WSJ What’s News - How China’s AI Power Threatens Silicon Valley

Episode Date: January 18, 2026

It’s been one year since Chinese AI developer DeepSeek released an experimental large language model that shocked the tech world with its advanced capabilities, despite strict chip import restrictions. WSJ Senior Global Correspondent Josh Chin and Oxford Analytica technology analyst Tatia Bolkvadze discuss how China’s AI prowess has only grown in the past twelve months, something that is now challenging Silicon Valley’s pricing power, and becoming a bone of contention in the U.S.-China trade war. Luke Vargas hosts. Further Reading:  The AI Cold War That Will Redefine Everything China’s Alibaba Links Qwen AI App to Vast Consumer Ecosystem The Row Over South Korea’s Push for a Native AI Model: Chinese Code China’s DeepSeek Unveils New AI Model That Could Halve Usage Cost Silicon Valley Is Raving About a Made-in-China AI Model Chinese AI Developers Say They Can’t Beat America Without Better Chips Learn more about your ad choices. Visit megaphone.fm/adchoices

How are the U.S. businesses of Philip Morris International invested in America? We're invested in advancing science, giving adults who smoke better options. We're invested in American manufacturing, helping local economies thrive. We're invested in community, supporting military veterans and their families, disaster relief, and economic empowerment. Because we're proud to be invested in America. See how at uspMI.com. Hey, what's news listeners. It's Sunday, January 18th. I'm Luke Vargas for the Wall Street Journal, and this is What's News Sunday, the show where we tackle the big questions about the biggest stories in the news by reaching out to our colleagues across the newsroom to explain what's happening in our world.

On today's show, one year ago, Chinese-made AI model Deepseek shot the world. Earth-shattering developments in the AI space. Deepseek said its chatbot was created with less than $6 million. The app has become the most downloaded free app in the U.S. Technology shares on Wall Street have fallen sharply. It's mind-blowing, and it is shaking this entire industry. Since then, China's AI prowess has only grown in ways that could change global tech norms, challenge Silicon Valley's pricing power, and become a bone of contention

in the U.S.-China trade war. This week, China's AI strategy and what it means for the world. Let's get right to it. Joining me to unpack China's approach to AI and consider the economic and geopolitical impacts of its growing adoption. I'm joined by a pair of great guests. First off, Tatia Bulkwadza is a technology analyst at Oxford Analytica. And Josh Chin is the Wall Street Journal's senior global correspondent in Asia. Josh, in a nutshell, what's China's AI strategy? China's AI strategy, as opposed to the U.S. AI strategy, China's is much more state-driven.

Things really picked up in early 2005. People remember the unveiling of Deep Seeks big frontier model called R1, which was basically nearly as powerful as the top models from Google, Open AI, Anthropic had produced much more cheaply. And this was such a surprise to some people in the U.S. And really started to wait people up to the potential of Chinese AI, but also to leaders in Beijing. And it was at that moment that the Chinese Communist Party really started to feel some hope and some confidence that they could compete. And that unleashed a whole guiser of government support. Tatia Deepseek is still very much around, but there's a different case study in the form of an AI model suite from Chinese company Alibaba that's Kwan that you and your colleagues have been drawing attention to lately that showcases sort of another element of China's AI strategy. Tell us about that.

All right. So Kuen did not get as much media attention in the West. as deep-seek and some of the other kind of smaller Chinese AI models did in the beginning. Quinn was launched in 2023 for the first time, but this sort of deep-seek moment really forced Alibaba to step up innovation and increase funding in AI. So Kwan managed to get pretty significant global footing over the past year. It has been downloaded more than 600 million times. It's been used pretty widely internationally, not just within China.

its customers now count companies like Airbnb. And also last December, Quinn became the most downloaded open source AI model. So beyond just sort of showcasing Alibaba's success in this domain, the success of Quinn and DeepSik and this group of other Chinese open source AI models is really showing that the strategy of making models cheap and freely available is working for Beijing. You mentioned adoption in the West. The U.S. venture capitalist Martin Cassata was quoted. and the economist saying 80% of U.S. entrepreneurs that he's coming across are using Chinese-made

