- 9 minutes ago
China's AI ambitions are no longer just about catching up to Silicon Valley. From chips and data centers to energy, talent and global markets, Beijing is building an ecosystem designed to challenge US tech.
Category
🗞
NewsTranscript
00:00Hardly a day seems to go by without some kind of disturbing headline from U.S. big tech companies
00:04when it comes to artificial intelligence. It might be that AI agents have supposedly gone
00:09rogue and are now hacking into websites and accounts. Or maybe it's that AI is going to
00:14take our jobs before it kills us. And when anyone says, hey, what if we had some regulation on all
00:19of this? Instead, what we get is a morally binding agreement between some big tech CEOs.
00:25But what seems to be forgotten in all of this, or at least rarely discussed for my taste,
00:31is how does China fit into this? Because their models seem pretty good and efficient,
00:38definitely more efficient than some of their U.S. counterparts. Are Chinese people, for example,
00:43worried about data centers or losing their jobs? Does Beijing think that these Chinese models are
00:48a risk to all of humanity? And what does the future look like if one of these two sides
00:54wins this race, whether it's China or the U.S.? Welcome to The Dip. This is the podcast about
01:00money, power, and consequences. I'm Cassandra. To help me get a lay of the land, I am joined by
01:07two brilliant guests. First up, Claire Chu on the line from Washington, D.C., where she's a senior
01:12China analyst at Jane's, a global defense and security intelligence company. She's also with
01:16the Atlantic Council's Global China Hub. Claire, thanks for joining us.
01:21Thanks for having me.
01:22And also on the line from New York City is Zhang Yuan Zoe Liu. She's a senior fellow for China
01:27Studies at the Council on Foreign Relations. Zoe, thanks for joining us as well.
01:32Thank you for having me.
01:33And Zoe, I want to start with you. Which AI chatbots are you already asking for vacation ideas
01:40or to spellcheck your emails? Is it the Chinese models or American?
01:45I think it's both. I use ChessGPT or Cloud when I'm here in the U.S. and I ask Yuan
01:55Bao or
01:56DeepSeek or Kimi when I am in China. And when I'm in China, I also found out that on Taobao,
02:04for example, all this e-commerce platform, I ended up asking AI which coat or which shoes I should buy.
02:13Yeah. That's a great life hack right there. But same with you, Claire. Which apps are already on
02:18your phone or desktop?
02:21So I'm a bit of a Luddite and I am a bit of an AI skeptic in my personal life.
02:27So I try to avoid
02:29using too much AI in day-to-day life with the exception of, you know, the platforms that have
02:36come pre-installed in all of my devices, like my iPhone summarizing my texts before I get to read them.
02:43Claire, what in your view is China maybe doing right now that the U.S. can learn from?
02:53I think one thing that China has done is stack several core competencies when it comes to China's
02:59industrial and technological development to be able to scale AI development. If we're talking
03:05about AI development in terms of accessing new markets and expanding global presence. So for
03:14example, China is powering a lot of data centers with renewable energy and building power generation,
03:22electricity generation projects near data centers, investing in energy resource intensive regions to be
03:30able to really scale effectively. And you see this domestically too. China's routing intensive AI
03:36workloads, such as model training to remote provinces like in inter-Bongolia, in Xinjiang,
03:42Gansu, where there is inexpensive land, the climate is cool, and there is massive wind and solar overcapacity.
03:50There's a lot to unbraid there. And I definitely want to get into each of those. But Zoe, you know,
03:55you've been following this power competition between the U.S. and China for a while. But when you look at
04:00the struggle between these two countries, where do you start? Are you starting with technology, politics, economics? Where?
04:07Yes, I start with FIRM and the most advanced large language models to begin with. But increasingly,
04:17I have also been following how the Chinese government is rebuilding a mission-driven financial system to
04:26fund this technology self-reliance or self-sufficiency in the broader context of strategic competition with
04:34the United States. And the reason I increasingly move beyond which country or which company is
04:41developing or deploying a more advanced frontier model is because fundamentally, a lot of these
04:49investments in high tech, whether it is AI, whether it's data center, a lot of this comes down to money.
04:57Do you have the money or not? Can you make the investment sustainable or not? And we all know that
05:02Chinese system is very much a bank-dominated system, which is by default not very good at financing
05:11technology and innovation. So increasingly, I'm moving to look at how China is rebuilding
05:17its mission-driven financial system to fund this tech competition.
