英伟达H200协议是战略重置还是暂时休战？

我会称这是美国的一个试点项目。

I would call this a pilot by The United States, a pilot project.

这是我们从美方听到的积极信号。

It's a positive message that we can hear from The US sides.

这是美国人的一种姿态。

It's a gesture from the Americans.

这是为了让美国AI泡沫多维持一段时间的尝试。

It's an attempt to keep the American AI bubble afloat for a little bit longer.

他们必须在利润与安全之间取得平衡，这是个重大问题。

They have to balance profits with security, and that's a big issue.

国内研发将继续加速。

Domestic R and D will continue to accelerate.

这些芯片的供应肯定不会减缓研发速度。

It certainly won't be slowed down by the availability of these chips.

新的合作方式势在必行。

A new way of working together is inevitable.

华盛顿再次改变了方针。

Washington has just reversed the course again.

现在允许英伟达向中国出售其先进的H200人工智能芯片，但25%的巨额收入将直接流入美国。

It is now allowing NVIDIA to sell its advanced H200 AI chips to China, with a hefty 25% revenue cut flowing straight to the U.

S.

S.

政府。

Government.

大家好，欢迎来到本次对话。

Hello and welcome to the chat.

我是张途轩。

I'm Zhang Tuhyun.

本次会议与我一同参与的还有博士。

Joining me on this session, Doctor.

乔米，中国国际贸易经济合作研究院美国与大洋洲研究所副所长、北京外国语大学的安迪·莫克教授，以及昆士兰科技大学客座教授沃里克·鲍威尔博士。

Jomi, the Deputy Director of the Institute of American End Oceania Study, Chinese Academy of International Trade and Economic Cooperation, Professor Andy Mock from Beijing Foreign Studies University and doctor Warwick Powell, an adjunct professor, Queensland University of Technology.

欢迎回到聊天环节，先生们。

Welcome back to the chat, gentlemen.

那么，特朗普政府为何选择此时批准H200芯片对华出口？

So why did the Trump administration choose this moment to approve h 200 exports to China?

您感到意外吗？

Were you surprised?

我们是否可以从第三国视角开始讨论，鲍威尔教授？

Shall we start from a third country's perspective, Professor Powell?

您又让我为难了。

You've put me on the spot again.

我总是希望能先听听其他人的观点，这样我的日子会好过些。

I always wish I could follow-up with everyone else first so that my life's a bit easier.

我认为特朗普总统关于这些芯片的声明，首先可以放在此前马来西亚会谈的背景下理解，最终在釜山举行的习特峰会上达成了一系列美方所谓的'胜利'成果。显然AI芯片问题是双方为缓解双边贸易关系紧张达成的共识之一。

Look, I think that the announcement by President Trump in relation to these chips can be understood firstly in the context of discussions that were undertaken a little while back in Malaysia and then culminating in the summit between President Xi and President Trump in Busan where a range of announcements were made concerning what the Americans would have called wins And clearly this issue of AI chips was part of the mutual understandings to try to take some of the heat out of the bilateral trade relationships.

其次，我认为这清楚地表明特朗普政府希望与拜登政府的部分做法保持距离。拜登政府对华AI芯片出口实施日益严格的管控措施，被特朗普——我认为确实如此——视为一项有缺陷的政策。

The second thing is that I think it's a clear indication that the Trump administration wants to distance itself from aspects of the approach that the Biden administration took and imposing an ever stricter regime of controls on the export of AI chips to China by the Biden administration was seen by Trump, and I think actually rightly so, as a flawed policy.

首先，这一政策显然未能实现遏制中国的野心，反而加速了中国本土芯片能力的发展。

It certainly did not advance the ambitions of curtailing China, number one, and in fact catalysed an acceleration of Chinese domestic chip capability development.

第三个原因在于，我认为特朗普政府正面临一些国内政治和经济挑战。

And then the third reason is, I think that the Trump administration has some domestic political challenges and economic challenges.

尽管对华高科技出口的政治考量对特朗普政府而言是个复杂难题，但他们仍需设法为人工智能产业注入活力——该产业在2025年前一直是美国GDP增长的核心驱动力，但如今越来越多人质疑美国AI产业是否仅是一个不断膨胀的泡沫。

And whilst the politics of high-tech exports to China is a complicated problem for the Trump administration, it nonetheless needs to find ways of breathing life into the AI sector, which has been central to the American GDP story for the best part of 2025, but around which there are growing doubts that the American AI sector is really nothing more than an ever expanding bubble.

