Will China win the AI race? You asked, we answered

Question asked by: Curio
Question: The internet search market devolved over time to near complete dominance by Google despite a number of early rivals. Is the AI market likely to end up with a clear, dominant winner, or will there continue to be multiple high performing AI models in there market for an extended period.

Answered by: John Thornhill
Answer: Interesting question. It depends. I suspect that only a handful of companies are going to dominate the field at the infrastructure layer because of the colossal costs involved. But there will be many thousands of AI-enabled companies that will grab the technology and apply it in productive ways.

Question asked by: Anonymous
Question: Despite calling it a race, where are potential mutually beneficial areas for international collaboration in AI?

Answered by: Eleanor Olcott
Answer: Another field I cover is biotech, and there’s huge positive potential for the US and China to work together on using AI for drug discovery and development. China is becoming an increasingly important hub of biotech innovation, especially in the rare disease space, where there’s still so much unmet need.If AI can genuinely cut the cost and time of developing new drugs — which early signs suggest it can — then China will almost certainly play a major role in that progress. And if the US continues to pull back or cut ties with Chinese biotech, it risks missing out on that wave of innovation entirely.

Answered by: John Thornhill
Answer: One of the most interesting features of this AI “race” is how open most of the research is. That means the field as a whole is advancing very rapidly. Some of the biggest gains from AI may well come from its application across the public sector. There is huge scope for the sharing of best practice and international collaboration in that area.

Question asked by: Paul Thind
Question: Why does anyone think that there will a winner? AI will have millions of applications and enhancement. Each new development will be countered, much like in the nuclear arms race. There will be no silver bullet.

Answered by: John Thornhill
Answer: Interesting point of view. Maybe everyone will be a winner if AI is adopted as widely as you suggest. But the history of the Industrial Revolution suggests that it will take a long time for the full benefits to accrue. In the meantime, there may well be lots of people who lose their jobs.

Question asked by: Anonymous
Question: Are AI researchers and companies in China as obsessed with ‘AGI’ as in the US? Are there any philosophical differences in how AGI is understood?

Answered by: Eleanor Olcott
Answer: Not at all. Nowhere near the same level of obsession you see in the US.Sure, companies like DeepSeek and Alibaba talk about AGI, and they’ve positioned themselves as China’s most advanced labs. But the term just doesn’t have the same buzz around it that you get in Silicon Valley. Part of that is cultural (or political). The heads of Chinese tech companies aren’t nearly as vocal or freewheeling as their American peers — you don’t hear them popping up on popular podcasts chatting about “the future of humanity.”But it’s also practical. Most Chinese AI companies are focused on the business case, not the philosophy. They want to land customers, ship products, and make money — not spend time debating what happens when machines become sentient.

Question asked by: Masif Group
Question: Where is UK in this race, with its defence sector, a major private-sector customer for green energy & power for autombiles and their AI driven automation, claims leadership, while their private contractors shift to China for production and AI integration to cut costs?

Answered by: John Thornhill
Answer: The UK has a very strong history in computing (Babbage, Lovelace, Turing etc). London is also home to Google DeepMind, one of the world’s leading AI research labs, which has been an incredible magnet for global AI talent. The UK also has 3 of the world’s top 10 universities with strong AI departments. The open questions are: whether the UK can build enough AI infrastructure quickly enough? And whether its most promising AI startups can raise enough growth capital to compete globally? Open questions.

Question asked by: chow ah beng
Question: In China’s journey towards AI dominance, what are some lesser- known but critical pain points that the country would need to successfully tackle?

Answered by: Eleanor Olcott
Answer: One of the biggest and maybe least understood (just because it’s so opaque) challenges is semiconductors. China still doesn’t have all the ingredients needed to make truly high-end chips on its own.No country completely owns the chip supply chain, but US export controls have effectively cut China off from the most advanced nodes at TSMC and the cutting-edge manufacturing gear from ASML. That’s a huge problem for Beijing’s AI ambitions, especially as it pushes domestic companies to switch over to locally made chips.Huawei’s upcoming Ascend 950 will be the first chip fully fabricated in China, whereas many of the earlier 910 series were actually made at TSMC. So this next chip generation will be a real test of how far China’s homegrown semiconductor ecosystem has come — and how much those pain points still bite.

Question asked by: Anonymous
Question: US and China are using different approaches in AI development. What do you expect the AI development impact to both countries in coming 5 years?

Answered by: John Thornhill
Answer: It’s true that US tech companies are mostly developing proprietary closed-weights AI models while China is mostly developing more adaptable open-weights models. OpenAI’s Sam Altman suggested that the US might be on the “wrong side of history” in that competition. Expect to see a lot more US open-weights models in future

Question asked by: Andre
Question: Do you believe that the copyright framework can and will survive the generative AI race?

