"Turpentine VC" | Venture Capital and Investing · Erik Torenberg

E68: How Theory Ventures Differentiates in a Crowded VC Market

·53 min·3 clips

OpenAI is earning about $1.3 to $1.5 billion in run rate, and Tom says small firms cannot play foundation models.

1. Turpentine VC E68 centers on Theory Ventures and Tom Tunguz's view of how a new firm differentiates in a crowded venture market. 2. Tom Tunguz is a general partner at Theory Ventures, and the episode also includes the host of Turpentine VC framing Theory's $450 million second fund. 3. The episode asks how Theory decided where to fit in, how concentrated the fund should be, and which AI and data businesses still fit venture returns. 4. Tom says Theory began with Monte Carlo analysis on historical venture returns to design a thesis-driven, highly concentrated portfolio. 5. He contrasts a Y Combinator-style broad index model with the classical venture model of early bets and significant ownership. 6. Tom says the US venture capital asset class grew from about $800 billion to $300 billion over 12 years, with 81% of those dollars from non-traditional investors according to PitchBook. 7. He predicts the asset class may settle around $150 billion to $180 billion as later-stage capital gets cut back. 8. Tom says the fundraising market was difficult six to nine months earlier because LPs were re-evaluating portfolios after public-market crashes, China tensions, and higher Fed funds rates. 9. He says smaller funds are attractive because the multiples are higher, and he advises managers to ask LPs how many emerging managers they backed in the last 18 to 24 months. 10. Tom argues that differentiation only works when product idea, product, and go-to-market strategy align, whether the strategy is AI, proprietary access, or a specialized roll-up fund. 11. On AI, Tom says OpenAI is earning roughly $1.3 billion to $1.5 billion in run rate and that Microsoft has two AI businesses of similar combined size. 12. He says foundation models are too capital-intensive for small firms like Theory, so he focuses on enterprise readiness for LLMs. 13. Tom says he and his team interviewed about 25 data leaders over a couple of months, and only one had deployed an LLM in production. 14. He flags risks around moving data into and out of LLMs, including indemnification terms and uncertainty in enterprise use. 15. Tom says Theory is also researching application-layer AI that could combine CRM and marketing software in new ways using English as an API. 16. In the data segment, Tom says data teams are becoming software engineering teams because data now powers personalization, risk scoring, and chatbots. 17. He says LLMs are non-deterministic and hallucinate, which makes BI and text-to-SQL difficult for enterprise data systems. 18. Tom describes his tone as analytical and framework-driven, with long, specific answers and frequent comparisons across venture, software, and data markets. 19. Listeners who follow venture strategy, AI infrastructure, and data platforms will get the most value. 20. Listeners looking for quick company news or light banter will probably skip this one.

As heard by us

A clear look at how Theory positions itself in a crowded venture market.

Theory Ventures comes across as a firm trying to stand out in a crowded venture market, with fund size, concentrated portfolios, and founder perception all part of the pitch.

Read the full review in PlayNext →

Why you'd press play

You want a clear map of how Theory Ventures tries to stand out.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from "Turpentine VC" | Venture Capital and Investing