我认为这些声明至少能让这个泡沫再维持一小段时间。

And these announcements, I think, will at a very minimum keep that bubble going for just a little bit longer.

那么这算不算出人意料呢？

So is it surprising or not?

结合这些因素来看，人们不该感到意外，但这确实有助于理解这些举措的深层含义。

Well, in the context of these factors, I don't think people should be surprised, but it's a way of understanding what these moves mean.

好的安迪，你怎么看？

Right Andy, what about you?

你感到意外吗？其实在此之前，英伟达CEO黄仁勋曾与特朗普总统会面，你认为黄仁勋是如何说服特朗普及其顾问团队的？

Were you surprised and, actually before this I think, NVIDIA CEO Jensen Huang actually met with, President Trump so how do you think, Jensen persuaded President Trump and his advisors?

云，首先我认为具体细节或许有些出人意料，但这件事本身的发生完全不令人惊讶。

Yun, so I think first of all, the details perhaps might be a little surprising, but the fact this happened isn't surprising at all.

我会将之称为美国的一次试点，即在我所称的'竞争下的治理性相互依存'框架下的试验项目。

I would call this a pilot by The United States, so a pilot project in something that I call governed interdependence under rivalry.

这意味着美国确实在尝试探索两个战略竞争者能否在保留最关键控制权的前提下保持部分联系。

What this means is that it really is an attempt by The United States to see whether two strategic competitors can stay partially connected without surrendering the levers that matter most.

这里我是从美国的视角来阐述的。

And here I'm talking from The US perspective.

我认为这个试点建立在四个事实基础上。

And this pilot, I think rests on four truths.

第一个事实是接受。

So the first one is acceptance.

美国现在认识到无法彻底封锁中国或延缓其发展。

The US recognizes now it cannot seal China off or slow it down.

第二，这仅是试点，属于探索性尝试。

Second, this is a pilot, so this is experimentation.

因此华盛顿并非遵循某种特定意识形态或教条，而是在高压高风险的复杂关系中谨慎摸索前行。

So Washington is not following a particular ideology, particular doctrine, but really is feeling its way forward in a relationship that is defined by high pressure and high stakes.

第三点是控制。

Third is control.

从美国视角来看，他们希望所有获准事项仍需通过美国设置的关卡，无论是许可制度、规则制定还是监管审查。

So again from The US perspective, what they want is whatever is allowed still runs through American gates, whether that's licensing, that's rules, whether that's oversight.

最后是可逆性。

And then finally reversibility.

我认为美国方面由于这是项实验，不希望被任何不可撤销的承诺束缚，这种策略允许灵活调整。

So I think The US again, because this is an experiment, doesn't want to lock in to anything that cannot be unwound and this approach allows adjustment.

这种战略旨在展现谦逊而非软弱，谨慎而不致僵化。

And I think this strategy is meant to demonstrate humility without weakness and caution without paralysis.

这承认了竞争与相互依存已紧密交织，不再是非此即彼的政策选项。

And it's a recognition that both rivalry and interdependence have become fused together and are not no longer policy choices.

关于黄仁勋如何说服特朗普总统。

About how did Jensen Huang persuade president Trump.

我想可能有几个原因吧。

You know, I think maybe a couple things.

对吧？

Right?

我认为他的论点首先是，从美国的角度看，如果英伟达不向中国销售产品，这将为中国竞争对手创造空间，而让中国企业使用美国技术栈更为有利。

So I think, you know, his argument, first of all, The US perspective is that if Nvidia does not sell into China, this will create the space for Chinese competitors, and it's better to have Chinese companies on the American tech stack.

我认为向美国政府收取费用——他们不称之为税，而是称为费用，无论是15%还是25%——这个提议对特朗普总统非常有吸引力。

I think the appeal to let the US government collect, and I believe they're not calling it a tax, they're calling it a fee, whether it's 15% or 25% is very very appealing I think to president Trump.

这又是个收入来源，而且他对关税为美国财政部带来数百亿美元收入一直引以为豪。

Again, it's another source of revenue and he's been very proud of the tens of billions of dollars that tariffs have been bringing into the US treasury.