Answered by: Eleanor Olcott
Answer: I think that ship has already sailed. It’s a total whack-a-mole situation.OpenAI might tighten its rules around generating images in the style of real artists — and that’s a good thing — but there are a million other platforms out there offering exactly the same service with no such restrictions. It falls to consumers to seek out art, literature and music to ask what produced what they are consuming. It remains a massive open question whether in the future people will care whether it originated from a human or a machine.

Answered by: John Thornhill
Answer: There’s no question the copyright framework is under massive threat from AI. But the AI companies themselves desperately need access to fresh human-generated content given they have already ingested the entire internet. And content creators desperately need a new way of making money. There must be a market solution in there somewhere.

Question asked by: FTinthedoor
Question: Why isn’t Lidl (Schwartz group) making models like Meituan is? Why is Europe so behind

Answered by: Eleanor Olcott
Answer: It’s a bit of an apples-to-oranges comparison, to be fair to Lidl. Meituan is a tech platform — a food delivery and services app — whereas Lidl is a retailer. So it’s not really in the business of building large language models, and neither are China’s own discount supermarkets, as far as I know.That said, your question gets at a really interesting point: in China, there’s huge pressure on tech companies — from investors and from Beijing — to show they’re making progress on AI. It’s become a kind of tech status symbol. The same dynamic exists in the US, where all the big players are racing to prove their AI credentials.Whether any of those models actually translate into growth is another question entirely.As for Europe, to understand why it’s lagging, you have to go back a bit — to the fact that Europe never really built the modern internet giants in the first place. Without that tech base, it’s naturally harder to jump straight into building frontier AI models.

Question asked by: Franco Spoltore
Question: Don’t you think the Draghi and Letta reports to the European Commission have clearly shown that without building an effective European federal State, via a profound reform of the existing treaties, Europe has no chances to compete with continental powers like USA, China and Russian federation?

Answered by: John Thornhill
Answer: Interesting debate that was explored in Martin Wolf’s latest column. It’s an open question whether Europe does better in periods of fragmentation or consolidation. What would massively help European AI startups, though, would be the completion of the single market, the introduction of a “28th regime” to enable them to operate more freely across the EU, and capital markets union, which could help provide more growth capital.

Question asked by: Anonymous
Question: While there has been a lot of talk regarding the Chinese export controls on rare earths, from discussions with C-suite, earnings calls and results, it seems that they are not having a noticeable effect on companies, at least from an investor’s “outside” perspective. Thoughts?

Answered by: Eleanor Olcott
Answer: Oh… these export controls ABSOLUTELY had an impact on companies when they came in. Several automakers in the US and Japan had to halt production. Since the pandemic, all the major companies that rely on China for some part of the supply chain have been talking about the importance of diversification for critical inputs and components. It’s amazing how few of them actually did it…

Question asked by: Hrsh
Question: How far away is AGI? If there is no AGI in near future, then what would winning mean?

Answered by: John Thornhill
Answer: Nobody knows. Some of the leading AI labs suggest it will be this decade. But no one can really agree what AGI means anyway. Far more important, I think, will be how powerful AI is applied in some narrow and critical economic and military domains.

Question asked by: Bardamuzes
Question: Isn’t winning the AI race a big risk for the US? AI is probably most likely to replace (high-end) service jobs. But this is essentially much of the US economy. While manufacturing shouldn’t be impacted in the same way? It might even make Chinese factories even more efficient

Answered by: Eleanor Olcott
Answer: It’s a really interesting way to look at it. I’ve started to think China might actually have an advantage in being behind.Here’s why: while Chinese factories are racing ahead with AI — automating production lines to offset an ageing workforce — the white-collar side of the economy has been much slower to adopt it. Meanwhile, in the US, the service sector — law, banking, consulting — is diving headfirst into AI tools.When I talk to friends in those professions, it’s clear the US is already much further along in adoption. And with that, we’re also starting to see the first big wave of layoffs linked to AI productivity gains.So there’s a world in which Beijing is quietly sitting back, watching the West roll AI out across every corner of the workforce — and thinking, “You go first.” If AI ends up disrupting jobs, education, even relationships, China’s slower, more controlled approach could actually look pretty smart in hindsight.

Answered by: John Thornhill
Answer: You’re right that the impact of AI is most likely to be felt among highly paid white collar workers. That means the geographic distribution of economic gains and losses is probably going to be very different from previous industrial revolutions. In other words the rich coastal cities in the US will be affected more than the Rust Belt. Many of the jobs that are affected – and much of the wealth that is generated – will be in places like NY, San Francisco and Los Angeles. And AI will certainly improve the competitiveness of China’s factories.

