The better AI gets at saving you money, the more your mortgage costs
American mortgages are cheap partly because most people can't be bothered to refinance when rates fall; once AI agents remind everyone to refinance on time, that option gets repriced and mortgage rates go up instead.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
AI reminders to refinance will make mortgages more expensive
American homeowners mostly sit still when rates fall: Matt met someone on a plane who volunteered their own mortgage rate and was about to sell and trade up to a bigger house, and he talked them out of it on the spot. Morgan Stanley research argues that if an AI agent prompts you to refinance whenever rates drop 200 basis points, refinancing will get far easier in future. That sounds like it saves you money, but more frequent refinancing means the prepayment option people previously "couldn't be bothered to exercise" actually gets exercised, lenders reprice for it, and everyone's mortgage rate rises by 10 to 20 basis points.
— Matt LevineThe smarter homeowners get, the fewer people can buy a home
Fabrice Touré, formerly head of mortgage derivatives at Goldman Sachs and now an academic in finance, puts it bluntly in a paper: policies that make refinancing more frequent also push up the equilibrium mortgage rate, and cut off the chance of getting credit for a large number of borrowers. The reason is that the American 30-year fixed-rate, prepayable mortgage carries an expensive option built into it, but most people exercise it badly, which is why lenders are willing to hold the rate down. Once regulation or AI makes everyone more rational, the option gets repriced, rates rise, and the pool of people that the subsidised low rate used to cover shrinks, so more households cannot afford a house.
— Matt LevineModels priced on past behavior are walking into a landmine
The shock hits investors first. Morgan Stanley's warning: historical prepayment behavior will understate future prepayment, so any model trained on past data has mispriced the fact that borrowers are becoming more willing to exercise the prepayment option. The result is that some mortgage-related assets look cheap when they actually rest on the assumption that borrowers stay lazy; once AI compresses refinancing friction from two weeks to ten minutes, all of these assets need to be revalued wholesale. Matt extends the logic: credit card points, high-yield savings accounts and life insurance policies are all the same, and a large part of the financial system runs on customers not pushing every option to its limit.
— Matt LevineAI stops the billable hour from working as a business model
AI's squeeze on professional services shows up first at law firms: clients believe AI can do the work of a row of junior associates, and ask for a discount. A firm bills by the hour on the surface, but the way it actually makes money is that partners sell junior associates' hours at a high price and pay those associates a lower wage; if a lot of the grunt work is done by AI, "charging for a junior associate's hours" no longer holds up. A senior lawyer's professional judgement is still worth money, but it needs a different pricing mechanism. The bigger problem is this: when new lawyers no longer have to spend hundreds of hours on basic work, where does this cohort of future partners train up their instincts?
— Matt LevineStar partners can go solo with AI instead of a platform
Since a senior lawyer's value does not have to be cashed out through headcount, the very top people will do this arithmetic first: rather than stay on a big platform supporting a partnership committee and training up a large crop of junior associates, go out and open a boutique with nobody in it but yourself and six AI agents. Chris Kircher, a former Quinn Emanuel partner, has already resigned and founded an AI-native law firm built for exactly that judgement. If enough senior people go solo, the big institutions cannot hold on to them even knowing the apprenticeship model matters: the money-making capacity walks out and the cash-burning training system stays, which makes large firms harder to keep alive.
— Matt LevineAmazon's ad auction hides a bidder in the dark
The FTC and 22 state attorneys general are suing Amazon, alleging its ad auction is not the second-price auction it advertises. In a normal second-price auction the highest bidder pays the second-highest bid; Amazon adds a "soft reserve price", the platform's own valuation of the ad slot, which works like a bidder hidden in the dark. Bid above the soft reserve and the soft reserve becomes the second price; bid below the soft reserve and the winner ends up paying their own highest bid. Amazon's response is that an ad auction is a black box computed in real time to begin with, nobody can know the soft reserve in advance, and advertisers place their orders looking only at return on spend.
— Matt LevineAn auctioneer calling bids at the chandelier is perfectly legal
Finance forbids an exchange from quietly entering buy orders for itself, but art auction houses have a legal version of the performance. The seller sets an undisclosed reserve price, and when the bidding in the room falls short, the auctioneer calls out bids into empty space, walking the price all the way up to the reserve; the term of art is chandelier bid, because they are often looking up at the chandelier while doing it. New York requires the auction catalogue to disclose that the auctioneer may bid on the seller's behalf up to the reserve price, and will not identify which bids came from the seller. Amazon's soft reserve is equally the platform bidding for itself, but it never spelled out the rules in that kind of fine print, and that is precisely the point the FTC will not let go of.
— Matt LevineSwapping one ETF for another dodges the dividend tax
A foreign fund, a Cayman structure for instance, owes withholding tax on dividends it receives from US stocks, which drags on the thin margins of a basis trade. Some in the market use two almost identical S&P 500 ETFs to play the timing gap: BlackRock's IVV and Vanguard's VOO pay their distributions about a week apart. The specific mechanic is to sell IVV before it goes ex-dividend and rotate into VOO, then rotate back when VOO is about to pay. After the ex-dividend date a fund's price falls by roughly the amount of the dividend, so the investor sidesteps the distribution that would be taxed, and earns back the same amount of money by selling high and buying low. This is a tax manoeuvre for institutional traders; a retail investor copying it would generate a large capital gains tax bill, and it is not a personal finance trick for TikTok.
— Matt LevineIn their own words · checked verbatim
many policies that lead to more frequent refinancing also increase equilibrium mortgage rates and thus reduce residential mortgage credit access for a large number of borrowers.
Matt Levine6:09
The best, most crazy thing the U.S. mortgage market does is offer 30-year fixed rate mortgages that are prepayable.
Matt Levine8:12
historical prepay behavior will understate future prepays. So models trained on the past mispriced this option that borrowers will be more inclined to exercise.
Matt Levine10:13
if all the grunt work can be done by AI, then, like, that model doesn't work.
Matt Levine13:16
Nobody buys ads based on like simplified descriptions of the auction mechanics. They buy ads based on like return on ad spend, right?
Matt Levine21:29
This is not something that an influencer should recommend on TikTok.
Matt Levine33:39
Figures
| Morgan Stanley's estimated rise in mortgage rates once AI raises refinancing frequency | 10-20 basis points | 5:08 |
| Number of state attorneys general suing Amazon alongside the FTC | 22 states | 18:26 |
| Gap between the distribution dates of the two S&P 500 ETFs, IVV and VOO | about a week | 30:36 |
Glossary
- second-price auction
- The highest bidder wins, but only has to pay the second-highest bid.
- soft reserve price
- A hidden floor price the platform or seller sets for itself in an auction, equivalent to adding a virtual bidder.
- chandelier bid
- The auctioneer calls bids into empty space on the seller's behalf to push the price toward the reserve; named for the habit of looking up at the chandelier.
- basis trade
- Buying the cash instrument while selling the future to capture the spread between them; in essence, providing funding.
- apprenticeship model
- The pattern where professional services firms have new hires do large volumes of low-level work and gradually train them into senior partners.
- withholding tax
- Tax deducted at source on payments such as dividends, commonly levied on foreign investors.
How to listen
People in mortgages, asset management, law firm management and ad auction operations, plus any founder who cares about how AI removing friction changes the way financial products are priced.
The first four minutes or so of small talk and introductions are skippable; from minute five, where the AI refinancing discussion starts, there is essentially no filler.