05:22Claire, you mentioned kind of your skepticism of artificial intelligence in some ways. That is a
05:28mentality that we're seeing across the U.S., push back to data centers, push back to using AI for
05:35different maybe homework assignments or work assignments, that kind of a thing. But I'm
05:38wondering, what is the Chinese view of artificial intelligence when we're talking about the average
05:43person?
05:47I think overall, very broadly speaking, there is maybe some more acceptance of new and emerging technologies,
05:54including artificial intelligence across platforms, across apps, across services. Part of that is,
06:01I think, the Chinese government's emphasis on control over the technology as it grows and scales
06:06and accelerates. But part of it also, there has been a significant whole society AI adoption push across
06:13China. The official policy is to integrate AI into all sectors of the economy and society. So we're
06:19seeing, I think one of the best known examples is the Chinese mini dramas that have been proliferating
06:25across the Chinese-speaking internet that have been generated or AI generated. So overall, I think
06:33China, I mean, zooming out to the government does view some of the pressure that the U.S. is exerting
06:39on
06:39it in terms of guardrail discussions as almost over security, as in the U.S. is using sanctions to
06:49suppress AI technology companies. But there are, I think, different views across society depending on
06:57demographic and industry.
07:00Yeah, that makes me wonder, what about skepticism versus optimism outside of the U.S. or China? What
07:05are you hearing, Claire?
07:09A lot of governments, a lot of countries around the world are looking to more broadly invest in digital
07:15technologies, whether that means digital economy, like Zoe said, or digital government services,
07:22and overall upskilling and improving capacity and internet connectivity. And China has been really
07:28active across the globe, really, in every continent and most countries in earlier generation technology,
07:35everything from, you know, 3G, 4G, 5G development, and now interconnected devices, IoT,
07:42big data, AI is huge. And as Beijing moves away from traditional large-scale infrastructure
07:50projects that dominated the going out policy in the 90s and the first decade of the Belt and Road
07:55Initiative, companies are turning to more commercial opportunities, less sensitive opportunities
08:01that align with countries' own strategic frameworks. And a lot of countries' domestic development
08:07priorities, again, focus on China's core competencies, like renewable energy, advanced manufacturing,
08:13and digital technology. So there have been a lot of opportunities for China to enter new markets and
08:19to establish a foothold for AI and AI-derived opportunities and AI-derived, I want to say platforms,
08:29but it goes far beyond that, right? We're looking at a lot of countries using Chinese models to
08:34a building on Chinese models for new projects in order to reduce the cost and to reduce or to close
08:45the technical gap. So a lot of AI development in, for example, Kenya, a lot of AI projects are being
08:52built on the backs of Chinese models. And China is doing a very good job of partnering with the local
08:57institutions, universities, in order to build these platforms and frameworks for local creators to
09:07then, you know, build next-gen tech. Right. Zoe, react to this. We've seen China take technology in
09:15the past, solar panels, electric vehicles are two topics that we see often here in Germany, take this
09:20kind of tech and run with it. Are you seeing something similar like Claire has described?
09:26Yes, I think domestically we do see Chinese consumers and Chinese companies from big companies to small
09:34and medium-sized companies. The sentiment to embrace AI is, I'd say, perhaps more optimistic here than in
09:43the United States. And one anecdote that I constantly hear people saying is that the bosses, the management
09:52class are embracing this AI efficiency gain throughout the workflow. So they ended up tasking the employees
10:03more. And from the employees side, they haven't really fully learned how to best deploy AI throughout
10:13their workflow. So as a result, you know, at the current status of AI deployment,
10:19the sentiment of evolution or Neijuan ended up, you know, AI ended up accelerating that kind of anxiety. But I'd
10:29say one
10:29interesting observation from my recent trip to China about AI is that ordinary Chinese consumers and companies, they do not
10:38view US and China are engaging in this fierce AI competition, especially in the context of military
10:48or framing it from the arms control perspective. You know, recently, President Xi Jinping and President Trump
10:56had between May and September had two summits and people have all been speculating to what extent
11:04AI guardrail or AI governance or AI safety discussion the two countries may have. But from Chinese
11:11developers point of view, they do not view this from an arms control perspective. Instead, some of these
11:18technocrats, they view it as a total factor of productivity gain, you know, they are obsessed with
11:24TFP, they are obsessed with supply side economy. And then on the other hand, they also view this as
11:30one of the levers they can pull through industrial policies, hopefully to innovate out of China's
11:37structural problem. Now, the fundamental question is, is this going to be another round of China over
11:42capacity? I'd say, on the one hand, the Chinese government understand this is a too high stake of a
11:50race to not invest. And then on the other hand, they have also learned about over capacity. They have,
11:55this is why when you look at a total AI related investment, by 2030, I think the United States is
12:07going to invest over 10 trillion overall, across the entire AI related industry, including data
12:13center by China is far below that. And I think one of the fundamental concern from the Chinese government
12:19point of view is that because the technology is changing so fast, they do not want to over invest in
12:25the early stage, because the more you invest upfront, the bigger the risk in terms of
12:29strandled asset when technology moved beyond what you already have now.
12:34You know, Zoe, China has typically found ways around trade challenges. So how's it going this time
12:40with talk that we hear about limiting the highest, I don't know if it's quality or the fastest chips
12:46from making their way into China?
12:49Yeah, this is another interesting question, right? I'd say chips are are fundamentally goods,
12:56and goods are fungible. So despite a lot of these export controls, you ended up having
13:03stories about from shale companies or frontier companies, either in Malaysia or in Hong Kong ended
13:10up smuggling chips into mainland China. But I'd say fundamentally from Chinese tech people,
13:16or the tech world, on the one hand, they would want to have access to the most advanced model
13:22in order to accelerate their own RMD. But on the other hand, their fundamental purpose is not to be
13:29reliant on American best chips. As Claire said, they are redesigning how chips work. They are also
13:36trying to make the best use of the mature chips that China is best at making right now. So they
13:45are
13:46attacking the problem from two sides. One, gaining time by buying chips as long and as fast as they
13:53can buy it. And then on the other hand, they are really trying to innovate either by redesign or
13:59rebuilding the entire industrial ecosystem so that they eventually don't have to be reliant to buy from
14:07US companies or any other companies. And from my latest trip to China, I talked with people,
14:13people are very confident not only they are going to have their own different version of
14:18lithiographic machines and enabled by quantum, they are also going to very quickly have breakthroughs at
14:25seven nanometers chips. So I guess China is all ready to go.
14:30Zhang Yuan, Zoe Liu, I know you have to leave us shortly. So thank you so much for joining us
14:35today on The Dip.
14:36Yeah, thank you very much for having me.
14:40You know, Claire, we were just hearing from Zoe about some of these restrictions that have been placed
14:45on Chinese imports of certain kinds of technologies, including these ultra high-end chips. But I can't
14:52help but wonder, are we in kind of the training montage of this like sports, maybe Rocky movie? Is this
14:58just where these companies or these models become leaner, meaner, stronger AI machines?
15:07I mean, I think that's absolutely a part of the conversation that comes to R&D. But around the
15:14world, rather than compete on high-end superiority, Chinese providers are marketing affordability,
15:20adaptability of their open source and open way AI models, especially as many countries around the world
15:26with varying baseline levels of digital capacity and infrastructure, have elevated AI technology
15:33and talent as a national priority. And in doing so are inviting Chinese companies and institutions
15:39to introduce China's ecosystem, to support technical upskilling, to localize application of AI across
15:46industrial processes, government systems, digital services. And we're seeing this in every region,
15:53right? There are countries like Malaysia that have high or Indonesia that have high potential for
16:02electricity generation, domestic energy resources that can power the growth of AI and hyperscale data
16:07centers. These countries also often, I mean, in the case of Southeast Asia, happen to have a massive
16:13digital economy that has generated significant infrastructure demand. So we're seeing a lot of data
16:18data investments for both the US and China in this region. But because of the cost and the openness,
16:25I anecdotally have heard a lot of factories in Vietnam and Indonesia are using Chinese models already
16:31on factory floors to optimize workflow and processes. And we're also seeing Chinese companies really hone in
16:39on this region as a new forefront of development for certain technologies. And International just opened a
16:44new global development center in Kuala Lumpur this summer to focus on the development of an R&D for
16:49AI payments and fintech. We're also seeing a lot of expansion in countries with massive potential
16:56digital markets and market potential. So particularly countries like Nigeria, which leads the African
17:01continental internet users and overall just tech savviness. Huawei, for example, is working
17:08government agencies there to launch agentic AI cloud infrastructure across telecoms, public
17:14administration, internet services, and is also working with Nigeria's National Commission for
17:20Colleges of Education on an AI initiative to upskill teachers around the country. You know, we're seeing a
17:26lot of this upskilling focus also in countries like Egypt and Kenya that have huge digital markets and
17:32potential. And so it's quite targeted the way that China is engaging globally. But we see we see the
17:41same also in low infrastructure countries and in highly developed countries just with slightly different
17:46approaches. So Claire, for those of us who aren't so technically minded, I've open weight is a new term
17:53for me. How would you explain it? So open weight, salute to open source, it's a model that is publicly
18:01available for anyone to download to run locally. And the weights I can't get into quite as technically
18:08as some other colleagues refer to train numerical parameters.
18:13But boil it down for me here. What's in it for these Chinese firms? Is it just like Netflix, you
18:20know,
18:20a few years ago offering me a very cheap rate and then increasing the price every year? Is this kind
18:26of a
18:26long game when it comes to the balance sheet? Or is it more of getting in there first, like you've
18:31described in some situations?
18:35I think all of those can apply. But especially the latter, right, China sort of missed the boat
18:41on sort of first second gen technologies when it comes to, for example, software that was and
18:46is still now largely led by the US and by the West. And so Beijing's interested in sort of leapfrogging
18:54and leading in emerging and advanced technologies. And so we saw this summer, right, China established
19:00the World Artificial, the World AI Cooperative Organization in Shanghai, 29 founding members
19:07across regions of the world. And through that and other initiatives is working with countries to
19:12develop their AI standards and their AI models. I think Kazakhstan is a great example of this.
19:19They're a company that was also founded in a lot of activity this summer called, what is it called?
19:27I think it's called QAI in partnership with a Chinese unicorn company that is called Zero One AI,
19:35which is backed by Alibaba and Xiaomi. So this new joint venture with the government of Kazakhstan
19:39will operate Kazakhstan's national AI platform under the leadership of Kazakhstan's former vice
19:45minister of AI, Dmitry Munn. And just a month before, Kazakhstan had joined the US-led Paxilica
19:51initiative. But China is still very, very active in Kazakhstan's emerging AI governance.
19:59Its CEO, Zero One AI CEO, joined Kazakhstan's Council for AI Development under the president,
20:06and also helped develop an AI-generated board director for Kazakhstan's sovereign wealth fund
20:13that was developed using a Kazakh language model called Allum LLM, which was also developed by this
20:19Chinese unicorn company. And we'll be able to vote on the fund studies. So there is a lot of potential
20:26that
20:27I think the US and China aren't fully aware of, but would like to position themselves to be able to
20:35explore in the future.
20:36But I'm glad you brought up Kazakhstan, because I think that that country is an interesting example,
20:41because it's my understanding that this was an example of a country that was trying not to pick
20:46sides between the US and China, right? And that didn't really seem to go so well. From what I
20:52understand, the US seemed to turn off the lights when it came to some of this technology. Besides the
20:57successes that you've already outlined, what does this example of Kazakhstan maybe mean
21:02for the future of this race between the US and China?
21:09I think that's very much ongoing, right?
21:14There's no really win, I think. I don't know if there's a predetermined end to this race or to
21:20competition. It very much will be ongoing. At times, the US will lead in some dimensions,
21:26China will lead in other dimensions. And I think the balance is visible across all of these regions.
21:33Again, like Zoe mentioned, the US has invested massive amounts of capital, including Southeast
21:40Asia, where China is very active. The US is also very active in Kazakhstan and is investing in a
21:49hyperscale data center in, I think it's called Data Center Valley, backed by Firebird, the US company,
21:55and NVIDIA. So I don't know if we can really conclusively say that specific countries are
22:04adopting China's model entirely or the US model entirely. But I am particularly interested,
22:10at least in my research, in the global south and in, I think it's sub-Saharan Africa,
22:14in North Africa, and these countries that are, I do think, largely overlooked by the US
22:19in terms of investing in the talent, investing in the local ecosystems,
22:24and really building that familiarity and visibility to have continued to expand in those markets and
22:33with those stakeholders.
