The Twenty Minute VC (20VC) · Sep 24, 2026 · 1h 21m · 1h 2m saved
With Harry Stebbings, Founder of 20VC · Jason Lemkin, Founder of SaaStr · Rory O'Driscoll, Partner at Scale Venture Partners
Meta gained about $100 billion of market value in the week its Muse assistant took the top spot on the App Store, by Rory O'Driscoll's count, and Jason Lemkin called it the first real rival to ChatGPT.
Most of the AI trade is priced on the frontier labs. The panel spent much of the hour on the businesses agents could route around, and on whether this is the top of the venture cycle; Harry Stebbings said the loudest warning about the top came from a firm that has paid some of the highest prices.
"I think it's the last stand of the unnecessary system of record."
Harry Stebbings, who founded 20VC and runs its venture fund, hosts this weekly news round table on The Twenty Minute VC with Jason Lemkin, founder of SaaStr, and Rory O'Driscoll, a partner at Scale Venture Partners.
The full episode is covered here so you can skip it. 81 minutes of audio, 18 minutes of reading.
Here are the 12 calls that matter.
Key Takeaways
Anthropic delayed its IPO to November so the October numbers can go in the prospectus, in O'Driscoll's reading
OpenAI's plan needs about $700 billion of capex to reach $350 billion of revenue, and much of it sits on other companies' balance sheets
Meta's Muse is a ChatGPT competitor that is free and has autonomous agents, Lemkin said
It built him a CRM that tracks 150 sponsors
Agents will route around Amazon, Resy and OpenTable, and Lemkin expects those businesses to be maimed, not killed
About 20% of today's LLM calls could move to cheap decision models like Jev, which O'Driscoll put at $20 billion of spend that could shrink to $4 billion
Stebbings called Menlo's warning about the top of the cycle "a blowhard piece"
Lemkin approved Factory at $5 billion because large companies do not trust the labs with their code; O'Driscoll called coding "the mother lode"
Lemkin passed on Crusoe, and O'Driscoll warned the data center trade is the most leveraged bet on AI growth
Stebbings, an Airwallex investor, called the China allegations "ridiculous"; Lemkin said Chinese ties are already blocking real deals
1. Anthropic Waits For November
Stebbings opened with Anthropic moving its $2 trillion IPO from October to November and asked whether it was a crack in the market or a wish for a better prospectus. O'Driscoll said it was the prospectus. Anthropic had an amazing second quarter and became bigger than OpenAI, and bankers would rather include the October numbers than list after the quarter ends but before the results can be shared.
Bankers want the numbers to do the talking
"If I'm a banker, I'm thinking if we do this in October, it's going to be a lot of explaining. If we do this in November, the numbers will talk." — Rory O'Driscoll
Lemkin said the delay still suggested "maybe there's just a hint of worry" about getting the 20 or 30 times oversubscription the company wants for the largest IPO of their lifetimes. O'Driscoll said the choice shows confidence, a willingness to take another month of timing risk for a better price.
"And there's always that 10% chance the world goes crazy and you look back and you go, damn, should have taken the 100 billion." — Rory O'Driscoll
Stebbings said a friend, a very successful investor, had asked him how frontier labs can go public when no one will sell them liability insurance. O'Driscoll dismissed the point: once a company has said publicly there is a 10% risk its product blows up the world, product liability is noise.
A $2 trillion company can insure itself
"This is a $2 trillion market cap company can self-insure." — Rory O'Driscoll
It does not need Munich Re, worth about $200 billion, to reinsure it, he said, and anyone buying the stock who does not know management considers the product dangerous has not read their tweets. Lemkin said the labs will instead keep about 200 lawyers in house plus the top law firms to fight liability suits, as tech companies fight patent trolls.
"The most important part of securities laws, security laws is not that you have sell stock that has no risk. It's that you sell stock where you disclose the risk." — Rory O'Driscoll
O'Driscoll praised Jensen Huang for saying his probability of AI destroying humanity is zero. Lemkin said Huang can say that because Nvidia's own open-source models have not caught up yet.