Question asked by: Mirko Gaslini
Question: What are the best options to solve the energetic issue regarding the AI data center energy cost to train and run AI?

Answered by: John Thornhill
Answer: Energy costs are going to be a critical component of the competitiveness of any country’s AI sector. With its massive investments in renewable energy, China has the edge on most others in this regard. High energy costs are going to be a particular challenge for many European countries.

Question asked by: LonMan
Question: As I understand it ASML is hugely important to AI in terms of chip manufacturing. Could either the US or China develop their own rivals?

Answered by: Eleanor Olcott
Answer: China is certainly trying to build its own advanced DUV and EUV technologies. China already builds DUV machines (leading company is SMEE) but it’s pretty mature technology. We have reported on efforts to build more advanced DUV machines, but we’re yet to see if these machines work on a factory floor.

Question asked by: sbelh
Question: Why is Russia not in the mainstream AI conversation?

Answered by: John Thornhill
Answer: The Soviet Union was one of the pioneers of early AI research. But Russia has fallen steadily behind in recent times. There are plenty of brilliant Russian AI researchers in Silicon Valley but Russia itself has suffered from a big brain drain of tech talent. That makes it all the more ironic that Vladimir Putin once said: “Whoever becomes the leader in this sphere will become the ruler of the world.”

Question asked by: Zelia
Question: Who do you consider the biggest Chinese competitor to Open Ai today?

Answered by: Eleanor Olcott
Answer: Alibaba is generally viewed as the most competitive LLM company in China. As China’s leading cloud company, Alibaba has invested heavily in AI infrastructure, it has many of the leading open-source models and is beloved by developers. It also has access to wealth of data from existing apps like Taobao, which make it a pretty formidable competitor in China.

Question asked by: Anonymous
Question: White House AI czar David Sacks recently stated that Huawei’s Cloud Matrix shows that China can compete against the US even without parity in AI chips by networking Ascend chips together. How realistic is this concern?

Answered by: Eleanor Olcott
Answer: Sacks has been very outspoken about the risk of the US losing its AI edge — and he’s right to flag the threat from Chinese AI exports. Chinese companies are already pushing cheap, capable models overseas and undercutting American labs.But when it comes to AI chips, we’re nowhere near a point where Huawei can rival Nvidia outside of China. Huawei’s Ascend chips have some serious limitations — especially when it comes to their software stack, which still lags far behind Nvidia’s CUDA ecosystem. The company’s throwing resources at the problem, but this isn’t something brute-force engineering alone can fix, even though that’s usually Huawei’s strong suit.It’s also worth considering who’s doing the talking. Some American policymakers — and even a few US chip founders who’d love to sell to China — have a vested interest in playing up Huawei’s capabilities. That doesn’t mean Huawei isn’t improving, but for now, the idea that it’s on the verge of overtaking Nvidia outside of China is more hype than reality.

Question asked by: DHIC
Question: It seems increasingly difficult for foreign nationals in China to visit AI companies in China or engage with execs who travel less frequently. In these conditions, how do foreign journalists understand and gauge the potential of Chinese AI companies and the comparative position of their tech?

Answered by: Eleanor Olcott
Answer: Definitely less open than a decade ago. As a journalist covering AI, that means official access is pretty limited — it’s harder to get inside the companies or talk to people on the record.But you can still build a clear picture of what’s going on by talking to customers, suppliers, investors, and employees. Being on the ground in China helps a lot — you pick up things that don’t make it into official statements.And interestingly, the open-source nature of China’s AI industry actually helps reporters too. Many labs are publishing their models openly, and companies like DeepSeek are releasing incredibly detailed technical reports about their training runs. That transparency gives us a surprisingly good window into how capable these models really are — even if we can’t just walk into their offices. The one area where that doesn’t apply at all is semiconductors — that’s a completely different story. Chip development is where opacity really kicks in.

Question asked by: Algats Finest
Question: While most people talk about AI technologies, what impact does AI investment have on residual areas of non-AI technology (both positive and negative)?

Answered by: John Thornhill
Answer: Interesting question. Economic historians suggest that in previous technological booms investors have made the most money out of complementary investments rather than out of the investments in the technologies themselves. In other words investors made more money out of the factories that were built to exploit electricity than from electricity power plants. It will be interesting to see which related fields flourish now.

