Uploaded July 2026 | Updated November 2026, 1 day ago
For years, Silicon Valley has justified spending hundreds of billions of dollars on AI infrastructure by promising that bigger models, bigger data centers, and bigger GPU clusters would inevitably lead to dominance.
But what happens when someone offers nearly the same capability—for a fraction of the cost?
In this episode, Carl Zha welcomes returning guest TP Huang to discuss why China's latest AI breakthroughs could fundamentally reshape the global AI industry.
From GLM and Kimi to DeepSeek and Huawei's rapidly advancing AI infrastructure, they explore why Chinese AI companies are closing the gap with OpenAI and Anthropic much faster than many expected.
The conversation examines whether the AI boom has become an investment bubble, why open-source AI threatens Silicon Valley's business model, how export controls may have accelerated Chinese innovation, and whether Huawei's domestic AI hardware ecosystem can eventually overcome America's chip advantage.
They also discuss recursive self-improvement (RSI), AGI, the economics of AI inference, data pipelines, and why the next phase of AI competition may depend less on raw compute and more on engineering efficiency.
Topics Covered
Is Silicon Valley facing an AI bubble?
Why China's AI models are becoming dramatically cheaper
Open-source AI vs proprietary AI
GLM, Kimi, DeepSeek and China's leading AI labs
Anthropic's concerns about Chinese AI
The economics of AI pricing
Why export controls may have backfired
Huawei Ascend chips and Atlas AI clusters
Compute versus software optimization
Recursive Self-Improvement (RSI) explained
Does AGI guarantee victory?
China's long-term AI strategy
If you enjoy thoughtful discussions about China, technology, geopolitics and the future of artificial intelligence, subscribe to Silk and Steel and hit the notification bell.
Question for viewers:
Do you think cheaper open-source AI will eventually replace proprietary models like ChatGPT and Claude?
Let us know in the comments.
#China #ArtificialIntelligence #OpenAI #DeepSeek
:::
⏱ Chapters
00:00 The AI bubble may be bursting
01:40 China's AI pricing threatens Silicon Valley
05:30 Why Anthropic fears open-source AI
09:30 Is China only six months behind?
13:00 Distillation controversy explained
16:00 AGI and Recursive Self Improvement
20:30 China's real AI advantage
22:40 Compute shortages vs software efficiency
26:00 Huawei chips vs Nvidia
30:00 Why cheap electricity matters
💥 Biggest Takeaways
China's AI companies are competing on cost, not just capability.
Open-source models may undermine the pricing power of OpenAI and Anthropic.
Chinese labs are approaching frontier performance much faster than expected.
Huawei's AI hardware ecosystem is becoming increasingly competitive.
Data—not compute—may become the biggest bottleneck.
AI competition is shifting from model quality alone to deployment economics.
Lower-cost inference could dramatically reshape enterprise AI adoption.
Silicon Valley's massive AI infrastructure spending depends on maintaining premium pricing.
For years, Silicon Valley has justified spending hundreds of billions of dollars on AI infrastructure by promising that bigger models, bigger data centers, and bigger GPU clusters would inevitably lead to dominance.
But what happens when someone offers nearly the same capability—for a fraction of the cost?
In this episode, Carl Zha welcomes returning guest TP Huang to discuss why China's latest AI breakthroughs could fundamentally reshape the global AI industry.
From GLM and Kimi to DeepSeek and Huawei's rapidly advancing AI infrastructure, they explore why Chinese AI companies are closing the gap with OpenAI and Anthropic much faster than many expected.
The conversation examines whether the AI boom has become an investment bubble, why open-source AI threatens Silicon Valley's business model, how export controls may have accelerated Chinese innovation, and whether Huawei's domestic AI hardware ecosystem can eventually overcome America's chip advantage.
They also discuss recursive self-improvement (RSI), AGI, the economics of AI inference, data pipelines, and why the next phase of AI competition may depend less on raw compute and more on engineering efficiency.
Topics Covered
Is Silicon Valley facing an AI bubble?
Why China's AI models are becoming dramatically cheaper
Open-source AI vs proprietary AI
GLM, Kimi, DeepSeek and China's leading AI labs
Anthropic's concerns about Chinese AI
The economics of AI pricing
Why export controls may have backfired
Huawei Ascend chips and Atlas AI clusters
Compute versus software optimization
Recursive Self-Improvement (RSI) explained
Does AGI guarantee victory?
China's long-term AI strategy
If you enjoy thoughtful discussions about China, technology, geopolitics and the future of artificial intelligence, subscribe to Silk and Steel and hit the notification bell.
Question for viewers:
Do you think cheaper open-source AI will eventually replace proprietary models like ChatGPT and Claude?
Let us know in the comments.
#China #ArtificialIntelligence #OpenAI #DeepSeek
:::
⏱ Chapters
00:00 The AI bubble may be bursting
01:40 China's AI pricing threatens Silicon Valley
05:30 Why Anthropic fears open-source AI
09:30 Is China only six months behind?
13:00 Distillation controversy explained
16:00 AGI and Recursive Self Improvement
20:30 China's real AI advantage
22:40 Compute shortages vs software efficiency
26:00 Huawei chips vs Nvidia
30:00 Why cheap electricity matters
💥 Biggest Takeaways
China's AI companies are competing on cost, not just capability.
Open-source models may undermine the pricing power of OpenAI and Anthropic.
Chinese labs are approaching frontier performance much faster than expected.
Huawei's AI hardware ecosystem is becoming increasingly competitive.
Data—not compute—may become the biggest bottleneck.
AI competition is shifting from model quality alone to deployment economics.
Lower-cost inference could dramatically reshape enterprise AI adoption.
Silicon Valley's massive AI infrastructure spending depends on maintaining premium pricing.










