Deep-tech investor talks about real-world AI across chips, healthcare, and data centers, and how AI’s economic upside will diffuse into society instead of being captured by a handful of giants
Phenomenal episode! The framing of proprietary data as the actual moat in vertical AI really clarifies why so many GPT wrappers collapsed after their initial ARR spike. What clicked for me was the raw acoustic ultrasound example, where the company grabbed that data when nobody valued it, then watched the market wise up and price everyone else out. It creates a compounding advantage that model quality alone can't replicate. This feels fundamentaly different from the horizotnal aggregation playbook we saw with search or marketplaces, where winner-take-most dynamics dominated. If AI value actually gets socialized through edge models and diffusion instead of captured by a few mega-caps, that reshapes the entire investment thesis. Curious whether the US can actually close teh infrastructure gap or if regulatory friction keeps handing industrial AI advantages to China by default.
Great insightful conversation over many gadgets (conceptual, physical, and somewhere in between), pitfalls, constraints, trendings, concerns, politics, east-west divergences, etc., in the landscape of AI (including the fast-commoditization of the ubiquitous buzzword “Agentic” and the bread-and-butter reasoning-power-hungry energy grid). Both of James’ (interviewed guest) and Grace’s insights led me to believe that there will not be a single winning country or enterprise in the “fight for AI dominance” that exploded in November of 2022 when the world witnessed the delivery of the first AI “baby” (ChapGPT). After 3 straight years of explosive growth (in terms of VC betting, GPU advancement, inference and diffusion breakthrough, monopolistic moat-building attempts, open vs closed models, etc.), this value-driven pie is growing bigger and bigger. I concur with James that the game-winning factors include but are not limited to the following: a) Vertical AI Optimization, b) Energy-Efficient AI Ecosystem, c) Customizable SaaS AI Edge Computing, d) Cost-Reducing Proprietary Data Adoption, and, e) Socialization of AI-delivered Values to 80% of the World’s Insight-Hungry Consumers. The ultimate winner(s) will be the farsighted VC pioneers without borders. [Personal Notes: I grew up in Hong Kong and was educated in the US with both business and IT background. I was an exchange student at the Zhongshan University (aka Sun Yat-sen University) in Guangzhou during my MBA program at the University of Southern California back in 1985. Next year, I will visit China (Beijing, Shanghai, and Guangzhou) in person after 30 years of absence. I will also visit my nephew’s startup company in Beijing (an entrepreneur building around Healthcare on Uber using AI). Grace’s easy-to-understand 5-star postings have been my inspiration in bridging the eastern and western frameworks for the adoption and development of AI.]
Phenomenal episode! The framing of proprietary data as the actual moat in vertical AI really clarifies why so many GPT wrappers collapsed after their initial ARR spike. What clicked for me was the raw acoustic ultrasound example, where the company grabbed that data when nobody valued it, then watched the market wise up and price everyone else out. It creates a compounding advantage that model quality alone can't replicate. This feels fundamentaly different from the horizotnal aggregation playbook we saw with search or marketplaces, where winner-take-most dynamics dominated. If AI value actually gets socialized through edge models and diffusion instead of captured by a few mega-caps, that reshapes the entire investment thesis. Curious whether the US can actually close teh infrastructure gap or if regulatory friction keeps handing industrial AI advantages to China by default.
Great insightful conversation over many gadgets (conceptual, physical, and somewhere in between), pitfalls, constraints, trendings, concerns, politics, east-west divergences, etc., in the landscape of AI (including the fast-commoditization of the ubiquitous buzzword “Agentic” and the bread-and-butter reasoning-power-hungry energy grid). Both of James’ (interviewed guest) and Grace’s insights led me to believe that there will not be a single winning country or enterprise in the “fight for AI dominance” that exploded in November of 2022 when the world witnessed the delivery of the first AI “baby” (ChapGPT). After 3 straight years of explosive growth (in terms of VC betting, GPU advancement, inference and diffusion breakthrough, monopolistic moat-building attempts, open vs closed models, etc.), this value-driven pie is growing bigger and bigger. I concur with James that the game-winning factors include but are not limited to the following: a) Vertical AI Optimization, b) Energy-Efficient AI Ecosystem, c) Customizable SaaS AI Edge Computing, d) Cost-Reducing Proprietary Data Adoption, and, e) Socialization of AI-delivered Values to 80% of the World’s Insight-Hungry Consumers. The ultimate winner(s) will be the farsighted VC pioneers without borders. [Personal Notes: I grew up in Hong Kong and was educated in the US with both business and IT background. I was an exchange student at the Zhongshan University (aka Sun Yat-sen University) in Guangzhou during my MBA program at the University of Southern California back in 1985. Next year, I will visit China (Beijing, Shanghai, and Guangzhou) in person after 30 years of absence. I will also visit my nephew’s startup company in Beijing (an entrepreneur building around Healthcare on Uber using AI). Grace’s easy-to-understand 5-star postings have been my inspiration in bridging the eastern and western frameworks for the adoption and development of AI.]