Question asked by: Hugh Ebbutt
Question: Which European country will lead in AI

Answered by: John Thornhill
Answer: For the moment, the UK is the third AI power behind the US and China in terms of number of AI companies and investment. But the French AI sector is developing very fast. And several other countries in Europe, including Sweden, Germany and Switzerland, are also becoming more significant players.

Question asked by: Fong Ma
Question: Are there any other countries in this so called race or are both the US and China that far ahead now?

Answered by: John Thornhill
Answer: The US and China are certainly the two dominant powers in AI at the moment. But the UK, France, India, Canada, the UAE and several other countries are fast developing very vibrant AI ecosystems.

Question asked by: Anonymous
Question: What do panellists think will cause the ‘bubble’ to burst? If you don’t think a bubble is building (blowing?) do you think wider foreboding in the market is justified and if so why?

Answered by: John Thornhill
Answer: It is in the nature of journalists to predict nine of the last five market crashes. How long this AI bubble will last – and what might provide the pin – are (for this journalist at least) impossible to say.But I think two things are happening simultaneously: we are seeing both a long-term investment boom in frontier models and AI infrastructure by the Magnificent 7 tech companies and a short-term speculative frenzy in both the public and private markets about the potential impact of the technology.Nobody knows where that boom ends and the bubble begins. But SoftBank’s decision to sell its $5.8bn stake in Nvidia and go all in on OpenAI suggests that Masayoshi Son still thinks the bubble still has some way to inflate.

Question asked by: Anonymous
Question: Can the US overtake China in energy production? Do they have the capacity?

Answered by: Eleanor Olcott
Answer: Not anytime soon. China’s energy capacity is still growing at a staggering speed. As energy analyst David Fishman put it to me recently: last year alone, China’s production increased by more than Germany’s entire annual demand.My colleague Ed White has a brilliant piece on China’s rise as an “electrostate.” For me, living in Beijing, it means far cheaper electricity bills than when I was in London!https://www.ft.com/content/f86782fa-9f2e-448a-b710-29e787dc9831

Question asked by: YB
Question: Since there is a block on asml euv equipment to china, won’t “the west” always keep an advantage. Yes, the Chinese have proven that they can go down to 7nm, but at very small scale. What is your opinion?

Answered by: Eleanor Olcott
Answer: It’s true that China’s chipmakers face big challenges. Local 7-nanometre processes mean lower yields and higher power use than Nvidia’s chips made on TSMC’s N4 node.That said, China’s fabs are scaling up fast, and the industry is very good at finding workarounds.Your question hints at a bigger point. The entire Chinese AI industry has been devising creative solutions to work around its limitations, whether through clever engineering or even more clever lawyering to find loopholes in export controls. So yes, the gap exists, but it’s less decisive in practice than you might think.

Question asked by: In the comments, no one knows you’re a dog
Question: Given China’s:
– More holistic approach to AI rather than the ‘hotrod’ hardware approach in the US – Electricity generating capacity surplus rather than the constraints elsewhere – High performance on open source LLMs (currently all the top 5 models)
What’s the limiting factors for China AI?

Answered by: Eleanor Olcott
Answer: The big bottleneck now is data centre capacity — specifically, local ones with the right chips to meet exploding demand for inference.Export controls didn’t really stop China from training this generation of models, lots of the big labs used overseas data centres, especially in Malaysia. But to actually run these models at scale, they need domestic facilities.That’s where things get tricky. China still has lots of Nvidia chips, either bought before the bans or smuggled. Still the US export controls and now Beijing’s own restrictions on lower-end Nvidia chips mean that Chinese firms have no choice but to start using local chips for inference. They work fine but are less efficient and more power-hungry, meaning slower performance and higher electricity bills.

Question asked by: G.O.M
Question: Could the AI competition between China and the USA be defined as a race? What does winning the AI Race look like?

Answered by: Eleanor Olcott
Answer: This is a great question. “Race” suggests everyone’s heading for the same finish line — and that’s not the case.In the US, the big labs are all sprinting toward AGI — artificial general intelligence. In China, the competition is more about building the best open-source models, optimising for cost and efficiency.So it’s really two parallel races, with very different goals, but they influence each other’s strategies along the way.

Answered by: John Thornhill
Answer: The implication of your questions is that comparing the US China competition to a race is not very helpful. You’re right. We don’t know what the finishing line looks like. It could, arguably, be the first country to reach artificial general intelligence, when a generalisable AI matches human intelligence across all domains. But that definition can be contested in all kinds of ways. And it could be that the most consequential breakthroughs will come before then in specific and narrow economic and military domains.

