David Epstein is General Partner at USF Ventures, the venture fund backing companies connected to the University of San Francisco. He was previously a General Partner at Crosslink Capital and has held management and CEO roles at more than half a dozen startups. He also teaches entrepreneurship and finance, and began his career at Data General as a computer designer, on the project chronicled in Tracy Kidder's Pulitzer Prize winning The Soul of a New Machine.
In this episode, Dave argues that the real AI bottleneck isn't chips, power, or capital. It's data. We get into why frontier labs backing open weight models is a defensive move rather than a principled one, where early stage startups can still win, and why he thinks jobs will disappear faster than they get created.
⭐This episode is brought to you by Podcast10x. We help founders and investors turn one podcast episode into a full month of content. Strategy, production, and distribution handled end to end. Learn more at https://podcast10x.com
What we cover:
→ Why "AI company" is no longer a category, and the pitch deck claim that has become his pet peeve
→ Why AI isn't a tool anymore, and what makes this cycle different from the dot com era
→ The real bottleneck: why we've exhausted the internet's data and what comes next
→ Money as the constraint nobody prices in, and the circularity in the current data center build out
→ How Chinese open weight models pull revenue out of token charges and subscriptions
→ Why big lab support for open models is defensive positioning
→ Who survives if open weights take share, and why consolidation is coming
→ Where early stage startups can still win: drug discovery, financial services, legal
→ Why the likely exit is a sale, not an IPO
→ Ethical investing as a return rather than a tax, and why it's tough to work with jerks
→ Why self-regulation rarely works, and what 2008 tells us about the current AI alliance
→ Why layoffs are just the beginning, and the Industrial Revolution parallel everyone forgets
→ Where the jobs actually are: management, human facing care, and the trades
→ What top tier VCs get right, and why VCs are also lemmings
→ Quantum computing as a data center accelerator, and the password problem it creates
→ Physics AI vs physical AI, and the validation problem sitting on top of both
→ Five year predictions: AGI, commonplace robots, and why consciousness doesn't matter
Connect with Dave Epstein:
LinkedIn: https://www.linkedin.com/in/thedavee/
USF Ventures: https://usfventures.com
Connect with Prashant Choubey:
LinkedIn: https://linkedin.com/in/choubeysahab
Subscribe to VC10X newsletter - https://vc10x.beehiiv.comVC10X website - https://vc10x.com
Timestamps:
(00:00) - Preview
(00:56) - Introduction to David Epstein and the Episode's Topics
(02:48) - How the AI Startup Landscape Has Fundamentally Changed
(05:08) - Comparing the Current AI Boom to the Internet Boom
(06:25) - Identifying the Next AI Bottleneck: Chips, Power, or Data?
(09:30) - Why Money is an Overlooked Bottleneck for AI Development
(11:14) - The Cyclical Nature of AI Investments and Financing
(12:46) - Analyzing Big Tech's Support for Open Source Models
(15:12) - Winners and Losers: Open Source vs. Frontier Models
(18:01) - How Early-Stage Startups Can Compete and Win in the AI Space
(20:53) - The Role of Ethics in AI Investment Decisions
(23:31) - The Challenge of Upholding Ethics in a Competitive Market
(26:21) - Implications of the OpenAI Model Escaping
(29:46) - The Future of AI-Driven Job Disruption
(32:46) - Where to Find Employment Opportunities in the AI World
(37:43) - How Top-Tier VCs Evaluate Founders and Make Decisions
(40:21) - The Most Exciting Emerging Areas of Innovation
(43:18) - Explaining Quantum Computing's Potential and Impact
(48:39) - An Ambitious AI Prediction for the Next 5 Years
(51:35) - Rapid Fire Round: USF Ventures' Investment Strategy
(53:06) - Conclusion
