Chinese AI Models Narrow the Gap, Apps Become the Moat, SaaS Faces Repricing

This episode of EUVC explores the accelerating convergence of AI models, with Chinese open-source models like DeepSeek V4 now benchmarking close to leading U.S. models such as Anthropic's Opus 4.6. The hosts debate whether the U.S. compute moat is eroding, highlighting how Chinese labs are training models on domestic chips and achieving rapid progress—potentially narrowing the frontier by just 78 days. The discussion shifts to strategic implications: as foundational models converge, the competitive edge is shifting to applications, making 'apps the moat.' Europe is urged to pivot from regulation to proactive strategy—leveraging public procurement, energy access, and single-market unification to build sovereign AI infrastructure. Cybersecurity breaches in the UK and France expose systemic risks from outsourcing sensitive data, while the hosts critique outdated data governance models. Sergey Brin’s return to DeepMind to close the coding gap with Anthropic underscores the race to make AI agents primary developers. SpaceX’s potential $60 billion deal with Cursor signals a strategic move to distribute AI via coding platforms. Meanwhile, the EU’s AI Act is being quietly gutted, raising concerns about regulatory retreat. The episode concludes with a warning about private credit’s vulnerability to AI disruption, which could pressure SaaS valuations and force legacy software firms to innovate or face re-rating. The hosts predict the next European trillion-dollar company may be a defense prime, not a software firm, and that Anthropic could reach $100 billion without ever going public.