Scuttlebutt

Scuttlebutt

Old GPUs could be worth zero if AI hits a power wall, says Parnassus's CIO

22 lessons: why he keeps most of his AI bet in second-order winners, how cheap mediocre stocks quietly cost active managers, and what a daily journal showed him about noise

Scuttlebutt's avatar
Scuttlebutt
Sep 30, 2026
∙ Paid

Insightful Investor · Sep 29, 2026 · 1h 6m · 46m saved

With Todd Ahlsten, CIO of Parnassus Investments

Old GPUs could be worth zero if AI hits a power wall, says Parnassus's CIO

About 45% of the Parnassus Investments portfolio sits in direct AI infrastructure, mostly semiconductor companies and the largest cloud providers. Most of the rest is in companies Todd Ahlsten expects to win from AI more slowly.

The market sorts companies into AI winners and losers every day. Ahlsten says the durable value is more likely to show up in precision agriculture, life sciences tools and payments, and that the chips behind the build-out keep their value only if there is enough power to run them.

"This business isn't about who has the highest IQ or even has the strongest work ethic."

Todd Ahlsten, CIO of Parnassus Investments, on the Insightful Investor podcast, joined the firm as an intern in 1995 and has run money through the dot-com bust, 9/11, the financial crisis and COVID. Parnassus was founded in 1984 and manages about $41 billion.

The full interview is covered here so you can skip it. 66 minutes of audio, 21 minutes of reading.

Here are the 22 lessons that matter.


This Scuttlebutt Summary is available in full to premium subscribers. If you're not one already, please consider subscribing to keep reading.


Key Takeaways

  • An old GPU could be worth zero if power runs short, even with every token sold

  • About 45% of the portfolio is direct AI infrastructure, and the biggest part of the rest is second-order AI winners

    • Life sciences tools, payments and precision agriculture, not a maximum bet on a neocloud

  • The silent killer for active managers is not a drawdown, it is a mediocre stock that lags a rising market for years

  • Quality means confidence a business will be worth more in 10 years, not a seven-paragraph answer

  • AI concentration only works if the companies buying the chips earn a profit from them

  • Most AI spending may be defensive, because falling behind can make the last two years of investment uneconomic

  • 90% of what he was stressed about ended up being noise, a journal he has kept since 2017 showed him

  • Patience is the one thing he thinks an AI model may not have, and the human edge in active management

1. Worth More in Pieces

Ahlsten joined Parnassus in 1995 as an intern, while a senior at UC Berkeley's Haas School of Business. He said research then meant annual and quarterly filings, the public library and a phone, and that founder Jerry Dodson taught him to treat how a company handles employees, the environment and governance as markers of quality alongside the financial statements.

A corporate raider taught him what a breakup is worth

His father was a pilot at Trans World Airlines. When Ahlsten was about 13, his father came home and said the airline might not be around much longer, because a corporate raider was buying shares and saw more value in breaking TWA up. His father said it was "like buying a car for $10,000 selling the pieces for 20."

"And that was just kind of a fascinating thing to me that the routes and the airplanes had value that weren't reflected in the price and it was worth more in pieces than a whole." — Todd Ahlsten

He said he was also a baseball statistics fan, started reading The Wall Street Journal every day after basketball practice, and is still fascinated about 40 years later.

2. Industry Structure First

User's avatar

Continue reading this post for free, courtesy of Scuttlebutt.

Or purchase a paid subscription.
© 2026 The Golden Age LLC · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture