@Asianometry & Dylan Patel — How the semiconductor industry actually works
Jon Y (Asianometry) (guest), Dylan Patel (guest), Dwarkesh Patel (host), Narrator
In this episode of Dwarkesh Podcast, featuring Jon Y (Asianometry) and Dylan Patel, @Asianometry & Dylan Patel — How the semiconductor industry actually works explores inside Chips and AI: Scale-Pilled Geopolitics, Taiwan Risk, NVIDIA Power The conversation explores how the modern semiconductor supply chain actually works, how fragile and stratified it is, and how it underpins the current AI boom. The guests dig into China’s catch-up strategy via espionage, talent poaching, and state-led centralization of compute, contrasting it with the more decentralized US ecosystem. They explain the technical and economic bottlenecks from process nodes and memory to data centers and power, and how AI demand is reviving old industrial sectors like power and networking. Finally, they link all this to AI scaling trajectories, OpenAI’s massive capital needs, and what happens to global tech and everyday products if Taiwan’s fabs go offline.
Inside Chips and AI: Scale-Pilled Geopolitics, Taiwan Risk, NVIDIA Power
The conversation explores how the modern semiconductor supply chain actually works, how fragile and stratified it is, and how it underpins the current AI boom. The guests dig into China’s catch-up strategy via espionage, talent poaching, and state-led centralization of compute, contrasting it with the more decentralized US ecosystem. They explain the technical and economic bottlenecks from process nodes and memory to data centers and power, and how AI demand is reviving old industrial sectors like power and networking. Finally, they link all this to AI scaling trajectories, OpenAI’s massive capital needs, and what happens to global tech and everyday products if Taiwan’s fabs go offline.
Key Takeaways
China could match or exceed US training runs if it centralizes compute.
China already imports large numbers of constrained NVIDIA GPUs and produces domestic accelerators; if Xi Jinping became truly “scale-pilled” and funneled most of this into a few national clusters, China could plausibly run frontier-scale (1e27–1e30 FLOP) models by the late 2020s.
Get the full analysis with uListen AI
Export controls are slowing but not stopping China’s semiconductor progress.
US controls effectively cap GPU performance sold to China, but tool exports and loopholes allow SMIC and Huawei to fabricate 7 nm–class chips domestically; sanctions also galvanize Beijing to treat semiconductors as a strategic, must-win industry.
Get the full analysis with uListen AI
The real bottlenecks are shifting from chips to data centers and power.
NVIDIA can manufacture millions of Hoppers/Blackwells, but building multi‑hundred‑thousand‑GPU clusters now runs into constraints on substations, transformers, cooling, fiber, and grid build‑out—areas where China is structurally advantaged over the US and Europe.
Get the full analysis with uListen AI
Process-node advances are increasingly funded by AI, not phones or PCs.
Moving to 3 nm, 2 nm and beyond is economically questionable on mobile alone; AI accelerators’ extreme appetite for density and energy efficiency is what makes N3/N2 viable and could push a large fraction of advanced TSMC capacity into AI by the late 2020s.
Get the full analysis with uListen AI
Semiconductor manufacturing knowledge is hyper-siloed and partially tacit.
Each engineer specializes in a tiny sliver (e. ...
Get the full analysis with uListen AI
A disruption in Taiwan would trigger a global “tech reset.”
Because Taiwan produces the bulk of leading-edge logic and huge shares of trailing-edge nodes (e. ...
Get the full analysis with uListen AI
AI investment is being driven by a Pascal’s-wager mindset among tech leaders.
CEOs at firms like Microsoft, Google, and Meta see the downside of under‑investing in AGI as existential for their companies; this justifies tens to hundreds of billions in CapEx before the revenue curve has caught up, as long as each new frontier model appears significantly more capable.
Get the full analysis with uListen AI
Notable Quotes
““If you are Xi Jinping and scale pilled, you must now centralize the compute resources.””
— Dylan Patel
“Leong Mong‑Song is a nut… He does not care about people, he does not care about business. He wants to take it to the limit, the only thing.”
