Inside the AI Sprint, Understanding Anthropic's Strategy | Tomasz Tunguz, Theory Ventures
The Peel with Turner Novak
Turner Novak
4.6 • 11 Ratings
🗓️ 15 May 2026
⏱️ 92 minutes
🧾️ Download transcript
Summary
Tomasz Tunguz is the Founder and General Partner of Theory Ventures.
We talk about today’s “all out sprint” in AI, Anthropic’s strategy, the three layers of AI business models, how AI compares to prior technologies, where to invest in AI today, and what Theory looks for in new investments.
Thank you to Numeral, Flex, and Amplitude for supporting this episode
Numeral: The end-to-end platform for sales tax and compliance https://www.numeral.com
Flex: Get premium banking and a net 60 day credit card at 0% APY https://home.flex.one/referral/bananacapital
Amplitude: AI analytics, all you have to do is ask https://www.amplitude.com
Timestamps:
(0:42) The “all out sprint” in AI today
(1:40) Why GPU prices are up 116% in six weeks
(6:34) AI infra end-state: “We’ll over build”
(9:12) Tokenmaxxing, and why AI needs to get more efficient
(15:48) AI models will resemble pharma more than software
(19:52) Why Anthropic still trades at a discount
(25:42) Anthropic’s strategy: commoditize the compliments
(30:29) Why OpenClaw is so strategic for OpenAI
(34:08) The three layers of AI business models
(38:18) Where to invest in AI today
(45:49) Who will survive SaaSpocalypse?
(52:15) Comparing AI’s impact to historical technology cycles
(57:34) How new technology historically impacts jobs
(1:05:58) Where AI is underrated today
(1:10:41) How people are actually buying AI products
(1:14:06) Why Theory’s investing in ads, inference, and email
(1:16:24) 2026 IPO pipeline, how VC has changed over 20 years
(1:20:56) What Theory looks for in new investments
(1:22:32) Starting Theory Ventures in 2022
(1:25:39) Running a monte carlo analysis to determine portfolio construction
(1:27:54) Tomasz personal AI projects
Referenced
Theory Ventures: https://theoryvc.com/
Tomasz Blog: https://tomtunguz.com/
Follow Tomasz
Twitter: https://x.com/ttunguz
LinkedIn: https://www.linkedin.com/in/tomasztunguz
Follow Turner
Twitter: https://twitter.com/TurnerNovak
LinkedIn: https://www.linkedin.com/in/turnernovak
Subscribe to my newsletter to get every episode + the transcript in your inbox every week: https://www.thespl.it/
Transcript
Click on a timestamp to play from that location
| 0:00.0 | Tomaz, welcome to the show. Pleasure to be here, Turner. Thanks for having me on. I know. It's kind of funny. We had never met before last week, and then we were at the Samuels thing, and then we're recording this podcast. And then next week, we're going to be at the Beyond Summit. So it's like three, three times right in a row. Three Pete. Let's go. Let's do it. |
| 0:22.4 | And so really quick for people who don't know, what is theory ventures? |
| 0:26.5 | We are an early stage AI-focused venture firm. We invest in anywhere from one to 45 million, |
| 0:31.8 | typically in B2B software and infrastructure companies. How would you kind of, I guess, |
| 0:37.2 | summarize sort of the state of AI today? |
| 0:40.3 | A little bit of an open-ended question, but how do you kind of think about everything that's going on? |
| 0:45.3 | All out sprint. That's the way it feels. I mean, okay, so why do I say that? The first is there aren't enough GPUs for anybody. |
| 0:55.1 | So people are sprinting to buy GPUs or rent them. |
| 0:58.0 | I think the second thing is model improvements. |
| 1:02.1 | Model only remains state of the art for about 41 days, |
| 1:05.5 | even though it's several hundred million or a billion to train, |
| 1:09.4 | maybe less. |
| 1:14.6 | And then there's also an all-out sprint for customer acquisition. Buyers are the most open they've ever been to trying new things. |
| 1:20.5 | And so if you can capture many of them, you'll have a big business. |
| 1:24.0 | And then the businesses themselves are growing at unprecedented rates. |
| 1:28.3 | So I think everybody is sprinting. |
| 1:30.9 | Yeah, I guess maybe the first one you mentioned, the GPU prices, what does that even mean? |
| 1:36.2 | Like for somebody who is not super familiar with, I guess, any of this? |
| 1:40.6 | Like, how would you just explain that to a smart person who is hearing this for the first time? |
| 1:45.3 | Yeah. So to run a machine learning or an AI model, you need a GPU, which is a particular |
| 1:51.4 | kind of chip that does lots of calculations in parallel at the same time, matrix math, |
| 1:56.0 | it's called. And if you have a MacBook, you have one. In fact, you have an excellent one. |
... |
Transcript will be available on the free plan in 6 days. Upgrade to see the full transcript now.
Disclaimer: The podcast and artwork embedded on this page are from Turner Novak, and are the property of its owner and not affiliated with or endorsed by Tapesearch.
Generated transcripts are the property of Turner Novak and are distributed freely under the Fair Use doctrine. Transcripts generated by Tapesearch are not guaranteed to be accurate.
Copyright © Tapesearch 2026.