At Beijing Capital International Airport, it is now required after landing to send carry-on bags through a scanner before leaving the arrivals building. As I approached the scanner belt with my bags, though, I saw a sign that I had never encountered before: Any passenger carrying a drone has to remove and declare it. A Chinese security officer pointed directly to the sign to confirm that I had no drones in my bag. I shook my head no. I don’t usually travel with a drone.
在北京首都国际机场，现在规定，乘客落地后，必须先将随身行李过安检机扫描，然后才能离开到达大厅。然而，当我提着行李走向安检传送带时，看到了一个我从未见过的标志：任何携带无人机的乘客都必须将其取出并申报。一名中国安检人员直接指着那个标志，确认我的行李中是否携带无人机。我摇头表示没有。我通常不带无人机旅行。
That sign was not the only new thing I saw on this trip. It was one of many. Having traveled to China regularly since the early 1990s, I found this journey totally different for a very specific reason: It was my first visit to China in the full-blown age of artificial intelligence.
那个标志并不是我此次行程中看到的唯一新鲜事物，只是众多新事物之一。自20世纪90年代初以来，我经常前往中国旅行，但我发现这次旅行完全不同，而且有一个很具体的原因：这是我在人工智能全面爆发的时代首次访问中国。
As I end nine days in the country, having visited Beijing and Xiong’an as well as China’s tech hubs in Hangzhou, Ningbo and Shenzhen, it is very clear to me that the A.I. era is already reshaping the geopolitical and geoeconomic relationships between the United States and China in ways that are actually pulling our two countries closer together geopolitically, while exacerbating our trade differences.
结束了为期九天的中国之行——期间我访问了北京、雄安，以及杭州、宁波和深圳等中国科技中心——我非常清楚地看到，人工智能时代已经在重塑美中两国的地缘政治和地缘经济关系，一方面，它实际上正在地缘政治层面拉近两国的距离；另一方面却加剧了双方的贸易分歧。
How so? Let’s go back to that drone sign at the Beijing airport. What is it telling us? It is telling us that while the United States and China are still economic and strategic rivals jostling for advantage, they both realize that there is now an intruder into their two-superpower marriage: the superempowered bad actor.
为什么这么说？让我们回到北京机场那个关于无人机的标志。它告诉了我们什么？它告诉我们，尽管美中两国仍然是为了争夺优势而相互角力的经济和战略竞争对手，但两国都意识到，在这两个超级大国的紧密结合之中，现在出现了一个闯入者——那个拥有超强赋能的恶意行为者。
I’m not talking about Chinese democracy advocates. I’m talking about agents of chaos — fraudsters, drug dealers, human traffickers, the kind of hackers who just stole $6.3 billion in cryptocurrency from Binance. I’m talking about bad individuals and small teams that can now arm themselves with off-the-shelf drones or A.I.-powered cyberweapons to wield the destructive force of a superpower all by themselves. And I am talking, even more important, about both American and Chinese A.I. models that are demonstrating greater and greater capacities to go rogue and become agents of mass disruption as they become increasingly self-directing, self-improving and self-preserving.
我指的不是中国的民主倡导者。我说的是制造混乱的势力——诈骗犯、毒贩、人口贩子，以及那种刚刚从币安盗走63亿美元加密货币的黑客。我指的是，现在那些恶意的个人和小团队可以用现成的无人机或人工智能驱动的网络武器武装自己，凭借一己之力就能施展出堪比超级大国的破坏力。更重要的是，我指的还有美国和中国的人工智能模型，随着它们日益具备自主决策、自我优化和自我保护的能力，正展示出越来越大的失控风险，可能成为导致大规模破坏的推手。
Both countries are just beginning to wrestle with the mind shift this requires in how they relate to each other — even as they continue to reflexively take the opposite sides on wars from Gaza to Ukraine to Iran. A Chinese blogger friend put it to me candidly over dinner in Beijing last week: “We still have real differences over issues like Taiwan. But sometimes the U.S. and China don’t know what they are fighting about anymore,” he said. “It’s just a habit.”
两国都才刚刚开始努力应对这种局面在双边交往中所要求的思维转变——尽管在从加沙到乌克兰再到伊朗的战争中，两国仍习惯性地站在对立面。上周在北京吃晚饭时，一位中国博主朋友坦诚地对我说：“在台湾等问题上，我们仍然有实质性的分歧。但有时美国和中国已经不知道他们到底在争什么了，”他说。“这只是一种习惯罢了。”
Well, it’s an obsolete habit that both nations must unlearn, and quickly, in the age of artificial intelligence, when global stability will increasingly require mutually assured collaboration between the world’s two A.I. superpowers.
好吧，这是一个两国都必须摒弃的过时习惯，而且要快；在人工智能时代，全球稳定将越来越需要世界上这两个人工智能超级大国之间建立起确保相互协作的关系。
To be sure, our definition of disorder is often not the same as China’s. Any regulatory system embraced by the United States needs to protect democratic values, free expression and peaceful protest. China is unlikely to have any of those in its regulatory scheme. But a lot of overlap remains between our self-interests, which lies in preventing artificial intelligence from causing violent harm — to infrastructure, to individuals, to countries and even to our species. That is the kind of disorder that Washington and Beijing should be able to agree needs controlling. We need these basic rules more than ever because the next steps for A.I. involve, literally, walking — as in robots.