AI models. Josh, are we getting kind of a Huawei 5G redo here where this very alluring technology gets the world to sign into bigger Chinese systems? Yeah, that's definitely a comparison people have made. When 5G was being developed, the U.S. actually really fought hard to keep Huawei, which is China's big telecom equipment manufacturer from installing 5G. equipment in the developed world. But that didn't actually keep Huawei from dominating globally. And actually, we're now at a point where Huawei owns many of the essential patents for 5G and has built out networks throughout the developing world. That is a success story that Beijing is constantly looking to replicate. And you are certainly starting to see Chinese AI models

being adopted around the world because they're so cheap. Right. Absolutely. I have to agree with Josh here. And I think it's important to note also that Huawei's previous success in the Global South is also paving away for Chinese AI models as well, because if you look at Huawei sort of phone adoption in the global South, that's enabling the adoption of Chinese AI models like DeepSik, because Deepak is the default AI chatbot on Huawei phone. So we've seen a very rapid adoption in countries like Iran, Ethiopia, Niger, and others. But also the open source strategy that we mentioned before, China is using this as a way to advance its image as the champion of the developing world. It's painting open source strategy as a way

to democratize access to AI. At the 24 G20 summit, President Xi famously noted that AI should not become a game of rich countries. And since then, Beijing has really stepped up convening international conferences, which are catered specifically to the global south. And you see a lot of the members here entering into partnerships with Chinese tech and AI firms, adopting their AI models, which are again open source. And these countries are later able to sort of develop their own national models off of Chinese products. All right. Chinese AI there differentiating itself from other products out there in the market and Beijing promoting that. We've got to take a short break. But when we come back, how that's all going over with America's tech giants. Stick around.

As companies seek to close growing gaps in skills and talent, Deloitte U.S. CEO Jason Garzadas believes it's important for organizations to understand their baseline of skills. There's so many organizations that can't ask and answer the fundamental questions about how much computer science or data management skills do I have or AI development skills in a given domain. By performing a skills inventory, leaders can truly understand where their efforts should be focused. Being blind to those gaps is the real miss. Visit Deloitte.com to learn how you're in. enterprise can help successfully cultivate talent.

Tatia, in addition to these Chinese AI companies, potentially opening up new markets for themselves, it sounds like they may be chipping away at the pricing power, big U.S. tech firms have been commanding. Right. I think the success of Chinese AI models, which are so low cost, in many cases free, is really exacerbating the really deep fears about. the AI bubble that's been forming in the US tech industry for a while. So beyond sort of pricing concerns about how to make your products more competitive in the global

market, I think a lot of these firms are forced to scale up innovation, pivot to other domains of AI. We've seen the rise of world models, which have more sort of practical applications. I think one big advantage of Chinese AI models has been that they reflect Beijing's overall strategy of utilizing AI with their industrial and economic growth goals, and that's been hugely appealing to not just developing countries, but also many enterprises. And I think companies in the West are now trying to make their products also more practically applicable. This really is a big divergence you're seeing in the world of AI right now. The default view in Silicon Valley is that

large language models are the path to superintelligence. That's what meta is pursuing. It's what Anthropic has been pursuing Google. But there are a bunch of skeptics, LLM skeptics, who think that you need to actually expand beyond large language models if you really want to achieve superintelligence, something like human intelligence, but more, right? You need actually models that can simulate the real world, including the physical world. And this is one area where China really does feel like it has an advantage because it has these factories because it builds so much stuff and already has robots in wired

factories, making things, collecting data. And so that's one area that a lot of people in China, I think, has immense problems. And if I understand this correctly, this is not just some philosophical debate about whose vision of the science fiction future is correct. The difference between the Chinese and U.S. private sector focus here maybe boils down to who the customers of AI will be, who might be paying for AI services in the future. Is that right, Tatia? Yeah, I think that's true. And this technology has a lot more avenues of growth for businesses. It promises to optimize decision-making. It's going to have a lot more benefits within sort of public sector decision-making as well. So real-life public applications are

better. The concerns about potential hallucinations are less as well. And I think world models also promise to address some of the concerns about data availability as well in the future. So this is, I think, one of the biggest trends to watch in 2026, and perhaps this will be the next stage of AI competition between the U.S. and China. I'm envisioning here, I don't know, the Egyptian government wanting to model the Nile River and how it would respond to, let's say, a bad harvest. Is that kind of what we're talking about here with world models, something with a sovereign AI, potential use case?

