored 2 out of 10. Amazon is diluting while it builds, which matters because everything above is per share eventually.

Free cash flow covers only 0.49x of short term debt. That scored 1 out of 10 and it's the worst single number in the whole model. It isn't a solvency problem given the cash position, but it does mean the buildout has left very little cushion between a year of cash generation and what falls due inside twelve months.

Altman Z sits at 1.87, technically the grey zone. I don't weight that heavily because Altman was calibrated on mid twentieth century industrials and misfires on asset heavy modern businesses, but it's there.

Beyond the numbers: AWS losing share to Azure and Google Cloud, regulatory pressure on the marketplace in both the US and EU, and the possibility that this capex is buying capacity that arrives after the demand has already been captured somewhere else.

WHAT I ACTUALLY THINK

I don't think the honest conclusion is "Amazon is 75% overvalued".

I think it's that a cash flow model can't value a business in the middle of a buildout, and anyone showing you a confident number on Amazon right now is hiding that problem rather than solving it. Mine included, which is why I'm showing you the number and the reason not to trust it in the same post.

The question underneath is simple to state and impossible to answer today. At $226 the market is paying for free cash flow that doesn't exist yet. Whether it shows up depends on what $131.8bn a year of capex eventually earns, and nobody knows that, including Amazon.

What I'd want before believing either side is a few quarters of capex flattening while AWS margins hold. Until then it's a bet on the buildout, not a valuation.

All figures from Amazon's 10-K filings. Price used is $226.