诚然，我们对“混乱”的定义往往与中国不同。美国采纳的任何监管体系都需要保护民主价值观、言论自由和和平抗议。中国的监管框架中很可能不包含这些内容。但我们的自身利益之间仍然存在很多重叠之处——防止人工智能对基础设施、个人、国家乃至整个人类造成暴力伤害。这正是华盛顿和北京应该能够达成共识、需要加以控制的那种混乱。我们比以往任何时候都更需要这些基本规则，因为人工智能的下一步发展，从字面意义上来说，就是“行走”——就像机器人那样。
I saw a demonstration at an A.I. robotics exhibition in Beijing where a Chinese humanoid robot replaced its own battery pack, lodged in its back. It just took a couple of minutes. As I stared at it doing this maneuver, I wondered: What happens when it also has autonomous intelligence?
在北京的一场人工智能机器人展览上，我看到了一项演示：一台中国制造的人形机器人自行更换了背部安装的电池组，整个过程仅用了几分钟。当我目不转睛地看着它完成这一操作时，我不禁思考：如果它还具备自主智能，会发生什么呢？
While the need for more collaboration against those trying to get the worst of their A.I. technologies may push Beijing and Washington closer together over time geopolitically, in geoeconomics the opposite is likely to happen.
尽管为了共同应对那些企图将人工智能技术用于恶用目的的势力，北京和华盛顿在地缘政治层面可能会随着时间推移而走得更近，但在地缘经济层面，情况很可能恰恰相反。
That is because A.I. is simultaneously superempowering something else: China’s manufacturing juggernaut. As of 2024, China already made almost a third of all manufactured goods sold in the world. What I saw on this trip, visiting some of China’s most advanced car companies and telecom manufacturers, is that A.I. is intensifying their efficiency and competitiveness even more.
这是因为人工智能同时也在极大地增强另一股力量：中国庞大的制造业机器。截至2024年，中国生产的制造业产品已占全球销售总量的近三分之一。在这次旅行中，我参观了一些中国最先进的汽车公司和电信制造商，我看到人工智能正在进一步提高它们的效率和竞争力。
You can see it in trade statistics. China’s trade surplus in 2025 — exports over imports — was a staggering $1.19 trillion. That was a 20 percent increase from the year before. This year it is expected to grow to $1.3 trillion — a trade imbalance out of line with global economic output not seen by any country since the end of World War II, when the United States was the only major industrial economy standing!
从贸易统计数据中可以清楚地看到这一点。2025年中国的贸易顺差——出口额减去进口额——达到了惊人的1.19万亿美元。这比前一年增长了20%。今年预计将增长到1.3万亿美元——这种与全球经济产出严重失衡的贸易状况，自第二次世界大战结束以来，任何国家都未曾出现过，而二战后美国还是唯一幸存的主要工业经济体！
Trust me, you have not seen anything yet.
相信我，好戏还在后头。
Because while American companies are burning through billions investing in frontier A.I. models that are so advanced they’re humbling the best mathematicians, Chinese companies are focused on deploying cheap, “good enough,” open-weight A.I. models that could make its large lead in advanced manufacturing simply unbeatable. To put it differently, America is spending vast sums on frontier A.I. models that are already so powerful the latest versions are too dangerous to be released. China, by contrast, starts the A.I. era with such a dominant base of advance manufacturing facilities, it can reap immediate and broad gains from its smaller “good enough" A.I. models.
因为，当美国公司正耗资数十亿美元投资于前沿人工智能模型——它们如此先进，令最顶尖的数学家都自叹不如，中国公司正专注于部署廉价、“够用”的开源权重人工智能模型，这可能使其在先进制造业中的巨大领先优势变得根本无法逾越。换种说法，美国正斥巨资开发前沿人工智能模型，这些模型已经如此强大，以至于最新版本因为太危险而无法发布。相比之下，中国在迈入人工智能时代的时候，已拥有如此强大的先进制造业基础，因此即使采用规模较小但“够用”的人工智能模型，也能立即获得广泛的收益。
When you combine this with China’s cowboy capitalism — scores of companies fiercely competing in every sector of the domestic economy — you get a pace of progress that is downright scary in some industries. At the Beijing Robot Industrial Park, where I saw that robot swapping out its own battery pack, another exhibit impressed me even more.
当你将这一点与中国的牛仔式资本主义（国内经济各个领域都有大量企业展开激烈竞争）结合起来，你会发现，某些行业的进步速度简直令人害怕。在北京机器人产业园，也就是我看到那个机器人自行更换电池组的地方，另一个展品给我留下了更深刻的印象。
It was two humanoid robots standing side by side. They were built by different Chinese companies to compete in the Beijing E-Town Humanoid Robot Half-Marathon race — 21 kilometers, or 13.1 miles. The first robot runner, which had a gold medal on a ribbon draped around its neck, won the 2025 race in two hours, 40 minutes and 42 seconds. The second robot runner, also sporting a gold ribbon, won the same race in April in 50 minutes and 26 seconds, beating what was then the human world record, 57 minutes and 20 seconds, held by Jacob Kiplimo.
那是两个并排站立的人形机器人。它们由不同的中国公司制造，准备参加北京亦庄举办的人形机器人半程马拉松赛——赛程全长21公里。第一个机器人跑者脖子上挂着一条系有金牌的绶带，它在2025年的比赛中曾以2小时40分42秒的成绩获胜。第二个机器人跑者也佩戴着金牌，它在今年4月举行的同一赛事中以50分26秒的成绩获胜，打破了当时由雅各布·基普利莫保持的57分20秒的人类世界纪录。
In a single year, Chinese advances in embodied artificial intelligence produced a robot marathon runner faster than the fastest human ever recorded. My Chinese guide pointed at the two robots and said: “Last year, this year. Next year — we’ll see.”
仅仅在一年之内，中国在具身人工智能方面的进步就创造出一个比有史以来跑得最快的人类还要快的机器人马拉松选手。我的中国向导指着这两个机器人说：“去年，今年。明年——我们走着瞧。”
We certainly will. Other than that, folks … nothing to worry about here.