所以我想这两个可能是主要原因，另外还有个更微妙的因素：通过声明'我们没给你们最好的，只给次好的'，他也为特朗普总统提供了对抗那些反对英伟达对华销售的鹰派的弹药。

So I think those are two probably the main reasons, and I think also, you know, is more of a subtle aspect of this, but he'd also has given president Trump the ammunition against the China hawks that don't want any NVIDIA sales to China by saying, look, we're not giving you the best.

我们提供的是次优选择。

We're giving you the second best.

我认为这再次为特朗普提供了一定的政治掩护。

And again, this provides, I think, some political cover for Trump.

但我认为，英伟达和黄仁勋采取的这种策略非常精妙。

But again, it's a very, I think, a very sophisticated approach that NVIDIA and and Jensen Huang is taking here.

你是说他们称之为费用而非税收。

You're saying that they're calling it fee, not tax.

但本质上就是税收。

But essentially, it's tax.

对吧？

Right?

有什么区别呢？

What's the difference?

确实。

Sure.

我的意思是，你想怎么称呼都行。

I mean, you know, you can call it whatever you want.

对吧？

Right?

但美国政府正从中获利。

But the US government is making money off of this.

是的。

Yeah.

这才是关键。

That's the key.

没错。

Yes.

周博士，这是否意味着之前计划将降级版H20芯片推向中国市场的方案实际上已经失败了？

Doctor Zhou, does this indicate the earlier plan to push the downgraded h 20 chip into the Chinese market has effectively failed?

嗯，我认为这更像是一种平衡。

Well, I would say that it's kind of a balance.

我不认为这是失败。

I don't think that is a failed.

目标可能不像最初禁止向中国出口H20芯片时那么明确了。

The target may be not so clear as the beginning when they are trying to forbid the exports of this h 20 to China, the chips.

我认为有些人可能会觉得这是美国用来遏制中国稳定发展、跟进并赶超美国竞争对手的一种手段。

I I think that's some of these people are may think that is a kind of measurement for US to stop China's stability, to follow-up, to catch up with US competitors.

但其他人会认为这也有积极的一面，尤其对中国半导体行业的供应商和设计师而言，他们确实渴望获得更大的市场份额。

But others will think that it has good news, especially from China side, from the semiconductor, the suppliers and designers, they really want to have a bigger market share.

这意味着当你试图更多地审视市场机遇时，我认为这并非总是唯一的选择。

So it means that when you are trying to do more to look at the market opportunity, I don't think that is always the only the single choice.

比如，假设未来中国企业能够购买H200芯片。

Like for, let us assume that in the future, the Chinese companies can buy H200.

那么他们是否还会继续大量采购H20芯片？我认为我们并非只能在这两者间做单选题。

Well, whether they were still trying to buy some more H20, I don't think that is only our one option to choose from these two sites.

根据不同应用场景和市场需求，可以充分发挥不同芯片的性能优势。

Based on the different applications and different kind of markets, are able to use the different abilities of the chips.

因此我认为，基于这两方面的考量，我们或许能就如何增强经济关系的韧性达成某种共识。

So I will say maybe our kind of consensus based on these two sides of the garments about what we can do to have a better resilience on the relations of the economy.

但对相关企业本身而言，这确实是个非常好的消息。

But it's also a very good news for the companies themselves.

在中国，与美国相比确实是一个截然不同的市场。

In China, it's a really are different market compared with The US.

美国市场的芯片使用规模主要不及中国庞大。

The US mainly are not as big as the size of the market use of the chips.

当你观察中国市场时，它正在增长、整合，并从一个领域扩展到另一个领域。

When you're looking at the China's market, it's growing, it's integrating, it is spreading from one area to another.

因此如果这些芯片在中国使用，我认为其利用率会比以前更高。

So if they are used here in China, I think that the chips can make better use than before.

所以我不会把这个策略当作一个领域来争论。

So I I will say that the strategy, I will not argue that as a field.

但我认为这可能是延缓中国追赶美国步伐的一种策略，不过显然相比初期效果已大不如前。

But I I think that it may be a kind of strategy to slow down China's the state to catch up with The US, but clearly, it's not so effective compared with the initial time.

没错。

Right.

鲍威尔教授，实际上有报道称英伟达原本可能只需支付15%的份额，但现在提高到20%，从您的角度看，这一增长是否意味着出口条件收紧还是对未来利润更有信心？

And professor Powell, actually, reports suggested NVIDIA might pay only a 15% cut, but now it's 20% so does this increase signal tighter export conditions or greater confidence in future profits from your perspective?

