Most startup advice assumes the biggest challenge is finding customers. But what happens when the demand is obvious and the real question is whether the technology can actually work at scale?
In this episode of Fund/Build/Scale, I sit down with Stanford professor and Inception Labs founder and CEO Stefano Ermon to discuss how a breakthrough research paper evolved into a venture-backed AI company serving enterprise customers.
Stefano explains why his team believed they needed a 10x advantage in speed, cost, or quality to compete with industry giants, and how they convinced investors to back an idea that was still largely unproven.
We explore the realities of building a deep-tech startup, including raising capital before product-market fit, assembling a world-class founding team, learning enterprise sales as a first-time CEO, pricing a new category of technology, and competing with companies like Google, OpenAI, and Anthropic.
Along the way, Stefano shares practical advice for technical founders trying to transform research into a business, de-risk ambitious ideas, and create evidence that investors can believe in before the market fully understands what they're building.
In this episode:
Why startups need a 10x advantage to stand outTurning a research paper into a venture-backed companyRaising capital when the biggest risk is technical, not market demandWhy Inception Labs built before talking to customersLearning enterprise sales as a first-time founderThe challenge of pricing a fundamentally new technologyWhat happened when Google announced its own diffusion modelsHow deep-tech founders can de-risk ideas before fundraisingThe hiring lesson Stefano wishes he had learned earlier📺 Watch this episode on YouTube
RUNTIME 46:03
EPISODE BREAKDOWN
(0:00) From Research Paper to Startup
(2:11) Why AI Is So Expensive
(4:34) Turning Research into a Company
(7:47) The Bet Behind Inception Labs
(9:01) Raising Money for an Unproven Idea
(12:24) Building the Right Team
(14:27) Dividing Founder Roles
(16:32) Building Before Customers
(20:07) When Customers Started Calling
(25:04) Pricing a New Category
(26:34) The Wrong First Customers
(31:16) Why Startups Can Beat Giants
(34:04) The 10x Rule for Startup Differentiation
(39:20) Building a Durable AI Company
(41:03) How to De-Risk a Deep Tech Startup
(42:40) The First 90 Days After Funding
(43:54) One Founder Lesson He Learned Too Late
LINKS
Stefano ErmonInception LabsWhat Are Diffusion Models?, Stanford HAIInception raises $50 million to build diffusion models for code and text, TechCrunch, 11/6/2025SUBSCRIBE
📥 Get the Fund/Build/Scale newsletter on Beehiiv: https://fundbuildscale.beehiiv.com/
📸 Follow Fund/Build/Scale on Instagram: https://www.instagram.com/fundbuildscale/
📺 Watch Fund/Build/Scale on YouTube: https://www.youtube.com/channel/UCFFH4cs2B1BKatPGs8SFRJw