我们当然会走着瞧。除此之外，伙计们……这里似乎没什么好担心的了。
What follows is what I gleaned from dozens of conversations and company visits in China on how the age of A.I. is reshaping China and China-U.S. relations.
以下是我通过在中国的数十次交谈和公司考察所了解到的，关于人工智能时代如何重塑中国以及中美关系的情况。
Let me start with some good news.
让我先说点好消息。
Six months ago, if you asked pretty much any A.I. expert if China and America would ever get together and agree on a set of guidelines to control their now increasingly dangerous A.I. models, the answer you would get is, “No way, no how” — with a special emphasis that China would never agree to such a thing. Think again.
六个月前，如果你问起几乎任何一位人工智能专家：中美两国是否会坐下来就一套规范达成共识，管控如今日益危险的人工智能模型？你得到的答案会是：“绝不可能”——而且对方会特别强调，中国永远不会同意这样的事情。现在，重新想想吧。
Right now, I would say that Beijing may actually be more eager to discuss A.I. guardrails than President Trump, who told the United Nations in September that the United States “rejects any attempt to construct a globalist scheme to control” A.I. — a theme he repeated after his summit with President Xi Jinping days later. Trump to the world: America’s frontier A.I. models are in the lead and I won’t collar them. Full stop. Get lost.
眼下，我想说北京实际上可能比特朗普总统更渴望讨论人工智能的监管框架，特朗普曾在9月份对联合国表示，美国“拒绝任何构建控制人工智能的全球主义计划企图”——在几天后与习近平主席举行峰会后，他又重申了这一主题。特朗普向世界传达的意思是：美国的前沿人工智能模型处于领先地位，我不会给它们套上枷锁。说完了。滚吧。
At the opening session of the 2026 Global Governance Forum in Xiong’an, I had a discussion with Wang Yi, the Chinese foreign minister and Politburo member, and John Thornton, co-chair of the Asia Society board of trustees. I was struck by Wang’s clear message in support of multilateral A.I. safeguards.
在雄安举行的2026年全球治理论坛开幕式上，我与中国外交部长兼政治局委员王毅以及亚洲协会董事会共同主席约翰·桑顿进行了讨论。王毅明确支持多边人工智能安全保障机制，这让我印象深刻。
As Wang put it to us: The “leading nations in A.I., China and the United States, have both the capability and the responsibility to jointly develop and manage A.I. effectively, ensuring its development remains under human control and benefits the people. We must not only mitigate risks but also cooperate to jointly discuss globally applicable rules and standards” to make sure A.I.’s development stays under human control and for the benefit of people.
正如王毅对我们所说的：“中美作为人工智能第一梯队国家，有能力也有责任共同发展好、管控好人工智能，确保其发展始终受控于人、造福于民。我们不仅要防范风险，也要开展合作，共同商讨全球适用的规则标准”，以确保人工智能发展始终受控于人、造福于民。
Of course, China, like the United States, is building A.I. tools to enhance weapons to fight abroad and improve every tool in its domestic surveillance kit to control dissent at home. But Beijing also knows — as that no-drones-in-your-luggage sign at the airport indicated — that A.I. is a two-edged sword and can be used by a lot of bad actors. That is why when you talk to U.S. administration officials about what the Chinese are doing with us behind the scenes, you’ll get a complex answer.
当然，与美国一样，中国也正在构建人工智能工具，以增强用于海外作战的武器，并升级其国内监控工具箱中的每一项工具，以控制国内的异见声音。但北京也知道——正如机场那个“行李中禁止携带无人机”的标志所显示的那样——人工智能是一把双刃剑，可能会被许多恶意行为者利用。正因如此，当你向美国政府官员询问中方在幕后针对美国采取了哪些行动时，得到的会是一个复杂的回答。
These officials have no illusions that Beijing would like to see more guardrails on America’s frontier models, because the United States is in the lead. But they will also tell you that China’s law enforcement agencies lately have become much more cooperative with their U.S. counterparts on combating cybercrime, fraud and money laundering — all of which A.I. is rapidly enhancing. The Chinese know that the open-weight A.I. models being distributed by Chinese companies are easier for software developers, including those that are bad actors, to adapt to specific, targeted applications than Silicon Valley’s frontier models.
这些官员不抱任何幻想，鉴于美国在该领域处于领先地位，北京自然希望对美国的前沿模型设置更多限制。但他们也会告诉你，最近，中国的执法机构在打击网络犯罪、欺诈和洗钱方面，与美国同行的合作变得更加密切——所有这些犯罪活动都在被人工智能迅速增强。中国人知道，中国公司发布的开源权重人工智能模型比硅谷的前沿模型更容易让软件开发者（包括那些恶意行为者）改编为特定的、针对性强的应用。
But don’t take my word for it. In September, the Chinese Communist Party’s top intelligence official, Chen Yixin, head of the Ministry of State Security, wrote in the journal China Cyberspace that emerging A.I. technologies were “directly threatening” China’s political, institutional and ideological security. Chen added: “Various hostile forces and individuals with ulterior motives abuse generative artificial intelligence technologies” to create rumors, spread misinformation and “incite confrontational sentiments at low cost and in large quantities.”
但别只听我的一面之词。今年9月，中共最高情报官员、国家安全部部长陈一新在《中国网信》杂志上撰文指出，新兴人工智能技术正在“直接威胁”中国的政治、制度和意识形态安全。陈一新还说：“各种敌对势力和别有用心之人滥用……生成式人工智能技术”，制造谣言、传播虚假信息，并且“低成本、批量化……煽动对立情绪”。
Read those words: at low cost and in large quantities.