— John (Asianometry)
“Semiconductor manufacturing and design is the largest search space of any problem that humans do because it is the most complicated industry that humans do.”
— Dylan Patel
“I don’t think you can stop the Chinese semiconductor industry from progressing. I think that’s basically impossible.”
— John (Asianometry)
“There’s no fucking way you can pay for the scale of clusters that are being planned to be built next year for OpenAI unless they raise, like, $50 to $100 billion.”
— Dylan Patel
Questions Answered in This Episode
If China fully centralized its AI compute, how would that change the global balance of AI capabilities and associated safety risks?
The conversation explores how the modern semiconductor supply chain actually works, how fragile and stratified it is, and how it underpins the current AI boom. ...
Get the full analysis with uListen AI
Are current US export controls on tools and chips actually optimizing for long-run AI safety, or just incentivizing China to build a more autonomous supply chain?
Get the full analysis with uListen AI
Given the master–apprentice, tacit nature of process knowledge, how feasible is it for AI to meaningfully accelerate semiconductor R&D in practice?
Get the full analysis with uListen AI
At what point do the economics of AI scaling break—does revenue growth plausibly keep up with 10–100x increases in training costs per generation?
Get the full analysis with uListen AI
How should governments and companies prepare for a sudden loss of Taiwan’s fabrication capacity, beyond just building a few redundant fabs abroad?
Get the full analysis with uListen AI
Transcript Preview
Leong Mong-sung is a nut. He's like, "We will make Samsung into this monster." He does not care about people, he does not care about business. He wants to take it to the limit, the only thing.
There's no fucking way you can pay for the scale of clusters that are being planned to be built next year for OpenAI unless they raise, like, $50 to $100 billion.
(laughs) Hold on, hold on. We've already lost John.
(laughs)
We've already accepted GPT-5 will be good.
But yeah, y- you got it, you know? (laughs)
You got it. You got it. It's like, bro, like, life is so much more fun when you just, like, are delusionally, like, you know?
We're just ripping bong hits, are we?
We're not even close to the dot-com bubble. Why wouldn't this one be bigger? We're gonna rip, baby.
(laughs)
You could-
Rip that bong, baby. (laughs)
You could raise AI for another two decades.
If you are Xi Jinping and scale pilled, you must now centralize the compute resources, right? They could have a bigger model than any of the labs next year.
Today, I'm chatting with Dylan Patel, who runs SemiAnalysis, and John, who runs the Asianometry YouTube channel.
Does he have a last name?
No, I do not.
(laughs)
No, I'm just kidding. John Y.
That's right, is it?
John Y.
Wait, wh- wh- why is it only one letter?
Because Y is the best letter.
(laughs)
(laughs) Why is your face covered?
(laughs)
Why not?
(laughs) No, seriously, why is it covered? (laughs)
Because I'm afraid of looking at myself get older and fatter over the years.
(laughs)
(laughs) Oh my God. But, but so seriously, it's like a- anonymity, right?
Anonymity.
Okay.
Yeah.
By the way, so d- d, uh, you know what Dylan's middle name is?
Actually, no. I know he told me, but-
W- what's my father's name?
I'm not gonna say, but I remember.
You can say, you can say it. It's fine.
Sanjay?
Yes. What's his middle name?
Sanjay?
That's right.
Wow.
Yeah.
So I'm Dwarkesh Sanjay Patel, he's Dylan Sanjay Patel. It's, like, literally my white name. (laughs)
Wow.
Yeah. I, uh, it's u- it's unfortunate my parents decided between my older brother and me to give me a white name and I could've been Dwarkesh Sanjay. Like, you know how amazing it would've been if we had the same name? (laughs)
(laughs)
Like butterfly effect and all-
Yeah, yeah.
... that probably would've all, would've turned out the same way, but like...
Maybe it would've been even closer if we would've met each other sooner, you know?
Yeah, yeah.
Who else is named Dwarkesh Sanjay Patel in the world?
Install uListen to search the full transcript and get AI-powered insights
Get Full TranscriptGet more from every podcast
AI summaries, searchable transcripts, and fact-checking. Free forever.
Add to Chrome