Yeah, absolutely. I think that could be one use case for governments when it comes to world models. It's also important to note that autonomous vehicles are powered by a type of world model. So China has a huge advantage in that domain already. I mean, China's EVs and autonomous vehicles have been vastly successful worldwide. So I think it is expected that given the fact that world models generally align well with China's industrial development model and larger vision of AI, there's going to be a pretty significant state support directed to this domain in the coming years. Josh, lots of advantages here for this Chinese approach.

What are the major challenges to it that you're hearing about? The biggest challenge for China is chips. This is an area where the U.S. has an immense lead. It owns all of the intellectual property, and it controls the supply chain for the sort of super high-end, you know, really cutting-edge chips that are used to train AI models. I talked to a former U.S. official who was behind efforts by the Biden administration to ban exports of U.S. chips to China who was saying it's a really difficult technology gap for China to close. They're probably as much as a decade behind. Just this last weekend, you had a bunch of

top Chinese figures in AI, including Justin Lin, who's Alibaba's lead for the Quinn project, who said that China won't be able to compete if it doesn't solve the chip gap. Though that edge in chips, I don't know, could it be a double-edged sword? It's really dependent on chip performance, especially those highest-end U.S. chips continuing to just achieve bigger and bigger gains over their Chinese counterparts, but if that were to sort of plateau or top out, that's a risk for Silicon Valley. That's certainly true. We don't know for a fact that these chips are going to keep getting better and better.

It's also a possibility that there's a new type of chip architecture out there that has yet to be invented that China may invent, right? There's just a lot of huge unknowns in terms of where this technology is going to go. For the foreseeable future, there definitely are some big challenges for China, and not just with chips, but also with financing. U.S. companies have much, much more money available to them. Yeah, Josh, just jumping in of a recent Morgan Stanley report that tried to assess who could win this AI race, gave credit to China and maybe an edge overall citing its state resources, its ability to scale. But for the U.S., as you mentioned there, not just higher salaries that help boost the talent pool, but specifically private sector money that can come in off the sidelines.

Tatia, we haven't spoken about it yet, but probably a big X factor in all of this is what Washington ends up doing. Right. Washington's policy has been somewhat in consideration. I would say over the past few years, under the Biden administration, we had the AI diffusion rule, which the Trump administration rescinded. So we saw the Trump administration allowing the shipment of H-200 chips to China made by Nvidia, with a caveat that Nvidia has to enter into a revenue-sharing agreement. So perhaps this is a new model, new approach as to where Washington is going to go next, sort of more targeted relaxation of. of export restrictions, which give China some access to U.S. technology. And perhaps Washington thinks that this is a way to keep China addicted to U.S. hardware and U.S. software.

Josh, I think Tantia there has given us a what-to-watch for in 2026. What do you be paying attention to in the coming months? One big question is to what extent Beijing will allow Chinese AI companies to buy these chips. They're trying to support Huawei. Huawei has its own chips. It's trying to build. And Huawei has said that it can match the computing power of American data centers by just stringing together huge numbers of sort of lower powered, less efficient chips. It's a strategy they call swarms to beat the Titan.

The Chinese government is committed to that. So it'll be really interesting to see exactly how many Nvidia chips they let in. China has been a huge market for Nvidia. There is a lot of interest in China in H-200 chips, particularly among private. private companies. Again, the question kind of just comes down to how much of that is Beijing going to allow to come to fruition and in what ways. Josh Chin is the Wall Street Journal's senior global correspondent based in Seoul. Josh, thank you so much. Always a pleasure. Thanks, Lou.

And Tatia Bokwadza is a technology analyst at Oxford Analytica. Tatia, thank you. Thank you for inviting me. And that's it for What's News Sunday for January 18th. Today's show was produced by Daniel Bach with supervising producers Sandra Kilhoff and Melanie Roy. I'm Luke Vargas, and we'll be back Tuesday morning with a brand new show. Until then, thanks for listening. Millions of Americans use credit cards to purchase goods every day. The Durbin Marshall credit card mandates put the secure transactions you rely on at risk, leaving you vulnerable to hackers and

foreign cybercriminals that want your data. Hackers win. You lose. Guard your card before it's too late. Tell Congress your data security is not for sale. and oppose the Durbin Marshall credit card mandates. Paid for by Electronic Payments Coalition.