The AI Trade Isn’t the AI Economy

Key claims

• The trigger: on November 5, 2026 the Nasdaq lost about $2.3 trillion, Nvidia fell more than 9% and chips about 12%, “the worst rout since 2020,” as the market decided the political winds had turned on AI infrastructure.
• Two clocks: the AI trade (a few companies spending on chips, data centers and power, and a crowd betting on whether demand arrives in time) reprices daily; the AI economy (code written, tickets answered, robots learning, drugs found faster) accumulates. Coding is “the richest vein” today, but not the only one for long.
• Timing doesn’t pay: from 2005 to 2024, $10,000 left in the S&P 500 grew to $71,750; missing just the 10 best days left $32,871. Seven of the ten best days came within two weeks of the ten worst.
• “Hunter” vs. “gatherer”: the trader’s story of striking at the top and bottom is a good story but not how most people get fed; the gatherer who fills the bag on good days and bad does.
• Every great building era overbuilds (canals, railroads, fiber). Many funders lost money, “and that’s part of the story, not a footnote to it,” but the infrastructure was used for decades.
• His one disconfirming signal: idle compute. “A data center humming in the dark, fully powered, with nothing to do.” If capacity keeps getting cheaper and still sits idle, the season is shorter than he thinks. He hasn’t seen it.

Notable quotes

There are two clocks in a week like this. One ticks by the second and resets every morning. The other moves slowly: one engineer, then a team, then an industry.

The feast and the famine keep the same calendar.

Every great building time overbuilds. The canal diggers dug too many canals. The railroads laid more track than there were trains to run.

The good news and the bad news are already in the price, the way weather is already in the air. You don’t argue with weather. You dress for it, and you keep walking.

One thing would change my mind, and it isn’t red screens. It’s a data center humming in the dark, fully powered, with nothing to do.

There may be more pain to come in the AI trade. I believe the AI economy underneath it is just getting started.

How it connects

• Late-Stage Venture Returns Are Marks, Not Distributions (Eric Fitzgerald): the same post-midterm selloff in neoclouds and chipmakers, read from the LP side.
• 🎯 Nick Mehta on AI Profits… (tweet) and AI productivity gains don’t automatically… (tweet): the same week’s worry that AI isn’t yet showing up in corporate margins, the gap between Syvertsen’s two clocks.
• Technological Revolutions, Bubbles, Dot-com Bubble: the overbuild pattern he leans on.
• AngelList