请注意这些字眼：低成本、批量化。
Chen’s government has the world’s largest state-run facial recognition surveillance system, which can pluck a single person out of a huge crowd — what I call “the big acting small.” What Chen was saying is that he also knows that A.I. is a dual-use technology that is enabling the small to now act big. That is the essence of the huge change in the geophysics of the world — the spread of cheap, precision technologies, like drones and cyber-hacking tools, that are now supercharged by A.I. That new superempowered actor can be another nation or a single foreigner or one of your own citizens. Either way, it’s a big new problem in need of a big new answer.
陈一新的政府拥有世界上最大规模的由国家运行的面部识别监控系统，它可以庞大的人群中精准锁定单个目标——我称之为“大角色以小规模行动”。 陈一新所表达的意思是，他也知道人工智能是一种双重用途技术，它正在使得小角色如今也能以大规模行动。这正是世界地缘政治格局发生巨大变化的本质——廉价、精密的技术（如无人机和网络黑客工具）正日益扩散，现在人工智能又为这些技术注入了强劲动力。这种获得超强赋能的新行为体可能是另一个国家、也可能是一个外国人，或者是你本国的一个公民。不管怎样，这都是一个亟待全新重大对策的全新重大问题。
Which is why I was not surprised that while I was in China last week the government issued a draft of A.I. rules that would give the governing Communist Party new powers to track or restrict exports of some advanced Chinese systems. Why now? Maybe because, as The Times reported on Oct. 14, A.I. agents attributed to tests being run by the Chinese tech giant Tencent recently went rogue and tried to “bypass anti-robot protections on websites,” among other shenanigans. Yup, the Chinese have the same problems with their biggest A.I. models as we do, even if they are six months to a year behind in their capabilities.
正因如此，当我上周在中国时，看到政府发布了一份人工智能规则草案，赋予执政的共产党新的权力，以追踪或限制某些中国先进系统的出口，对此我毫不感到惊讶。为什么是现在？也许是因为，正如《纽约时报》10月14日报道的那样，据说由中国科技巨头腾讯运行的测试中的人工智能智能体最近出现了失控，除了搞出一些恶作剧之外，还试图“绕过网站上的反机器人保护机制”。是的，中国人在他们最大的人工智能模型上也遇到了和我们一样的问题，即便他们的能力比我们落后六个月到一年。
So, I am going to go out on a limb here: In the coming year, China and America will be seriously exploring collaborative controls on A.I. models. Because, as America’s frontier models reach the point of training themselves 100 percent through recursive self-improvement (without human direction) and as China’s open-weight models that learn off ours through distillation get strengthened in their wake, I predict that 2027 will be the year when both countries fully grasp what an amazing opportunity A.I. is for humanity — and what an unprecedented threat it is to humanity.
所以，我要在这里大胆预测一下：在未来一年里，中美两国将认真探索对人工智能模型实施协同控制。因为，随着美国的前沿模型达到通过递归自我改进（无需人类指导）实现100%自我训练的程度，以及中国通过知识蒸馏向我们学习的开源权重模型随之变得越来越强大，我预测，2027年将成为两国完全领悟到这一点的一年：人工智能对人类来说是多么惊人的机遇，同时又是多么前所未有的威胁。
This collaboration will be tricky for America. Because, unlike the U.S. government, the Chinese government of President Xi is worried not only about A.I.’s empowering criminals at home, but also about political opponents. Managing that complexity is precisely what makes U.S.-China relations in the age of A.I. different.
这种合作对美国来说将非常棘手。因为与美国政府不同，习近平主席领导的中国政府不仅担心人工智能会在国内为犯罪分子赋能，还担心它为政治对手赋能。如何应对这种复杂局面，恰恰是人工智能时代中美关系与以往不同的关键所在。
Given all of these competing pulls, it may not be possible for the United States and China to come up with a shared system of A.I. guardrails as much as they might need to — at least not without a perfect A.I. storm. That is, a large-scale A.I.-driven event that is big enough to frighten both countries into collaborating on effective controls — but not so big as to end the world.
考虑到所有这些相互拉扯的力量，中美两国可能无法像两国所需的那样，彻底制定出一套共同的人工智能监管框架——至少在没有一场“完美人工智能风暴”的情况下无法做到。也就是说，一场由人工智能驱动的大规模事件，其规模大到足以吓得两国合作建立有效的控制机制，但又没有大到能导致世界末日。
I repeat: 2027 could be that year. Let’s hope the storm is perfect.
我再说一遍：2027年可能就是那一年。但愿这场风暴是完美的。
Now let’s look at how A.I. is reshaping economic relations between China and the United States and the rest of the world. Every time I visit China, I come away more struck than from the trip before by the stark contrast in its economy. If China were a person, it would have the upper body of an Olympic weight lifter — rippling with muscles that represent its export-driven advanced manufacturing economy — and the lower body of an Olympic skier who just crashed into a mogul, broke his leg and can now only hobble on a crutch.
现在，让我们看看人工智能如何重塑中国与美国以及世界其他地区的经济关系。我每次造访中国，都会比上一次更加深刻地感受到中国经济内部的巨大反差。如果把中国比作一个人，那么它的上半身就像一名奥运举重选手，肌肉发达，代表着出口驱动的先进制造业经济；而下半身则像一名在滑雪比赛中撞上雪墩、摔断了腿的奥运滑雪运动员，现在只能拄着拐杖蹒跚前行。
That lower body represents a domestic Chinese economy that is still so weakened by the explosion of its large, quarter-century-long housing bubble — which ravaged its middle class and left hundreds of millions of people with mortgages that are underwater or with wiped-out savings — that many Chinese are holding tight to every yuan they can. The frenzied overbuilding of high-rise apartments with cheap money in the early 2000s got so crazy that, as the Times Beijing bureau chief, Keith Bradsher, wrote in February 2025, China now has “abundant housing for the country’s 1.4 billion people and enough empty apartments for another 300 million people.” Put another way, Bradsher said this past spring, China has enough empty apartments to house almost every person in America, which has a population of about 350 million.
