Kimi, Odyssey, Tyler Cowen Joins, 5 Cups of Coffee, Aston Martin x COD

Watch on YouTube ↗  |  July 20, 2026 at 20:49  |  2:45:37  |  TBPN
Speakers
John Coogan — Co-Host, TBPN
Tyler Cowen — Economist
Danny Yeung — CEO and Co-founder, Prenetics
Tarek Mansour — CEO and Co-Founder, Kalshi
Connor Love — Partner, Andreessen Horowitz
Khalil — CEO and Co-founder, Natural
Tony Xiao — CEO and Co-founder, Sunday Robotics

Summary

The episode opens with a debate over China's Kimi K3 open-source AI model and its implications for semiconductor, compute, and cybersecurity markets. Guests include Tyler Cowen on AI, open source, and Europe; Danny Yeung on Prenetics and IM8 growth; Connor Love on American Dynamism venture capital; Khalil on agentic payments; Tarek Mansour on Kalshi's GPU futures; and Tony Xiao on robotics. Later segments cover film industry economics, coffee health, and Aston Martin's Call of Duty tie-in.

  • Kimi K3 open-source model is seen catching frontier models, triggering a US-China AI cold war debate.
  • Semi Analysis pushback argues K3 is positive for Nvidia, while compute constraints support Neoclouds and Microsoft.
  • Tyler Cowen argues open-source AI cannot be banned and European tariffs against China would fail.
  • Danny Yeung presents IM8/Prenetics growth and the General Catalyst financing as a public-market disconnect.
  • Connor Love discusses defense, space, and manufacturing investment opportunity under American Dynamism.
  • Khalil frames agentic payments as a large winner-take-most market.
  • Tarek Mansour introduces Kalshi GPU forward curves, positioning compute as a massive commodity derivative market.
  • Tony Xiao updates Sunday Robotics on 99.1% laundry folding and one-shot learning toward physical AGI.
Ideas
John Coogan Co-Host, TBPN 10:59
Open-source AI increases cybersecurity demand.
Open-weight frontier cyber capability increases hacker tooling and attack customization. Defenders such as CrowdStrike and Palo Alto Networks had months of preparation and systems patching, and open-source AI will force more advanced defense layers, supporting cybersecurity demand.
John Coogan Co-Host, TBPN 19:43
Kimi K3 is positive for Nvidia.
Kimi K3's linear attention and lower KV cache requirements are not bearish for Nvidia. Because K3 has more than 2.88 trillion parameters, inference requires the NVL72 scale-up domain to store weights, and wide expert parallelism actually increases networking bandwidth demand. This makes the model a positive for Nvidia, HBM/DRAM, and networking rather than a DeepSeek-style bear case.
John Coogan Co-Host, TBPN 21:06
Compute shortage revives Neocloud trade.
Kimi K3 is turning away paying customers because compute is capacity-constrained, which is the Neocloud thesis in one post. CoreWeave has lagged, but flexible capacity ready to serve new open-weight workloads can become profitable quickly, and Neoclouds should be back.
John Coogan Co-Host, TBPN 22:24
Microsoft can fund AI capex best.
The AI economy is severely compute-constrained across land, power, grid, chips, construction, and cooling. Companies with massive net income from non-AI products can fund hundreds of billions in capex, which is a bull case for Microsoft.
Tyler Cowen Economist 64:13
Europe loses a trade war with China.
If Europe starts a major trade war with China, Europe will lose. Tariffs will not rebuild competitive industries, they postpone adjustment and raise input costs for other sectors; Europe lacks the alternatives to shift into, so it is in a no-good-way-out pickle.
Danny Yeung CEO and Co-founder, Prenetics 101:12
Prenetics is undervalued versus IM8 growth.
IM8 reached $100m ARR in 11 months and projects over $200m in year two, shipping to 43 countries from day one; Prenetics owns 100% and holds about $140m cash while the Nasdaq-listed parent trades at a $354m market cap, creating a public/private valuation disconnect.
Tarek Mansour CEO and Co-Founder, Kalshi 127:35
GPU futures market could be enormous.
Compute is becoming a core commodity and Kalshi is building a forward curve/futures market for GPU pricing. Historically derivative markets run 10-15x the underlying, so compute could become the largest derivative market; the current flat H200 curve implies GPU pricing stability and innovation roughly matching consumption.
Up Next

This TBPN video, published July 20, 2026, features John Coogan, Tyler Cowen, Danny Yeung, Tarek Mansour discussing PANW, CRWD, NVDA, CoreWeave, MSFT, VGK, PRE, GPU futures. 7 trade ideas extracted by AI with direction and confidence scoring.

Speakers: John Coogan, Tyler Cowen, Danny Yeung, Tarek Mansour  · Tickers: PANW, CRWD, NVDA, CoreWeave, MSFT, VGK, PRE, GPU futures