So Amazon just announced $200 billion in capex for 2026 to handle AI infrastructure, and naturally the stock dipped in pre-market. I get why people are nervous about this — the whole cloud capacity overhang concern is real. But I've been thinking about something Andy Jassy said on the earnings call that kind of reframes the whole thing.

First, context: AWS just hit a $142 billion annual revenue run rate with 24% growth last quarter. That's the strongest growth in 13 quarters. Pretty solid. And yeah, a lot of that is AI-driven right now.

Here's where it gets interesting though. Jassy said something that stuck with me: customers really want AWS for both core workloads and AI workloads. Not just AI. That distinction matters way more than people realize.

Why? Because if AI demand actually cools off — which is the whole risk people are worried about — Amazon's still generating growth from non-AI projects. They're not putting all their eggs in the AI basket. And on top of that, Jassy mentioned they're monetizing new capacity immediately as they add it. So this isn't some speculative infrastructure build that might sit idle.

This changes how I think about the spending risk. It's not like they're betting everything on AI staying hot forever. They're building capacity that's already getting used across multiple workload types. The infrastructure is paying for itself as it comes online.

I think a lot of people are overweighting the downside scenario here. Yeah, there's always execution risk, but the fundamentals of what AWS is doing actually look pretty solid to me.