这下半身代表着中国国内经济，它仍因其长达四分之一个世纪的巨大房地产泡沫的破裂而元气大伤——这摧残了中产阶级，让数亿人背负着资不抵债的房贷或积蓄化为乌有——以至于许多中国人如今都尽可能攥紧手里的每一分钱。在21世纪初，廉价资金带来的高层公寓过度建设热潮变得如此疯狂，以至于正如《纽约时报》北京分社社长柏凯斯(Keith Bradsher)在2025年2月写道的那样，中国现在“除了为14亿人口提供了充足的住房外，也建造了足够另外3亿人居住的空置公寓”。柏凯斯在今年春天曾表示，换句话说，中国的空置公寓几乎足以容纳美国的全部人口（美国人口约为3.5亿）。
In short, China is an export giant and domestic consumption invalid. Unfortunately, the government, instead of injecting money into the domestic economy through improved pension programs, is doubling down on subsidies to build up its global lead in advanced manufacturing — and telling Chinese consumers to suck it up.
简而言之，中国是一个出口巨人，却也是一个国内消费羸弱的病人。遗憾的是，中国政府没有通过改善养老金体系等方式向国内经济注入资金，反而加倍补贴先进制造业，以巩固其全球领先地位，同时要求中国消费者默默忍受。
Therein lies a huge problem for the global economy at the dawn of the A.I. revolution. Because what I saw on this trip is that the Olympic weight lifter part of the Chinese economy is now guzzling down protein shakes labeled A.I. and is becoming an even more dominating competitor in every advanced manufacturing realm — from socks to robots.
这正是人工智能革命初期全球经济面临的一个巨大问题。因为我在这次旅行中看到的是，中国经济中那个“奥运举重选手”部分现在正在狂饮贴着人工智能标签的蛋白质饮品，在从袜子到机器人等各个先进制造领域成为更具统治力的竞争者。
Sure, American, European, Japanese and Korean manufacturers are also benefitting from A.I. enhancements. The problem is that China has so much more installed advanced manufacturing and therefore is going to reap outsized gains. Come along with me on my whirlwind tour of some major Chinese advanced manufacturers.
当然，美国、欧洲、日本和韩国的制造商也受益于人工智能的增强。问题在于，中国现有的先进制造业规模远远超过其他国家，因此有望从人工智能带来的变革中获得远超其他国家的收益。请随我一起进行一次旋风式的访问，参观几家中国主要的先进制造企业吧。
On my visit to Alibaba headquarters in Hangzhou, Xiong Xiong, chief of staff to Alibaba’s co-founder and chairman, Joe Tsai, pointed out to me that in Formula 1 racing, you rarely overtake your competition on the straightaways. “You win in the turns,” he said, “because in the turns it is not how fast you go, it is how daring you are — how fast you brake and how you respond to my braking — and do you have the stability to control your car at high speed.”
在杭州访问阿里巴巴总部时，集团联合创始人、主席蔡崇信的办公室主任熊雄（音）向我指出，在一级方程式比赛中，车手很少能在直道上超越对手。“胜负往往在弯道决定，”他说，“因为在弯道上，关键不在于你开得有多快，而是你有多大胆——你多晚刹车、怎么应对我的刹车——以及你有没有足够的稳定性，在高速下控制住你的车。”
Well, one day we may look back at the years 2019 to 2023 as one of those turns and China came out of it now leading the pack on the straightaway. Two things happened in that period: Covid-19 and A.I. both exploded around the world. Because of Covid, virtually all American students and many American business executives left China. I was also not able to make my own annual trips.
或许有一天，我们回头看2019年至2023年，会发现那几年正是这样一个弯道，而中国如今已经驶出弯道，在直道上领跑。那段时期发生了两件大事：新冠疫情和人工智能浪潮相继席卷全球。由于疫情，几乎所有在华美国留学生以及许多美国企业高管都离开了中国。我也无法像往年那样每年访华。
Here is one example of what happened while we were sleeping.
下面就是一个例子，说明在我们沉睡之际发生了什么。
When I visited China, before Covid, Xiaomi was a smartphone company. When I came back in 2023, it was also an electric car company. It put four wheels on its phone. But it didn’t stop there. When I visited its almost entirely roboticized car factory on the outskirts of Beijing last Monday, I found Xiaomi was a superconnected, A.I.-enabled E.V. company. My hosts noted how Xiaomi’s self-developed “multi-material performance simulation system” uses A.I. to quickly figure out what the company believes is the optimal alloy formula — from 10.16 million possibilities — to produce what it called “a perfect combination” of strength, resilience, stability and low carbon footprint for its newest vehicle bodies.
疫情前我访问中国时，小米是一家智能手机公司。当我在2023年再来时，它也是一家电动汽车公司了。它给手机装上了四个轮子。但它并没有止步于此。上周一，当我参观其位于北京郊区、几乎完全由机器人操作的汽车厂时，我发现小米是一家超级互联、人工智能赋能的电动车企业。接待我的人提到，小米自主研发的“多元材料性能AI仿真系统”如何利用人工智能，从1016万种可能中迅速找出公司认为的最优合金配比，为其最新车身生产出强度、韧性、稳定性和低碳足迹的所谓“完美组合”。
Trump hates electric cars. What he has never understood is that China’s hypercompetitive E.V. industry birthed a powerful, deeply integrated supply chain of advanced battery technology, robotics, material science and, now, A.I. Today, any Chinese inventor or entrepreneur looking to manufacture absolutely anything can seamlessly leverage this hyper-efficient, foundational ecosystem to innovate better, faster and cheaper than most any global competitor. Global manufacturers used to say you had to be in China “to sell to China.” Now they say, you have to be in China and have access to its supply chains “to sell to the world.”
