AI Agents and the $1.4 Trillion Fee Cut Wall Street Isn't Pricing In

$1.4 trillion a year. Joseph Chalom thinks that's the size of the check AI agents are about to hand back to investors, and it comes straight out of the pockets of the firms that have spent a century collecting it.

The Forecast

Chalom runs SharpLink, and before that he spent years at BlackRock, so he isn't some outsider lobbing grenades at the front door. He laid out the thesis in an X post on Wednesday. His team's model projects that AI agents will eliminate nearly a quarter of global finance fees by 2035, with the resulting investor savings landing around $1.4 trillion annually.

He framed the whole thing as a $4 trillion prize sitting behind a $180 billion blind spot, which is one of the cleaner ways I've seen a sitting public company CEO describe his own industry's revenue line as a rounding error waiting to happen.

The timing is the interesting part. AI agents have gone, in roughly 18 months, from demos that could barely book a restaurant to systems that can compare fund fees, execute a trade and shuffle stablecoins across a chain without a human clicking anything. The plumbing is still ugly. But it exists.

Who Pays

Finance fees are a tax on friction. Asset management, brokerage, advisory, market making, and the 1% a year that plenty of wealth managers still charge to rebalance a portfolio twice and call it a strategy. Take the friction out and the fee has to justify itself.

Advisers will tell you they sell trust and behavior coaching, not spreadsheets, and they're not wrong. A lot of investors will keep paying for a human on the other end of the phone when markets get ugly. But 'nearly a quarter' suggests Chalom's model doesn't give them much credit for that.

So who actually pays for trust when the fee disappears?

I'm not entirely convinced by the 2035 deadline. Fee compression in this business is famously slow. Index funds took roughly three decades to crack the active management cartel, and active managers still run trillions that passive products haven't touched. History suggests otherwise on the speed, if not the direction.

The skeptics will add that agents inherit liability, compliance headaches and fiduciary duty the moment they touch client money, and none of that comes cheap. Fair. Granted, the incumbents have distribution and brand on their side, and those moats have held for a long time.

The bit I'd watch is less glamorous than the model itself. The winners here probably aren't the firms with the smartest agents. They're the ones controlling custody, settlement and payment rails when an agent needs to move money at 3 a.m. with nobody awake to approve it.

What to Watch

Ignore the headline number and track three things instead. Fund expense ratios, advisory fee schedules, and how much trading volume gets routed through agent-native payment rails over the next 24 months. If those start bending, the thesis has legs.

My read is that the direction is right and the date is fiction. Anyone building a 2035 revenue model on today's pricing is asking for trouble, and anyone dismissing agents as a toy is making the same mistake the wirehouses made about ETFs in 1995.

Time will tell, though. In the meantime, watch the basis points. They move before the headlines do.