Jul 29, 2026 · 58m · big-technology
How China Caught U.S. AI — With Grace Shao
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Host Alex Kantrowitz and tech analyst Grace Shao explore how China continues to close the artificial intelligence frontier gap with the United States through open-source collaboration, extreme hardware optimization, and elite talent retention.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 34.8% of the talking time here. How this is scored →
Alex as informed peer 5.2 Guest teaching 5.9 Guest disagreement 2.2 Alex pushing back 2.6
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 29:31 Grace pushes back on state subsidization theory
Grace openly calls the popular US argument around Chinese state subsidization 'really funny' and firmly clarifies that these labs operate as private entities seeking IPO capital.
Hardest push from Alex ▶ 8:32 Alex presses on the 996 grind cultureAlex refuses to let the culture explanation pass without directly questioning whether China's progress is merely the result of grueling 9am-to-9pm 6-day work weeks.
Biggest teaching moment ▶ 49:40 The nuanced drivers of talent repatriationGrace systematically deconstructs Western assumptions about why researchers leave the US, detailing spousal credentialing barriers, native culture, and modern quality of life parity in Asian metros.
Alex holds their own ▶ 44:04 Alex frames the macro shift in app paradigmsAlex demonstrates deep analytical insight by synthesizing Grace's reporting into a novel thesis: the historical trend has flipped, leading to US AI super-apps and Chinese specialized vertical applications.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| STEM Talent and Compute-Driven Specialization in China | 5 | 6 | 1 | 1 | Alex opens by citing benchmark rankings and technical assessments of Kimi K-3 from Nathan Lambert to set up the discussion. Grace details the sheer volume of Chinese STEM talent, domestically educated researchers, and how compute bottlenecks force labs to specialize rather than compete across every layer. | |
| Lab Culture, Mission Focus, and Work Ethic Debates | 5 | 6 | 3 | 4 | Alex quotes Lambert regarding organizational culture and pushes on whether Chinese labs are simply outworking Americans via 996 schedules. Grace pushes back against the pure grinding stereotype, citing DeepSeek's low churn and founder statements rejecting performative overtime. | |
| Vertical Integration, Proprietary Data, and Shared Open-Source R&D | 6 | 6 | 1 | 2 | Alex explores the dynamics of open-source collaboration versus closed-model labs. Grace provides insider examples from Meitu and Chinese open-source ecosystems where labs cite and build upon each other's research openly. | |
| The U.S. Open-Source Pivot and Monetization Pressures | 6 | 6 | 2 | 2 | Alex points out the irony of US frontier labs pivoting to endorse open source. Grace explains that enterprise self-hosting and token cost considerations mean US companies must compete in open source or risk losing developers to Chinese alternatives. | |
| Smart Distillation vs. Genuine Innovation in Chinese AI | 6 | 7 | 3 | 4 | Alex directly raises the distillation controversy and IP theft allegations against Chinese labs. Grace reframes the discussion by drawing a clear distinction between unsophisticated dumb distillation and smart distillation used broadly across the industry. | |
| Commercial Models and China's Strategic Open-Source Diplomacy | 6 | 7 | 5 | 5 | Alex challenges Grace on the commercial logic of open sourcing weights and suggests a state-driven commoditization strategy. Grace bluntly dismisses the simplistic state subsidization narrative and explains API monetization along with Belt and Road soft power diplomacy. | |
| Sponsor Segment: Gravity AI Agent Security Documentary | 7 | 6 | 2 | 2 | Following the sponsor break, Grace explains how Chinese AI firms are pivoting to Europe and Southeast Asia while building pragmatic B2B vertical tools like Alibaba's Axio. Alex synthesizes this into an insightful macro theory that China is building vertical apps while the US builds super apps. | |
| Navigating Compute Bottlenecks and Domestic Chip Solutions | 6 | 6 | 2 | 4 | Alex channels OpenAI's compute moat thesis, asking whether chip shortages will cap Chinese parity. Grace acknowledges the bottleneck while highlighting Huawei-DeepSeek co-optimization and China's massive energy grid advantages. | |
| Reverse Brain Drain and China's Talent Retention | 4 | 8 | 3 | 3 | Alex asks why elite researchers like Yang Zhilin return to China rather than remaining in the US. Grace offers a thorough, multi-faceted breakdown encompassing spousal career friction, family proximity, cultural comfort, and the shifting calculus of urban quality of life. | |
| Robotics Progress, Supply Chain Dominance, and Adoption Timelines | 5 | 7 | 2 | 2 | Alex asks about the state of Chinese robotics and physical AI. Grace details Shenzhen's manufacturing speed and 50% hardware cost advantage while tempering consumer humanoid hype by citing a realistic 10-year enterprise timeline from Pony AI's CEO. | |
| Episode Conclusion and Where to Follow Grace Shao | 1 | 0 | 0 | 0 | Alex wraps up the conversation, plugs Grace's Substack, and thanks her for returning to the show in a standard friendly outro. |
Statements from this episode (23)
Moonshot's Kimi K3 Ranks Top Three Across Major AI Benchmarks
“This is from Nathan Lambert's Substack Interconnects. It comes number two on the VALS AI Index. Number three overall on artificial analysis is Intelligence Index number one in the front end cone arena, and it has many more impressive results.”
Shao: 40% to 50% of Leading AI Researchers Are of Chinese Descent
“And then we see like the leading researchers really, really a lot of them, if not, you know, maybe 40, 50% of them are of Chinese descent or heritage.”