特朗普讨厌电动汽车。他从未明白的是，中国竞争极其激烈的电动汽车产业催生出一条强大、高度整合的供应链，涵盖先进电池技术、机器人技术、材料科学，以及现在的人工智能。如今，任何想制造任何东西的中国发明家或企业家都能无缝利用这个超级高效的基础生态系统，比大多数全球竞争对手更好、更快、更便宜地创新。全球制造商过去常说，要想把产品卖到中国，就必须在中国设立业务。如今，他们则说，要想把产品卖向全世界，就必须在中国布局，并接入中国的供应链。
You want to make shoes with electric rollers and flashing lights that sing the Chinese national anthem on the left shoe and the American national anthem on the right shoe? Someone here will have the parts and tools for you to do it in 24 hours.
你想制造装有电动滚轮和闪烁彩灯的鞋子，让左脚鞋唱中国国歌、右脚鞋唱美国国歌？这里会有人在24小时内为你提供完成这项工作所需的零件和工具。
In Ningbo, I visited the Geely electric vehicle manufacturing plant. I watched an engineer use the company’s latest charging station to power up a car from Geely’s newest A.I. luxury E.V. brand, called the Lynk & Co, in eight-plus minutes with a charge that will last some 500 miles. That is not a typo — that is almost like filling up a tank of gas. “In the future, we intend to improve it further,” the Geely demonstration engineer laconically told me. Yes, I thought, I am sure you will.
在宁波，我参观了吉利电动汽车制造厂。我看着一名工程师使用该公司最新的充电桩，用八分多钟的时间为吉利旗下最新的人工智能豪华电动汽车品牌领克的一款汽车充电，充完电足够跑约800公里。这不是笔误——这几乎就像加满一箱油一样快。“未来，我们打算进一步改进，”吉利的演示工程师轻描淡写地告诉我说。是的，我心想，我确信你们会的。
In Shenzhen, I toured Huawei Technologies’ assembly line for its newest Mate 90 smartphones. Zha Jun, who oversees Huawei’s research and development, explained that this line used to have 86 workers per shift — doing mostly manual inspections — and the rest of the work force was robots. It recently shrunk that number down to nine people by turning to A.I.-enabled image recognition systems.
在深圳，我参观了华为最新款Mate 90智能手机的装配线。华为研发团队负责人查钧解释说，这条装配线过去每班有86名工人——主要负责人工质检——其余的劳动力是机器人。最近，通过改用具备人工智能图像识别功能的系统，将每班所需的工人数量减少到了九人。
When I asked what kind of data center it used to run the A.I. systems, Zha said: “We have no need for a data center. Huawei can get all the A.I. benefits it needs for these kinds of manufacturing improvements using our own open-weight, domain-specific A.I. models with less than one billion parameters.” Those are the ones with circuitry that amounts to the size of a small suitcase. (Parameters are a measure of the connections and weights that an A.I. model uses to process information and are used as a general proxy for its complexity and capacity to learn.) These small A.I. models, Zha added, were “good enough” to make big improvements in manufacturing.
当我问起运行这些人工智能系统使用的是什么样的数据中心时，查钧回答：“我们不需要数据中心。华为利用自身不到10亿参数的开放权重、面向特定领域的AI模型，就能获得这类制造升级所需的所有人工智能收益。” 这类模型所需的运算硬件也就一个小手提箱那么大。（参数衡量的是人工智能模型用来处理信息的连接和权重，通常用来代表其复杂性和学习能力。）查钧补充说，这些小型的人工智能模型“够用了”，能够为制造业带来巨大改善。
“Good enough” A.I. was a phrase I heard at a lot of factories on this trip. It describes the improvements to businesses produced by the less-than-frontier A.I. models.
“够用的人工智能”是我在这次旅行中走访许多工厂时经常听到的一个表述。它描述了那些并非处于最前沿的人工智能模型为企业带来的改进。
“China and the U.S. are taking different directions when it comes to A.I.,” Ren Zhengfei, Huawei’s founder and chief executive, explained to me over tea at Huawei’s research campus in Shenzhen, one of many technology hubs where the company employs some 110,000 researchers, many of them engineers and scientists with advanced degrees. In the United States, Ren said, A.I. research revolves around “what does it mean to be human these days and what is the meaning of the world, and of God? In China, A.I. research is more practical” and is being deployed in big industries, like steel.
“在人工智能方面，中美两国正走向不同的方向，”华为创始人兼首席执行官任正非在深圳的华为研究园区与我喝茶时解释道。这是该公司的众多科技中心之一，在那里雇佣了大约11万名研发人员，其中许多是拥有高学历的工程师和科学家。任正非说，在美国，人工智能研究围绕着“在今天作为一个人意味着什么？世界和上帝的意义是什么？而在中国，人工智能研究更为务实”，并正被部署在诸如钢铁这样的大型工业中。
“China produces one billion tons of steel a year,” Ren said, “so if that industry can achieve a 1 percent improvement by using A.I. to increase the efficiency in how it mixes the various chemicals that go into every single steel-making furnace,” the additional profits generated for a single furnace will exceed 10 million yuan ($1.5 million) per year. In the past, China was not a world leader in medicine. With the help of A.I., China is now able to strikingly accelerate its ability “to discover and produce new drugs” by rapidly testing different molecular combinations.
