Amazon: A Wonderful Business

Not Yet a Wonderful Investment

Over the past week I spent time diving deeply into Amazon. It began as a modest position in my portfolio - one I owned without fully underwriting expected future returns. That bothered me. So I stepped back and rebuilt the thesis from first principles.

Amazon is clearly a phenomenal business. The question is not quality. The question is price relative to future returns.

Under a disciplined bull case, we assume AWS compounds at 15% annually for ten years, growing from roughly $142B to more than $575B in revenue while maintaining a 35% operating margin. We assume Retail, including advertising, compounds at 8–12% annually with blended margins of 6–8%, recognizing that advertising mix shift could expand margins while aggressive international and grocery investment could compress them. Under those assumptions, total operating income could exceed $300B in a decade.

Applying a 25× multiple implies a valuation of approximately $7–8T. From today’s ~$2.15T market capitalization, that equates to roughly 13–14% annualized returns. Applying a 20× multiple instead results in a valuation closer to ~$6T, implying approximately 11% annualized returns. Even under optimistic operating assumptions, expected returns from today’s price fall in the 11–14% range.

That is good. It is not extraordinary.

The swing variables are Cloud/AI TAM and Amazon’s structural cost advantage. If Cloud TAM expands toward $4–5T+ over the next decade and Amazon sustains durable cost curve superiority through custom silicon and scale, those return outcomes are achievable. If TAM stabilizes closer to $2–3T or cost advantages narrow due to competition, returns likely compress below those levels.

The work to gain conviction here is not in the historical financials - it is in understanding the demand side and the economics of infrastructure. How fast are enterprises actually deploying AI into production? What does inference cost trajectory look like? Is sovereign AI fragmenting or expanding the market? Does Amazon’s custom silicon create a lasting cost advantage, or will it be competed away? Observable confirmation would include sustained AWS operating margins above 35% while absorbing rising depreciation, clear Trainium and Graviton adoption indicating structural displacement of merchant silicon, and enterprise AI deployment rates showing TAM expansion beyond consensus.

At present, I do not possess differentiated insight into either of those two critical variables sufficient to justify paying full price. That means I am at risk of a Type I error - allocating capital at fair value without edge. Therefore, I require either a materially lower entry price (approximately $150-160 for a 12% hurdle, lower for higher required returns) or objective evidence of durable cost leadership and sustained ROIC expansion.

This is not a question of admiration. Amazon is extraordinary.

It is a question of arithmetic.

And arithmetic, not enthusiasm, determines allocation.