AI funds had their worst single day since the COVID crash

Hedge funds just had their worst day since the COVID crash. The scoreboard moved both ways in the same fortnight.

About $39.2 billion as of October 16, per tokenized asset market data. That is up from roughly $39 billion at the last Tape.

Flat, in other words. The line resumed its stall after one week of movement.

Treasuries still anchor the category at about $14.9 billion. That is the base everything else sits on.

The action moved to equities. Tokenized stocks hit a record $3.9 billion on October 12, per Token Terminal.

Read that against what happened next. Sixteen issuers vetoed OKXICE’s listing filing in the same period.

So the supply is growing and the permission layer is tightening at once. Both are now features of this metric.

This is where the board cracked. Goldman Sachs flagged it on October 22.

All three major fund categories lost more than 1% that day. It was the worst single session since the COVID rout.

Tech-focused funds fell about 10% across October. The damage concentrated exactly where the exposure was.

Goldman still puts the broad group up 8.8% so far this year. That trails the 15.5% for human stock pickers last printed in July.

The gap is roughly seven points and it has not closed since summer. Fourteen weeks of a stale comparison, now with a fresh loss attached.

One detail explains the mechanism. Goldman said funds sold AI stocks at a 3.5-to-1 ratio between October 8 and 14.

That is positioning unwinding, not a thesis changing. Machines that made their year on the AI trade just gave a chunk of it back.

Two readings, two directions, and both are real.

Stablecoin transfers hit $891 billion in the third quarter. Coinbase and The Block Research published the figure on October 20.

That is four times the first-quarter 2023 level. The multi-year trend is intact and steep.

The short-term reading went the other way. Thirty-day volume fell to $6.68 trillion as of October 16.

That is down about 15% from $7.84 trillion at the September peak. Four straight declines now.

Market cap dipped to $300.5 billion, per DefiLlama on October 24. Supply flat, velocity cooling.

The two figures are not in conflict. One measures a quarter against a three-year baseline, the other measures a month against last month.

The sharpest entry this metric has produced all year. Barclays said on October 22 it will cut 5,000 jobs by year end.

That is 5% of its staff. About 60% of the cuts land on support roles, coders included.

The quote is the part to keep. CEO C.S. Venkatakrishnan called AI the icing on the cake.

Read the hedge in that phrasing. He is attributing most of the cut to something other than AI, while naming AI anyway.

The context makes it sharper. Barclays expanded its Anthropic deal on October 1, targeting half its developer population on Claude Code by year end.

Three weeks later, 60% of a 5,000-role cut hits support and engineering. Barclays targets more than £2 billion in savings from 2026 to 2028.

This is the second bank this month to name machines in a layoff. DNB cut 400 technology roles on October 6 and cited agentic AI.

Three new entries since the last Tape.

JPMorgan put its MONY and JLTXX tokenized funds on Solana on October 22. That follows JPMD landing on Ethereum three days earlier.

DTCC moved its tokenization service into commercial launch this month. The central depository is now selling the product, not piloting it.

Barclays grew its Anthropic collaboration on October 1. That one appears twice on this board, once as adoption and once as compression.

The SEC’s Innovation Exemption from September 17 handed issuers a 30-day per-venue veto. Sixteen companies used it against OKXICE.

ESMA ordered EU firms to stop all services for non-compliant stablecoins by January 8, 2027. The last channel for offshore tokens in Europe closes then.

Four US banks blocked outside AI agents from customer accounts in the week to October 16. The $1.6 trillion non-interest deposit pool held.

The split this fortnight is cleaner than usual, because the machines took rails and headcount while the AI trade gave back a chunk of its year in a single session. That is not a contradiction, it is a reminder that the infrastructure story and the positioning story run on different clocks and should never be read as one signal. Watch whether Goldman’s next print closes any of the seven-point gap to human stock pickers, because a stale July comparison is doing a lot of work on this board.

Territory: +machines on rails and cost, +incumbents on returns.

Human stock pickers still lead AI-run funds by about seven points this year, while the pipes underneath got faster and cheaper in the same month.