20VC: 70% of Neolabs Will Die | There Will be a $100BN US Open-Source Model | Data is a Trillion $ Market | Governments Cannot Regulate Models: It is Too Late | The Cyber Attacks to Come Will be Insane with Anastasios Angelopoulos @ Arena
Summary
Harry Stebbings interviews Anastasios Anagnostopoulos, CEO of Arena, on the rise of Chinese open-source AI models like Kimi K3 beating US counterparts. They discuss the future of AI sovereignty, the potential for a trillion-dollar American open-source company, and the risks of cyberattacks and fake AI job applicants. The conversation also covers model routing, pricing pressures, and why government regulation of model releases is impractical.
Topics discussed
Introduction: AI inference margins and guest Anjini Midha
Sponsors: JP Morgan and Base44
What is LMSYS Chatbot Arena and its importance
Open source vs. closed source models and cannibalization
Business incentives for fine-tuning and data moats
US vs. China AI competition and open source lag
Chip export controls and regulatory trade-offs
Data sovereignty, local hosting, and security risks
Model routing, commoditization, and enterprise adoption
Pricing pressure, IPOs, and valuation realities
AI safety, guardian models, and fake applicant attacks
Neolab valuations, revenue, and the 'zero risk' trap
Data providers as complementary goods to AI training
Arena's business model: Agentic evaluations and margins
Vertical AI risks: Harvey AI and model provider competition
Founder reflections: Focus, university, and hindsight
Market hype, compute debt, and underrated neolabs
Future of AI, human flourishing, and closing remarks
Listen ad-free on Castria