从英伟达的角度来看，我认为这是重获部分市场份额的小小代价。

Well I think from NVIDIA's point of view it's a small price to pay to regain some market share.

别忘了英伟达在中国的市场份额从95%降到了零。

Let's not forget that NVIDIA's market share in China went from 95% to zero.

能通过任何产品重新进入这个市场都被视为额外收获。

Getting access to this market again with any product would be seen as a bonus.

英伟达CEO过去几个月明确表示，虽然公司未来预测不包括中国市场准入，但确实——实际上也确实——进行了强力游说，主张解除出口限制。别忘了英伟达专门为中国开发的芯片最终并未获得中国买家青睐，这向中国生态系统证明了美国人只允许向中国销售实质上的次级产品，这简直是侮辱，最终加速了中国自主发展。所以我认为英伟达显然会抓住任何机会重返这个全球最大市场。

And NVIDIA's CEO made that very, very clear over the last few months that whilst the company's forward forecasts did not include any access to the Chinese market, it would certainly and well, in fact it did it lobbied quite hard and argued the case that restrictions on its capacity to export should be lifted let's not forget the China specific chips that Nvidia did develop ultimately did not meet with much favor from Chinese buyers and in the end it proved to the Chinese ecosystem that the Americans were only going to allow what is in effect B grade product to be sold into China I mean it was an insult and ultimately contributed to the acceleration of China's own development So, look, I think NVIDIA quite clearly will grab at any opportunity to get a foothold back into the largest market globally.

事实上要向政府支付25%的费用实在是微不足道的代价。

And the fact that it's having to pay 25% to the government is really a small price to pay.

关于这一切我还想说，这明显引发了人们对美国公共政策诚信度的质疑，特别是在裙带资本主义指控方面。

The other thing I'll say about all of this is that quite clearly, it raises questions around the integrity of American public policy, particularly insofar as the allegations of crony capitalism are concerned.

很明显在美国，那些资金最雄厚的群体也最能接触政府决策者，最有能力推动——姑且这么说——政府政策朝对他们有利的方向改变。

And it is quite clear that those who have the deepest pockets in The United States also have the greatest access to government decision makers and have the greatest ability to encourage, shall we say, changes to government policy in their favour.

聊天休息室

The chat lounge.

聊天室以更轻松的方式解析热点问题的观点与看法。

The chat lounge unpacks views and opinions on hot issues in a more casual way.

是的。

Yeah.

他们是否想卖给中国是一回事，但中国市场是否接受或在多大程度上接受则是另一回事。

Whether they want to sell it to China is one thing, but whether the Chinese market will or to what extent if the Chinese market will accept it is another.

不过在深入探讨这点之前，让我先请马克·弗教授发言。

But before we dive into that aspect, let me turn to Professor Mark Fur.

对，你提到H200是第二好的AI芯片。

Yeah, you mentioned, H200 is the second best AI chip.

那么能否简要告诉我们，与那些降级版芯片（比如NVIDIA计划销往中国市场的H20）相比，它实际先进程度如何？

Then can you briefly tell us, how advanced it is actually compared with, those, like, H20, the watered down chip NVIDIA intended to sell to the Chinese market.

另外，与性能更强大的Blackwell芯片（仍受出口禁令限制）以及即将推出的Rubin世代相比，在性能、内存带宽、训练或推理能力方面存在多大差距？

And, compared with the more powerful Blackwell chips, which are still under, export ban and the upcoming rubin generation, how wide are the gaps, you know, in performance or memory bandwidth and training or inference capabilities there?

Twigene，在不深入技术细节的情况下，我认为可以说——或者说我的观点是——NVIDIA正在通过我称之为'n减一'的策略，实施类似波音公司的战略安抚手段。

Well, Twigene, without getting into a lot of the technical details, I think what we can say or what I would say is that what NVIDIA is doing is they're pursuing a Boeing strategy of strategic appeasement through what I call n minus one.

如果我们把Rubin视为最前沿的先进芯片，Blackwell是当前第二代，而Hopper（即AH200）则属于上一代产品。

So if we think about Rubin as the most advanced frontier chips, Blackwell is the second the current generation, and then Hopper, which is the AH200 is the older generation.

这可能是最容易理解的方式。

That might be the easiest way to think about it.