Question asked by: HG
Question: Market fragmentation and legacy technologies make US companies struggle with AI readiness. Chinese companies roll out AI-native stacks faster without these encumbrances. Example: transportation (routing, safety, ERP). How can US policymakers respond?

Answered by: Eleanor Olcott
Answer: I’d actually push back on that. There’s a running joke among Chinese VCs that “China is where enterprise software companies go to die.”There’s a reason there’s no Chinese Salesforce or Adobe — companies here just don’t like paying for software. That’s why so many AI firms are focused on consumer apps or looking overseas for paying customers.So while China’s tech stack might be newer, the local market isn’t exactly fertile ground for enterprise adoption. That’s a real drag on “AI readiness.”

Question asked by: Gift Kugara Mawire
Question: Is the AI race about innovation or implementation? Could China’s centralised data and industrial policy give it an edge in deploying AI across its economy faster than the U.S.’s decentralised, market-driven model?

Answered by: Eleanor Olcott
Answer: Depends on the sector. In manufacturing, China absolutely has the edge. The government is directing a huge campaign to modernise factories — automating equipment to boost efficiency and offset demographic decline.But in the US, AI adoption is more diverse and market-driven. Companies are using AI to cut costs, improve productivity — often meaning layoffs. In China, the enterprise software market is smaller and stingier, so you just don’t get the same breadth of innovation for enterprise solutions specifically.

Answered by: John Thornhill
Answer: It’s about both. The US still leads in frontier AI research and innovation. But China is extremely well placed to apply AI across all kinds of industries. Last year, US-based organisations produced 40 of the world’s strongest AI models, compared with 15 from China. But China has learned to do more with less. Its big bet on open-weight models, which can be more easily and cheaply adapted by developers, also suggests it may be faster in the implementation game.

Question asked by: Omar from Saudi Arabia
Question: What role do the Arab Gulf countries play in U.S.-China AI competition?

Answered by: Eleanor Olcott
Answer: The Arab Gulf has become ground zero for China’s AI exports.You’ve got SenseTime and Huawei selling surveillance systems, Chinese cloud providers building data centres, and ByteDance setting up shop to develop enterprise AI tools for local clients.China’s AI companies need growth abroad — and the Middle East is their first stop.

Question asked by: Ursula W
Question: Will quantum computing not take a substantial part of the current AI market? If see the announcement of HSBC/IBM regarding bond market trading and Google with drug discovery

Answered by: John Thornhill
Answer: The interplay between AI and quantum is going to be a fascinating area to watch. AI is helping to accelerate the development of quantum technologies and vice versa.Although the scale of investment into quantum in no way matches that into AI, it has been picking up fast over the past year. Several companies are now predicting they are on track to deploy a fault-tolerant quantum computer by 2030, which would be a big deal, if true. But AI and quantum solve different kinds of problems, so they are more likely to be complementary than competitive.

Question asked by: Noostrata von Llull
Question: Will Chinese companies stop releasing open source / open weights models and pivot to a proprietary logic?

Answered by: Eleanor Olcott
Answer: Highly unlikely. Open source is now national policy in China. After seeing how successful these models have been, Beijing made it part of its AI strategy at home and abroad. Officials literally talk about spreading “Chinese open source AI” as a form of soft power.Even companies that started with closed systems — like Baidu — had to pivot to open source to stay in line with this direction. So a full pivot away from open weights would be politically and strategically very difficult.

Question asked by: Alex N.
Question: What’s the pros and cons of AI providers in China ? How do they differentiate from US and French AI providers ? (Mistral AI, Chatgpt, gemini)

Answered by: Eleanor Olcott
Answer: From a developer’s point of view, it really comes down to open vs. closed source.Chinese open-source models are cheap, flexible, and easy to tinker with. Developers can download them for free and only pay to run inference — basically, to power the AI applications. That cost has been driven down aggressively by the likes of DeepSeek and Alibaba. You can also tweak the model weights to suit your own needs, which gives developers a lot of creative freedom.The cons? There’s still a 6–12 month lag between the best Chinese open-source models and the top US closed-source ones like GPT-5. And that gap really matters for more complex, research-heavy tasks — anything that needs real reasoning and discernment about what’s relevant or accurate.

Question asked by: Anonymous
Question: That’s all we have time for. Thanks FT readers for your questions and John and Eleanor for your thoughtful responses. We’ll be back with another Ask an Expert next week on Thursday at 1pm GMT. Visit www.ft.com/ask-an-expert.

Answered by: John Thornhill
Answer: Thanks for all the testing questions! Great to see all the interest in this subject. John

Answered by: Eleanor Olcott
Answer: Thank you all for your questions. You have given me some ideas for future story ideas… stay tuned!