Why Meta and Amazon Tanked on AI Capex — and Why I See It as a Buying Opportunity

Meta dropped 25–30% from its all-time high. Amazon slid from $240 down under $200. Neither dropped because the business broke. Both dropped for the same reason — they announced they're pouring tens of billions of dollars into AI infrastructure.

That pattern is exactly why I added to both positions.

What Actually Happened

Meta's most recent quarter can be summarized in one line. Annual capex up roughly 40% year-over-year. Operating margins compressed by the same magnitude. Short-term earnings momentum slowing.

The market reacted fast. After putting in a 2025 all-time high, Meta sold off 25–30%, then kept drifting lower. I made my first buy in that initial drop and doubled down further on the way down.

Amazon's picture was nearly identical. The stock fell from around $240 to below $200. Two things came up at once. First, like Meta, guidance for tens of billions in AI infrastructure and data center spend. Second, AWS growth decelerating slightly versus expectations.

The combined narrative — short-term margin pressure plus a slowdown in the most profitable segment — produced the simple "sell" reflex.

What the Market Missed: Capex and Moat

The reason I treated this drop as a buying opportunity comes down to one question. Is the capex going up because the business is breaking, or because the business is widening its moat?

Meta's spend is firmly in the second category. The AI infrastructure — data centers, GPU clusters, in-house AI silicon, recommendation training pipelines — once built, becomes an asset that improves the unit economics of the core business: ad targeting precision, Reels recommendation quality, AI features in Messenger and WhatsApp. Infrastructure laid down once produces revenue every year afterward without needing to be re-spent.

Amazon's setup is structurally identical. AWS is scaling capex into a moment when AI training and inference demand is exploding. That's a leading indicator of future revenue. The dollars going in now are determining the AWS revenue curve for the next 5 to 10 years. Margins look squeezed in the short term because the cost is recognized before the revenue catches up.

The accounting timing is the asymmetry. Costs land immediately on the income statement; the revenue from those costs lands across years. That mismatch makes short-term P&L look bad while long-term value gets stronger. The window where part of the market reacts only to the short-term P&L is the buying opportunity.

Where This Logic Could Break

The "capex equals widening moat" equation isn't automatic. Two scenarios break it.

First, if the capex doesn't actually convert into revenue. Build out AI infrastructure and the targeting improvements never show up in ad ARPU? That's just sunk cost. Second, if the capex cycle never ends. If the AI arms race runs longer than 5 years, the expected "margin recovery next year" gets pushed out indefinitely.

The way to track these is two metrics per quarter. Are AI-related revenue or user metrics actually rising in line with the capex? — for Meta, ad ARPU and Reels engagement; for Amazon, AWS revenue growth. And is capex guidance being revised up further, quarter over quarter?

Through the data so far, the first condition is being met. Meta's ad ARPU has been steadily improving since the AI recommendation rollout. AWS's absolute revenue base keeps expanding. The second condition is still open. If next quarter's capex guidance gets revised up again, the buy thesis needs reassessing.

Buy Discipline: Where Price Meets Fundamentals

The buying rule I used on both names is simple.

For Meta: 25–30% off the all-time high was the first entry, further drawdowns were the doubling-down level. Compare the PE at the high to the PE now, and the same business is essentially being marked down 30% on the back of one capex guidance line. Unless the underlying business is 30% worse, the price is wrong.

For Amazon: I bought on the move below $200. As of recording, the position is up over 15% year-to-date from that entry. AWS isn't dead — its growth rate just came in slightly below expectations. A company with a sprawling retail business and a fast-growing ad business growing alongside that, trading under $200, looked like short-term noise.

Both names share the same setup. The competitive moat is intact or widening, only the short-term margin is compressed, and the market is reacting to the short-term P&L only. That combination is the simplest buy scenario you can find.

Signals to Watch Next

Three things to track in the next earnings prints. First, does capex guidance get revised up again? Another raise validates the bear case. Second, do AI-related revenue or usage metrics come through? — Meta's ad ARPU and Reels time, Amazon's AI workload share inside AWS. Third, where do operating margins find their bottom? Distinguishing a one-off compression from a structurally lower new equilibrium matters a lot.

If all three come in negative simultaneously, the entry thesis needs to be reworked. If two of three improve, the market will probably start re-rating these names higher again.

FAQ

Q: Between Meta and Amazon, which is the better entry? A: Depends on time horizon. Amazon is more diversified, so short-term volatility is lower and AWS revenue visibility is stronger. Meta's ad business has higher margins, but you have to verify quarter by quarter that AI capex ROI is showing up in ad ARPU. For stability, Amazon. For upside, Meta.

Q: What if capex just keeps going up forever? A: That's the largest risk. If the AI arms race lasts more than 5 years, the "margin recovery" expectation gets pushed out every year. In that scenario, narrow your selection to companies where revenue growth dominates the capex burden. Amazon is stronger than Meta in that scenario because of the diversification.