我记得很多年前与波音公司一些高管交谈时，他们提到要进入中国市场就必须采取这种战略安抚手段，因为需要技术转让。

So I remember many many years ago when I was talking with some senior executives from Boeing, and they said that what they had to do was pursue this strategic appeasement, because to enter the Chinese market they had to transfer technology.

但他们实际采取的做法——经过精心战略规划——总是提供n减一或n减二代产品，n代表最先进的技术代际。

But what they did, and they planned this very carefully and strategically, it was always n minus one or n minus two, n meaning the most advanced generation.

我认为英伟达现在采取的就是这种策略。

So I think this is what NVIDIA is doing.

某种程度上可以说这就是谈判艺术的精髓。

In some sense we could say this is the essence of good negotiation.

虽然各方都不完全满意，但都能获得部分所需。

No one is happy, but everyone gets some of what they need.

显然，我认为英伟达出于各种原因需要中国市场。

So obviously, I think for NVIDIA for various reasons, you know, it needs the Chinese market.

我认为它从战略上相信，可以抢占先机或至少延缓真正竞争对手在芯片性能前沿的崛起。

I think it believes strategically that, you know, it can preempt or at least slow down the emergence of true competitors at the frontier of chip performance.

但这里也需要指出，组件级别的优势并不必然意味着胜利，因为这些芯片最终要装入服务器机架进入数据中心，而中国已取得巨大进步，特别是华为，在构建芯片集群方面，虽然单个芯片不如英伟达最先进的产品，但系统整体性能相当，只是功耗更高。

But here also it's important to point out that component level superiority does not necessarily mean you win, because again these chips go into server racks that go into data centers, and China's made enormous progress, especially I think Huawei, in terms of coming up with clusters of chips that are individually not as good as the most advanced NVIDIA chips, but the system performs about as well, and however it does use more power.

但同样关键的是，在中国，每千瓦时的成本要低得多，而且政府对战略产业提供支持。

But again, here's the thing that with China, China has a much lower cost per kilowatt, and the government also provides support to strategic industries.

所以这个问题并不容易回答，但我想向观众强调的关键点是：美国正在推行一种战略。

So I think it's it's not an easy question to answer, but the key point I think it's important to make for our audience is that The US is pursuing one strategy.

英伟达身处这一战略之中，但遵循着自身战略绥靖的逻辑，我认为这最早是由波音公司开创的。

NVIDIA lives inside this strategy, but is pursuing its own logic of strategic appeasement, which I think was pioneered by Boeing.

这就是正在发生的事情。

And you know, this is what's going on.

而正如你暗示的重大问题是：中国将如何应对这种情况？

And the big question as you alluded to, how will China react to this?

关于中国，我认为我们需要考虑两个层面，对吧？

And China, I think we have to think there's two pieces, right?

显然，一方是中国政府，另一方则是作为英伟达客户的中国科技公司。

Obviously, there's the Chinese government, and then there are the Chinese tech companies that are the customers of NVIDIA.

也就是阿里巴巴、腾讯、百度等这些企业。

So, you know, the the Alibaba's, the Tencent's, the Baidu's, you know, etcetera.

我相信我们稍后会讨论到这点。

So I'm I'm sure we'll get to that.

但我认为这里的关键词是'战略安抚'。

But, I think the key phrase here is strategic appeasement.

没错。

Right.

那么让我们看看中国市场或北京方面会如何回应这种战略安抚策略。

Then let's find out how the Chinese market or Beijing would respond to such a strategic appeasement strategy.

我们先从官方的回应开始。

Let's begin with the official response from the authorities.

目前看来外交部和商务部的官员对此事表态并不积极。

It seems foreign affairs officials and commerce officials are not so ardent about making comments on it.

他们仅表示已注意到相关报道。

All they said was, we've noted the reports.

博士。

Doctor.

周博士，您能谈谈对这种回应的解读吗？

Zhou, can you tell us your interpretation of such a such response?

我认为政府目前仍在等待相关公告或事实依据，毕竟这只是新闻消息，据我了解尚未发生实质性的交易。

Well, I think that for the government, they really are still waiting for, you know, some of this announcement or kind of facts because you just heard from the news, but there's no physical kind of transactions really happened in my understanding.

在缺乏具体事实的情况下，我们很难简单判断这是好事还是坏事。

Maybe without that kind of facts, we can not just say it's a good thing or a bad thing.