“中国每年生产10亿吨钢铁，”他说。“如果该行业能通过使用人工智能来提高每座炼钢炉中各种化学材料混合的效率，哪怕只获得1%的改善，”一座高炉每年产生的额外利润也将超过1000万元人民币。在过去，中国并不是世界医学领域的领导者。如今在人工智能的帮助下，中国正通过快速测试不同的分子组合，惊人地加快其“发现和生产新药”的能力。
He concluded: “That is why we call A.I. in the U.S. ‘Socrates A.I.’ and we name our approach ‘Lu Ban A.I.’” Lu Ban was a legendary ancient Chinese master carpenter and engineer.
他总结道：“这就是为什么我们将美国的AI称为‘苏格拉底AI’，而将我们的方法命名为‘鲁班AI’。”鲁班是中国古代传说中的木匠大师和工程师。
Now add one more increasingly important factor: rare earths, which in the age of A.I. will be more important than ever — not only for manufacturing but for running all the chips that go into every data center as well.
现在再加上一个越来越重要的因素：稀土，它在人工智能时代将比以往任何时候都更加重要——不仅对于制造业，对于每个数据中心里所有的芯片也是如此。
My traveling partner on this trip has been Keith Bradsher, who will soon be coming out with a book, “The Race for Rare Earths: Why China Is Winning the Contest for the World’s Most Critical Resources.” The book traces how decade by decade, the United States and Europe lost their early leadership to China in these vital metals that go into cars, drones, wind turbines, fighter jets, smartphones, missiles, lasers, A.I. chips and practically every other high-tech invention you can think of over the past 40 years, including most consumer electronics.
此次行程的同行伙伴柏凯斯很快将出版一本名为《稀土争夺战：为什么中国正在赢得对世界最关键资源的竞争》(The Race for Rare Earths: Why China Is Winning the Contest for the World's Most Critical Resources)的书。这本书追溯了在过去几十年里，美国和欧洲如何一步步将他们在这些关键金属领域的早期领先地位输给了中国。这些金属被广泛用于汽车、无人机、风力涡轮机、战斗机、智能手机、导弹、激光器、AI芯片，以及过去40年里你能想到的几乎所有高科技发明中，包括大多数消费电子产品。
In 1964, Deng Xiaoping, later to become China’s paramount leader, visited the world’s biggest rare-earth deposit, in northern China. In 1973, he began planning for China’s rise in rare-earth technologies, and by the early 1990s he foresaw the geopolitical power of China’s tightening grip on rare earths, declaring: “The Middle East has oil. China has rare earths.”
1964年，后来成为中国最高领导人的邓小平视察了位于中国北方的全球最大稀土矿。1973年，他开始规划中国在稀土技术领域的崛起。到了20世纪90年代初，他预见到了中国收紧对稀土掌控的地缘政治力量，并宣称：“中东有石油，中国有稀土。”
The United States and Europe are only now restarting a rare-earth industry — from the mines to the manufactured magnets — but will need at least five years and probably 10 to catch up, according to Bradsher’s book, which suggests ways to speed up the recovery. Meanwhile, Beijing has wielded its almost total control over the world’s supply of crucial rare earths to force the Trump administration and European Union to back down repeatedly on tariffs and to put heavy pressure on Japan not to support Taiwan.
美国和欧洲现在才刚刚开始重启稀土产业——从矿山开采到磁体制造——但柏凯斯在书中说，至少需要五年、甚至可能10年时间才能赶上中国，该书还提出了加速复苏的建议。与此同时，北京利用其对全球关键稀土供应几近绝对的控制权迫使特朗普政府和欧盟在关税问题上屡次让步，并向日本施加巨大压力，要求其不协助台湾。
You add it all up and China’s dominance in global manufacturing is extraordinary. Jens Eskelund, the head of the European Union Chamber of Commerce in China, citing information from members, told me that 90 percent of global manufactured goods move in containers. The percentage of global container exports by China rose from 31.8 percent in 2019 to 37.5 percent in 2025, he said. “The rolling three-month average for the most recent months is now 40 percent,” he said. The reason? China’s manufacturing output rose 5.3 percent during the first eight months of this year, he said. (Meanwhile, China’s domestic retail sales increased only 0.4 percent year on year in August, he pointed out.)
综合以上所有因素，中国在全球制造业中的主导地位是非同凡响的。中国欧盟商会主席彦辞(Jens Eskelund)引用成员提供的信息告诉我，全球90%的工业制成品是通过集装箱运输的。他说，中国在全球集装箱出口中所占的比例从2019年的31.8%上升到了2025年的37.5%。“最近几个月的每三个月滚动平均值现在已经达到了40%，”他说。原因何在？他表示，今年前八个月中国的制造业产量增长了5.3%。（同时他指出，8月份中国国内零售销售额同比仅增长了0.4%。）
In 2019, China exported about 2.5 containers to Europe for every container shipped in the opposite direction, Eskelund told me. During the first eight months of 2026, the ratio widened to six containers moving from China to Europe for every one container moving from Europe to China.
彦辞告诉我，在2019年，欧洲每向中国运送一个集装箱，中国就向欧洲出口约2.5个集装箱。在2026年的前八个月，这一比例扩大到了6:1，即中国向欧洲运送六个集装箱，欧洲向中国运送一个集装箱。
“I tell the Chinese: ‘Guys, trade becomes inherently volatile if you cannot point to value creation for both sides.’” Eskelund added. “China actually understands the world’s concerns. For China, though, the question is: Why should we accommodate these concerns when the world has not put up any serious resistance?”