Shao: Compute Limits Force Chinese AI Labs Into Distinct Specializations
“What you've seen is a lot of these tiny labs faced with compute constraint and in some ways capital constraint, they're forcing them to specialize rather than compete across every dimension. So, you know, for the sake of, you know, deep seek, it's really, real…”
Shao: DeepSeek and Moonshot Founders Focus on AGI Over Consumer Apps
“I think both, you know, DeepSeek and Kimi, if not seen as kind of the top two labs right now coming out of China, both founders have openly talked a lot about management of people you know, removing distractions, really focused on the pursuit of AGI, not kind …”
Shao: DeepSeek Lacks Capital Constraints and Pressure to Commercialize
“The fact that they don't actually have that much pressure to commercialize because they have enough money, he says, like, look, we have enough capital, we are not capital constraint compared to maybe other, other labs, and we are really, really focused on just…”
Shao: Western Rivals' Infighting Highlights Undervalued Unity of Chinese Labs
“I think people are a bit undervaluing this case, especially when we're seeing a lot of other labs maybe having a bit of infighting or, you know, internal mission or value misalignment.”
Shao: Meitu Focuses on Fine-Tuning Open Models Over Building Massive Counterparts
“When I speak to the mate who CFO, which is a leading like creative AI company he was saying, look, we don't really need the most like the biggest model. We don't need the best model in that sense, but we need the best model fine tune for our use case. So we ca…”
Shao: Chinese AI Labs Open-Sourced Models to Disarm Global Skepticism
“It's essentially in the beginning, it was like a branding strategy for a lot of these labs. And when I spoke to them, because they said, look, if we want more developers on our ecosystem or on our using our APIs, they need to know what's out there, especially …”
Shao: Self-hosting Chinese open-source AI eliminates China data flow risks
“Essentially if you self host these models or you use them through these inference service providers like fireworks, these models become yours or American, if you want to put it right. So the whole fear mongering around does the data or whatever go to China doe…”
Shao: Kimi K3 is cheaper than frontier models while nearing parity
“And even despite, you know, Kimi K three being said is a token hungry model. It's not as cheap as other Chinese models. The task per, I think the task per token usage is still quite high. It's still cheaper. It's still cheaper than the most frontier models and…”
Shao: Quant Hedge Fund Backing Allows DeepSeek to Ignore Immediate Profits
“On the China side, deep seeks people have talked about how they are not revenue driven. So whatever money they, sorry, they're not profit driven. So whatever revenue they make, they want to put that money into R&D, given that they have the quant fund kind of l…”
Shao: Smart Model Distillation Breaks Terms of Service, Not the Law
“Now there's smart distillation, which is kind of operating in the gray area. It is again, not IP theft. It's not breaking the law. However, it could be breaking what is considered, you know You know, your own term services.”
Shao: MiniMax AI Projected to Reach $1.2 Billion ARR by Year-End
“I think, you know, if you look at, I think Moon Minimax AI I think AI's run rate AR is already something like one billion now, and then Minimax projected to be one billion or to 1.2 billion by the end of the year.”
Shao: Most Chinese AI labs avoid government funding to protect IPO plans
“And in fact, some, most of them actually don't want to take government money because there's some hindrance as well, right? Like, and when they go public and et cetera on their structure.”
Kantrowitz: Selling Raw AI Intelligence Alone Won't Build Trillion-Dollar Companies
“I think that if there was a belief that you could build a trillion dollar company just by selling intelligence from the API alone not gonna happen most likely.
but what could happen is you, if you're the developer of the intelligence, you can build products w…”
Kantrowitz: US May Shift to AI Super Apps While China Proliferates Niche Apps
“And I think maybe because of China's embrace of open source and because of the U S closed AI model we might end up seeing a proliferation of individual
AI apps in China while the U.S. Goes to super apps.”
Shao: Huawei and DeepSeek Are Collaborating Closely on Hardware Optimization
“Huawei is working very closely with deep seek on how to optimize hardware and software and try to figure out this.”
Shao: China's Energy Grid Prevents Power From Becoming an AI Compute Bottleneck
“China's energy infrastructure side is not a problem, so if you have enough electricity and power to be powering these chips, even if you need to use double amount of energy, that's not a Kind of a bottleneck for the AI compute side.”
Shao: Non-transferable spousal credentials drive AI researcher return to China
“Personal reason I think is the main driver for most people, honestly, like I I've spoken to researchers where hate to overgeneralize, but their wives or spouse or whoever are maybe teachers, lawyers you know, healthcare professionals in China. And those are no…”
Shao: U.S. Quality of Life Advantage No Longer Drives Chinese Immigration
“The fact is, I think, 30 years ago, for any average Chinese immigrant to go to North America, it is a no brainer, because the quality of life in almost any major city is going to be higher than a major city in China. But that kind of, I'm not talking about pol…”
Shao: Chinese robot hardware is at least 50% cheaper than elsewhere
“It's said that, you know, a lot of these robots, whether they're humanoid or industrial robots, the raw kind of hardware itself can be like at least 50% cheaper than produced elsewhere.”
Shao: AI robotics and autonomous vehicles are bottlenecked by physical data
“And then the real, real, real bottleneck for these, all these companies, whether you're talking about autonomous driving or industrial robotics or human robotics with, in terms of AI integration is that they don't have enough physical data.”
Shao: Mass-Market Humanoid Robot Deployment Is 10 Years Away, Not Five
“People love to hype about it. They're saying, look, human robots are going to be deployed in the next three to five years in the mass market. He gave me a rough number of 10 years again, and I believe him, because I've seen a lot of these robots. They frankly …”