但直觉上这是件好事，因为中美双方都在佛山有所动作，这向市场传递了非常积极的信号。

But feeling that it is a good thing because, you know, both China and The United States will meddle in Foshan and we have some, you know, sending some very good sentiment to the market.

若能开放市场并解除部分限制，对双方营造更良好的互动氛围都是利好消息。

So if we're able open the market to lift some of the restrictions, I think it's a good news for both sides to have a better atmosphere for, you know, coping with the other side.

还有个问题您可能尚未提及，是关于所谓Edge 20的安全性问题。

And it's even our another question, maybe you haven't touched on that yet, is about the safety or security problems, about the so called Edge 20.

所以我们不知道Edge 200是否也会面临同样的情况。

So why We don't know whether it will be the same, Edge 200.

实际上，当市场试图做出反应时，他们必须在利润与安全之间取得平衡。

So actually, when the market is trying to do some response, they have to balance profits with security.

这才是我们真正面临的大问题。

And that's really our big issue.

对政府而言，我们不仅仅是想向企业本身收费。

For the government, we are not just trying to charge for the companies themselves.

所以我认为，可能有些人对这则新闻抱有非常复杂的情绪。

So I I think that maybe some of these very sophisticated feelings for this news.

是的。

Yes.

确实如此。

Indeed.

你提到了一些安全隐患。

You said some security concerns.

实际上，英伟达公司目前仍在接受关于其所谓后门程序的调查，对吧？

Actually, NVIDIA was, it's still under investigation into its alleged backdoors, right?

那么，周博士，这些挥之不去的担忧会在多大程度上影响客户信心呢？

So how much might the lingering concerns affect customer confidence, Doctor.

周？

Zhou?

我认为这是个非常严重的问题，因为如果企业可以向你出售产品，却又能以某种方式控制它，甚至能直接停止你使用芯片，这将极其危险——尤其是当这些芯片被用于金融系统或其他实时关键业务时。

I think that is a very serious problem because if the companies can just sell you something, but they can control that from a certain kind of way, and they can just shut down your using of the chips, it will be very dangerous, especially when the chips are used for some important thing, like for the finance errors or other kind of, you know, the on time activities.

我认为这不仅在理论上，实际上也非常危险。

I think it's very dangerous and not just virtually.

我们不可能把所有系统都重启一遍。

We cannot just restart everything again.

这将迫使我们直面大量潜在风险。

And it will give us a lot of expulsion on the possible risks.

基于此类风险，我认为我们绝不能放任所谓的后门程序存在。

So according to this kind of risk, I don't think that will just leave the so called the backdoor there.

我们必须解决这个问题。

We have to fix this problem.

这也正是政府需要做的——尝试从芯片供应商那里获取更多信息，并规范这些潜在危险的可能性，因为这不仅关乎企业自身。

And that is also, you know, what the government need to do to trying to have more information from the suppliers of the chips and also to regulate these possibilities of possible dangerous, because it's well, not just for the companies themselves.

其中一些企业正在服务市场。

Some of these companies are serving the market.

如果发生不良事件，我认为这些断连的溢出效应将会广泛蔓延。

And if there's something bad thing happened, I think that is a widespread of those spillover effects of these disconnections.

如果他们试图改变芯片功能而非仅仅是停止它，同时植入不良信息或虚假信息，我认为我们也需要警惕这些可能性。

And if they're just trying to, you know, to change the chip's function, not just to stop it, while putting some more bad information or disinformation, I think that is also possible for us to be alert about the possibilities.

因此，如果服务器由这些芯片支持，我不认为这对中国市场中的某些核心业务会是可行的。

So I don't think that will be even possible for some of the core businesses in the market in China if the servers are supported by those chips.

所以我们必须明确并确保在真正接受这些芯片作为重要业务和活动的支持之前，这些事情不会发生。

So we have to make it very clear and guaranteed to ensure that these things will not happen before we can really accept those chips as, you know, the support for the some of those very important businesses and activities.

鲍威尔教授，您对此持相似态度吗？

Professor Powell, do you share a similar attitude toward this?

你能理解吗

Can you understand

听着，当局最近已明确表示对安全问题存在担忧

Look, the authorities have made it clear in recent times that there are concerns around security questions.

毫无疑问，这些担忧已影响到产业界的战略决策者，他们需要高度关注此类问题

And no doubt those concerns have worked their way through to strategic decision makers across industry who need to be very mindful about these sorts of issues.