“我告诉中国人：‘伙计们，如果你们不能证明能为双方创造价值，贸易就必然会变得动荡。’”彦辞补充道。“中国其实理解世界的担忧。但对中国来说，问题在于：既然世界并没有进行任何认真的抵抗，我们为什么要在意这些担忧呢？”
It is a good question. And one that prompts me to leave this trip with some free advice to China and America from someone who believes the world will be a more prosperous and stable place if people in both countries thrive.
这是一个好问题。这促使我在结束这次行程时，作为一个相信如果两国人民都能繁荣发展，世界将会变得更加繁荣稳定的人，向中国和美国提供一些免费的建议。
To China: You are going too far too fast too alone, and in the process you are creating too many economic relationships with too many developed and developing countries that are not win-win but win-lose. As a result of your domestic economic weakness, combined with your huge advantage in homegrown engineering talent, unparalleled manufacturing supply chains, the unrivaled work ethic of your people and your high savings rate plowed into the world’s best transportation and electricity infrastructure, China just can’t stop itself from making everything for everyone.
对中国：你走得太远、太快，也太形单影只了。在这个过程中，你与太多发达国家和发展中国家建立了太多非双赢、而是“赢-输”的经济关系。由于国内经济疲软，加上本土工程人才的巨大优势、无与伦比的制造供应链、无可匹敌的国民职业道德以及高储蓄率，被投入到全球最顶尖的交通和电力基础设施中，中国已经无法阻止自己为全世界制造一切。
I don’t know when, I don’t how, but the rest of the world will not let that happen — not when robots and A.I. are also creating steadily more worker layoffs. China is courting a real backlash. As Eskelund bluntly put it to me: “There is now a real risk we get into a trade war. We are not going to tolerate a world where China makes everything, and we just give museum tours.”
我不知道是什么时候，也不知道会以什么方式，但世界其他地区不会任由这种情况发生——尤其是在机器人和人工智能也正稳步导致更多工人失业的时候。中国正在招致真正的反弹。正如彦辞对我直言不讳地说的那样：“现在我们陷入贸易战的风险很切实。我们不会容忍一个中国包揽一切制造，而我们只能带着人去参观博物馆的世界。”
To my fellow Americans (and our E.U. allies): China is not going to slow down in manufacturing as a favor to anyone. If we want to influence China’s geopolitics and geoeconomics in the age of A.I., we have to create the economic competitiveness that would compel it to do so. That requires the United States to be as serious as the people here are.
对我的美国同胞（以及我们的欧盟盟友）：中国不会放慢制造业的步伐来作为一种恩惠。如果在这个人工智能时代，我们想要影响中国的地缘政治和地缘经济，我们必须创造出迫使它做出改变的经济竞争力。这要求美国必须像这里的人们一样认真对待。
In my view, the Chinese today are getting the best out of a bad system of government — by thinking long-term and executing on a plan to align their country with the economic trends of the 21st century.
在我看来，今天的中国人正在从一个糟糕的政府体制中榨取最好的结果——通过长远思考并执行一项让他们的国家与21世纪经济趋势接轨的计划。
But one of the weaknesses of China’s system has been on dramatic display lately: the sudden disappearance and arrests of senior officials accused of corruption — so many that it could start to slow down the whole system here, if high-level bureaucrats fear making spending decisions that could land them in jail. Or as one Chinese economist put it to me, quoting a Chinese proverb: “If the water is too clear, there will be no fish.”
但中国体制的弱点之一最近戏剧性地暴露了出来：被控腐败的高级官员突然失踪和被捕——人数之多，以至于如果高层官僚害怕做出可能让他们入狱的开支决策，这可能开始拖累这里的整个系统。或者正如一位中国经济学家引用一句中国谚语对我说的那样：“水至清则无鱼”。
And America? By doing no long-term thinking, appointing incompetent cabinet secretaries and having each new administration undo major initiatives of the previous one — like on clean energy — we’re getting the worst out of a good system.
而美国呢？因为缺乏长远思考，任命不称职的内阁部长，并且每一届新政府都会推翻前一届政府的重大举措（比如在清洁能源方面），我们正在把一个好体制折腾出最坏的结果。
Just look around. Republicans are spending more than $200 million to elect a terrible man — an impeached Texas attorney general, Ken Paxton — to be Texas’ next senator. Trump is savaging America’s scientific research budget and driving away the most talented foreigners who want to come to America to work, study and invent. Trump gutted the electric car industry in favor of gas guzzlers, while Democrats did not think it was a bad idea to leave our southern border porous, which helped to get Trump re-elected.
抬头看看四周吧。共和党人正斥资超过2亿美元帮助一个糟糕的人——被弹劾的得克萨斯州总检察长肯·帕克斯顿——竞选得克萨斯州参议员。特朗普正在大肆削减美国的科学研究预算，并赶走那些想来美国工作、学习和发明的最优秀的外籍人才。特朗普摧毁了电动汽车产业以迎合高油耗汽车，而民主党人则不认为让我们南部边境形同虚设是个坏主意，这反过来又有助于特朗普的再度当选。
The only way we will restore some balance to the economic relationship between China and America (and between China and Europe) is not by begging Beijing but by rebuilding our leverage — by being serious again about our governance, by making cost-competitive products and by planning beyond next week and then executing on those plans.
我们要在美中（以及中欧）经济关系中恢复一些平衡，唯一的方法不是去乞求北京，而是要重建我们的筹码——再次认真对待我们的国家治理，制造具有成本竞争力的产品，以及制定目光不仅限于下周的长期计划并切实去执行。
I know just what to call this strategy — I saw it on a red baseball cap the other day: “Make America Great Again.”
我完全知道该怎么称呼这个战略——前几天我在一顶红色棒球帽上看到了它：“让美国再次伟大。”
If only we could start electing people who knew how.
要是我们能开始把那些知道怎么做的人选上去就好了。
