The Prof G Pod with Scott Galloway

China Decode: How an AI Price War Could Spark a Market Correction

November 18, 2025

Key Takeaways Copied to clipboard!

  • Chinese AI models, exemplified by Kimi K2 costing only \$4.6 million to train, are aggressively undercutting US models on price and efficiency, leading US companies to adopt them despite security concerns. 
  • The US-China AI race is diverging into two distinct models: the US pursuing ultra-premium, expensive frontier models (the 'iPhone equivalent'), while China focuses on lean, open-source, and cheap alternatives (the 'Android of AI'). 
  • Geopolitical tensions between China and Japan are escalating over Taiwan, driven by Japan's hawkish stance and historical grievances, which is already impacting Japanese stocks reliant on Chinese tourism. 

Segments

Chinese AI Cost Advantage (Unknown)
Copied to clipboard!
  • Key Takeaway: None
  • Summary: None
US vs. China AI Strategy
Copied to clipboard!
(00:04:02)
  • Key Takeaway: The US AI approach is characterized as ‘bubbly’ due to high spending, while the Chinese approach is ’lean and mean’ by focusing on lower cost and high efficiency.
  • Summary: The US AI efforts are seen as potentially bubbly due to billions invested in data centers and large models, whereas China prioritizes building models at a much lower cost. Former Google CEO Eric Schmidt suggested many governments will standardize on Chinese models because they are effectively free. Nuance exists, as some US models (like Meta’s Llama) are open source, and some Chinese models are also open source.
Chinese AI Architecture Efficiency
Copied to clipboard!
(00:12:52)
  • Key Takeaway: Kimi K2’s efficiency stems from using a ‘mixture of experts’ design, utilizing only a fraction of its trillion parameters per query, unlike dense US models.
  • Summary: Kimi K2 employs a ‘mixture of experts’ design, which only utilizes specific parameters when queried, contrasting with dense models that use all parameters. Although Kimi K2 has over a trillion parameters, it only uses about 32 billion at a time, pointing to a different research evolution in Chinese AI. This efficiency contributes to the lower training and inference costs observed.
Japan-China Diplomatic Firestorm
Copied to clipboard!
(00:15:20)
  • Key Takeaway: A diplomatic clash between Japan and China over Taiwan, sparked by PM Takeichi’s comments, signals historical mistrust and Japan’s shifting security posture amid US uncertainty.
  • Summary: China reacted furiously to Japanese Prime Minister Takeichi’s explicit suggestion of a military response to a Taiwan attack, leading to threats and state media insults. This spat highlights deep historical grievances from WWII and Japan’s concern over an ascendant China, prompting moves to question its non-nuclear principles. The tension is already impacting Japanese stocks reliant on Chinese tourism, with airlines and department stores seeing declines.
Starbucks Retreats in China
Copied to clipboard!
(00:25:00)
  • Key Takeaway: Starbucks is selling a majority stake in its Chinese business to local partner Boyu Capital after losing ground to hyper-competitive, localized rivals like Luckin and Cotti.
  • Summary: Starbucks is reducing its stake to 40% in China after nearly 30 years, unable to compete with local rivals offering bizarre, tailored drinks and aggressive pricing. Local chains like Luckin utilize hyper-efficient, app-based ordering systems that dominate the market, unlike Starbucks’ premium, slower product rollout. This trend of Western brands struggling is also seen with Apple and LVMH, as Chinese tastes evolve toward domestic alternatives.
Future Predictions on Robotics and Policy
Copied to clipboard!
(00:33:54)
  • Key Takeaway: James King predicts the ‘ChatGPT moment’ for humanoid robots—the ability to assess surroundings and complete novel tasks—will occur first in China by the end of 2027.
  • Summary: A Chinese robotics executive predicts the ‘ChatGPT moment’ for humanoid robots within three years, defined as the ability to find, pour, and deliver a glass of water autonomously. Alice Han agrees China is leading in embodied AI and humanoid robotics, noting the shift of US AI leaders toward world models. Alice also predicts the Trump administration may ban Chinese AI models by 2026 due to security concerns aligning tech companies and national security hawks.