当然，另一个维度是这些芯片具有我们可称之为的地缘政治半衰期

The other dimension to all of this, of course, is that there is what we could call a geopolitical half life to these chips.

即它们本质上是一种技术产品，存在使用期限限制

And that is that they ultimately are a technical product that has a limited use by date.

未来无法保证能获得升级支持

There are no guaranteed upgrades down the track.

极有可能出现一系列限制措施，特别是在固件方面

It's highly likely that there are going to be a whole range of restrictions, particularly in firmware.

我预见企业不可能不担忧未来缺乏生态系统支持的问题

I can't foresee that enterprises won't be concerned about the absence of ecosystem support down the track.

这类技术最多两到三年就会过时。

And these sorts of technologies are obsolete in two to three year cycles at best.

因此一旦考虑到这点，我认为企业面临的问题之一是如何利用这个所谓的短期机会窗口。

So once you take that into account, I think one of the issues you've got for enterprises is how they make use of what you might call a short term opportunity window.

这些芯片的性能优于现有的国产设备。

These chips are a better performing chip than available domestic equipment.

它们价格相对低廉，一旦上市就能立即部署，适用于大量中端模型的训练和推理工作。

They're relatively inexpensive And they can of course be, once they're available, you can deploy them immediately for a whole raft of mid range model training and inference work.

我认为小型企业，尤其是初创公司很可能会抓住这个机会，这将成为主要市场。但更具战略眼光的买家不太可能制定将这项技术嵌入关键系统的策略——没人会用这种技术来构建、设计或架构关键系统。充其量，大型企业可能将其视为临时过渡方案，但绝不会作为战略支柱。

So I think smaller firms, particularly startups are likely to leap at this opportunity and that will be the obvious marketplace but your more strategic buyers are unlikely to be developing strategies that would see this technology embedded as part of a mission critical system no one's going to build or design or architect mission critical systems around this particular technology at best I think larger operations may see them as stop gaps but they certainly won't be strategic anchors.

国内研发将继续加速。

Domestic R and D will continue to accelerate.

这些芯片的供应肯定不会减缓研发进程。

It certainly won't be slowed down by the availability of these chips.

我认为最终政府和战略技术参与者都会意识到——或至少会在决策中考虑到——使用这种芯片存在重大运营风险，而且未来还存在地缘政治武器化的未解风险。

And I think ultimately both the government and also strategic technology players will realise or at least have baked into their thinking that there are significant operational risks going forward with this particular chip, but also there are unresolved risks around geopolitical weaponization down the track.

这将使英伟达很难超越一个相对有限的份额。

That's going to make it very difficult for NVIDIA to go beyond a relatively modest share.

这里是粉笔休息室。

This has been the Chalk Lounge.

稍后回来时，我们将探讨这是否会促使其他政府或公司重新考虑他们对华技术出口的限制。

When we come back, check out whether this will prompt other governments or companies to reconsider their own restrictions on tech exports to China.

别走开。

Don't go away.

狂风呼啸着。

The strong wind was howling and whistling.

他是十九世纪中叶第一位从耶鲁大学毕业的中国公民。

He was the first Chinese citizen to graduate from Yale University in the mid nineteenth century.

我出生于11月17日。

I was born on the November 17.

她的五官很突出。

She had prominent features.

我们中有三人已到了能帮上忙的年纪。

Three of us were old enough to lend a helping hand.

他游走于两种截然不同的文化之间，并进一步实现梦想，促进中美两国人民之间的理解。

He navigated between two vastly different cultures and moved further to realise his dream and promote understanding between the people of China and The United States.

解明兴是汉阳人。

Jie Ming Xing was a native of Hanyang.

我并未意识到危险。

I realized no danger.

中国正在真正觉醒。

China is really awakening.

快来加入我们，通过自传《我的中国与美国生活》一同探索容闳的非凡人生旅程。

Come and join us in discovering the incredible journey of Yong Weng in his autobiography, My Life in China and America.

请访问radio.cgtn.com及各主要播客平台收听Audible有声故事。

Check out the Audible stories on radio.cgtn.com and all major podcast platforms.

只需搜索播客《书海无涯》，即可找到《我的中国与美国生活》。

Just search for the podcast Books and Beyond and find My Life in China and America.

欢迎回到聊天室

Welcome back to the chat lounge.