22:35Earlier, you mentioned some of the energy projects that are being built next to data centers and so
22:41on in, for example, Inner Mongolia, a very rural part of China, right? This reminds me of what we're
22:47seeing in the US where data centers are being built in very picturesque and sometimes rural
22:53parts of the country. But in the US, there's a lot of pushback, people saying, you know,
22:58we don't want these data centers here, we don't see the benefit of them being here. Is there a similar
23:04pushback from your average person in Inner Mongolia or other rural parts of China too? Or
23:11how is this kind of buildup that we're seeing from the Chinese government being received when it actually
23:16is affecting people's livelihoods or neighborhoods?
23:22I mean, there is a lot of land in Xinjiang and in the Mongolian Gansu and some of these
23:28largely desert regions that China is building these large scale facilities.
23:32And so, you know, the setup hasn't been official to China. If there is massive pushback or if there
23:44has been discontentment from local populations, I think China has set itself up and has equipped
23:55its local government services to address those concerns before they can create real obstacles
24:02for tech companies and for their ultimate commercial goals.
24:08You know, I wanted to talk a little bit more when we're talking about this race,
24:12which you make a great point, right? Like, in order for it to actually be a race,
24:16there should be like an endpoint. And there isn't going to be an endpoint in this,
24:18it's going to be an ongoing competition and conversation. But one of the parts of this
24:24is going to be AI governance. And we saw China, Chinese President Xi Jinping and President Donald
24:30Trump having what seemed like a very friendly meeting from my vantage point in Washington last
24:37month. I'm wondering, you know, what does this tell you about who's going to be operating
24:43the future operating system of, you know, the global digital economy? Are these two going to be
24:48working together? Or are we going to see a little bit of maybe a simmering struggle in your view?
24:55I think this will sort of be an ongoing struggle, much like other technology sectors, because
25:00ultimately, the two really are so intertwined, right? We're seeing companies like Visa and MasterCard
25:06are working with Chinese Spintech companies to set up auditing standards. And this is at very,
25:11the very large multinational companies too, with other operating across the world.
25:18Overall, it's hard to say, especially sitting in Europe, where you are, I know there is emerging
25:24distrust of Chinese hardware, of US software. So I do think there are other voices in this
25:29conversation too, who don't wish to be beholden to one or the other. The concept of strategic autonomy
25:36has been really, I think that the phrase of the last year or two, for a lot of these middle
25:41powers
25:42countries and countries really that aren't the US or China. So I do think there are implicit guardrails
25:49in place by virtue of everyone not wanting to be boxed in by one single system or one single
25:56set of standards. When that said, a lot of the soft initiatives that China has launched
26:02do serve to increasingly embed and integrate Chinese standards and technology without, say,
26:10formal partnerships or allegiances or alliances, right? This means a lot of the partnerships to
26:16train government employees on AI. This means university partnerships and vocational training
26:22across the world in different countries, working with Chinese companies, Chinese institutions,
26:27to teach students to operate, you know, Chinese systems, and to familiarize them for the fee for
26:37support to support Chinese investment investment projects in the future. So it really remains to be
26:44seen. And without a clear definition of competition, like we've been talking about, it's really hard to
26:51say where this goes.
26:52Claire, you've been so generous with your time. One last question, looking into your crystal ball,
26:56do you think in five years time, we are all going to be using more Chinese AI or American AI?
27:04I think it depends on where you live in the world. And at least for myself,
27:10most likely you at US, you know, companies, USAI, but for a lot of the world, likely not. And not
27:17just
27:18in your personal capacity, but in your workplace, just moving about in the world, your digital payment
27:23systems are very likely to be backed by Chinese, Chinese AI, and Chinese hardware too. So it's,
27:32it's really hard to, it's really hard to know, it's hard to make those choices because of how
27:36interconnected both of our digital ecosystems are.
27:40Claire Chu, thanks so much for making the time.
27:44That is it for this week on The Dip. Thank you very much for tuning in. And don't forget to
27:48leave
27:48a comment or a question. Any feedback is welcome wherever you're listening to this or watching it,
27:53of course. And if you're not subscribed in the podcast app of your choice, go ahead and do so.
27:58Just look for The Dip. But until next time, thank you very much for watching The Dip.
28:03Bye!