Fate of Reality

August 31, 2026

The AI Interest Rate Switch

An Apollo economist says long rates could fall hard in the next six months whether AI succeeds or fails. Both roads lead down. What differs is everything else.

The facts:

  • Torsten Slok, chief economist at Apollo, says the risk is rising that long-term interest rates are "a lot lower" six months from now (Apollo Daily Spark, via Brad DeLong)
  • His chart: the 30-year Treasury yield falling from 5% back toward 3% (DeLong)
  • The mechanism if AI succeeds: tech companies generate trillions in revenue, AI is "massively deflationary," rates get pushed lower (Slok)
  • The mechanism if AI fails: the bubble bursts, the Nasdaq drops 50%, investors rotate out of stocks into Treasuries, long rates fall dramatically (Slok)
  • His claim about timing: "over the next six months, the market will make up its mind about which AI scenario is playing out" (Slok)
  • Today's market narrative is inflation and deficits; Slok says the narrative is about to become the success or failure of AI (Slok)
  • DeLong's caveat, quoting the old rule: "never make large, directional predictions of interest rate movements" (DeLong)

Interest rates have been a Fed story for as long as most people have been paying attention. Inflation, meetings, a guy at a podium. So a serious economist saying the next big rate move depends on whether AI works is the kind of claim that stops the scroll.

The strange part of Slok's argument is that both branches point down. AI delivers, productivity explodes, that's deflationary, rates fall. AI flops, the Nasdaq craters, everyone hides in Treasuries, rates fall. Heads or tails, cheaper money.

But the two versions of "cheaper money" aren't the same country. One is a boom where borrowing is cheap because things are getting more efficient. The other is a recession where borrowing is cheap because half your retirement account is gone and nobody wants to build anything. Same headline. Very different Tuesday.

the six months thing

DeLong flags the obvious gap: why six months? Slok doesn't really say. There's no scheduled event where AI gets graded. No earnings date settles "did it work." Which suggests his real claim isn't about AI at all, it's about attention. Markets run on narratives, he says so himself, and he's betting the narrative flips from inflation to AI by early 2027.

Maybe the market picks a winner in six months. Or maybe it does the thing markets do when a question won't resolve: chop around, punish everyone, and wait. Trillions are riding on a question that doesn't have a deadline, and that's a recipe for a long ugly limbo, not a clean verdict.

why this one matters

Long rates aren't an investor thing, they're gravity. They set what a mortgage refinance costs, what a car loan costs, whether the shop down the street can afford to expand into the empty unit next door.

Worth remembering DeLong's own warning as he passes the prediction along: never make large, directional predictions of interest rates. He compares it to a land war in Asia. Then he spends the whole post taking it seriously anyway, because the argument is that good. Watch for the moment the financial news stops leading with inflation prints and starts leading with the AI trade. Slok's real prediction is that the switch is coming. The direction after that is the part nobody gets to know yet.

Source: Brad DeLong, Grasping Reality, quoting Torsten Slok (Apollo).