我们继续讨论华盛顿允许英伟达向中国出售其先进的H200人工智能芯片的话题

We continue our chat on Washington allowing NVIDIA to sell its advanced h 200 AI chips to China.

马克教授，黄仁勋最近实际上表示他不太确定中国是否需要这些旧款芯片

To professor Mark, Jensen Huang actually said recently he he was unsure if China would want the older chips.

你认为他这只是出于谨慎，还是你对他的言论有其他解读？

Do do you think he's just being, you know, cautious, or what's your interpretation of his words?

你对此有何预期？

And what's your expectation here?

彼得，你知道，我认为这真的取决于具体情况

Well, Peter, you know, I think it really depends.

对吧？

Right?

但请让我先简要谈谈中国政府和官方在此问题上的立场

But let me just talk a little bit about China's position here, the government position first.

我认为中国正面临一个棘手的问题，因为其有两个表面看似矛盾的目标。

I think China's facing a difficult problem with this because it has two goals that on the surface appear contradictory.

第一个目标是保持中国科技企业当前的强大。

So the first goal is to keep Chinese tech firms strong today.

这包括阿里巴巴、腾讯等企业，既要保持它们的全球竞争力，又要快速推进其AI模型的发展。

So this would be the Alibaba's, the Tencent's, to keep them globally competitive, but also to really advance quickly with their AI models.

其中的逻辑其实非常简单。

And the logic is really very simple.

如果拥有最好的芯片，就能构建更好的模型。

If you have the best chips, you can build better models.

如果能构建更好的模型，就能产生最大的经济影响。

If you can build better models, you can have the biggest economic impact.

如果我们接受AI是新电力这个观点，想想世界在有电之前和之后的变化。

So if we accept this idea that AI is the new electricity and think about what did the world look like before electricity and after electricity.

这些科技企业能否真正推动生产力、全球竞争力、出口以及全球影响力至关重要。

That it's really crucial that these tech firms are able to really drive productivity, global competitiveness, exports, as well as global influence.

因此，如果中国企业使用性能较弱的芯片进行训练，他们推出的模型也会较弱。

So if Chinese firms train on weaker chips, they ship weaker models.

而如果他们推出较弱的模型，就会失去发展势头。

And if they ship weaker models, they lose momentum.

这里存在一定的路径依赖问题。

And there is a certain path dependency here.

所以一旦失去这种势头，可能很难再重新获得。

So if they lose that momentum, it might be very hard to regain.

为了实现这一目标，政府会欢迎英伟达的芯片。

So for this goal, the government would welcome NVIDIA chips.

然而第二个目标是通过缩小芯片差距来降低长期脆弱性。

Now the second goal though is to reduce long term vulnerability by closing the chip gap.

显然，依赖对手提供前沿计算能力会带来非常现实的胁迫风险，正如我们所看到的美国惯于将全球金融体系、航运和其他关键技术武器化。

And obviously depending on arrival and adversary for frontier compute creates a very, very real coercion risk as we've seen The US with its track record of weaponizing the global financial system, ships, other key technologies.

因此，对芯片的控制实质上是对模型性能上限的控制，进而也是对经济潜力和国家安全上限的控制。

So control over chips is really a control over the ceiling of model power and therefore over the ceiling of both economic potential as well as national security.

正因如此，北京方面自然在推动国产GPU、NPU（无论你如何称呼它们）、先进制造业的发展，以及实现独立自主所需的漫长产业升级之路。

Because of this, Beijing of course is pushing domestic GPUs, NPUs, whatever you want to call them, advanced manufacturing and the long industrial arc needed to stand independently.

因此，这两个目标看似再次将我们拉向相反的方向，而无法同时满足这两个目标。

So these two goals then again appear to pull in opposite directions and one cannot satisfy both of these goals at the same time.

所以我认为真正将要发生的不是二选一，而是寻求平衡。

So I think really what's going to happen is not choosing one or the other, but balance.

这种平衡可以这样表述：中国必须在今天取得足够进展以抵达未来，又要在明天保持足够独立性以实现完全的技术主权。

And the balance can really be said in this way, China must progress enough today to reach the future and be independent enough tomorrow to be fully sovereign in it.

因此我认为，我们必须密切关注政府具体决策的实施情况。

So I think again, have to watch how the specifics play out in what the government decides to do.

如果我们从企业层面观察，就必须理解英伟达销售的不仅是一块芯片。

And if we look at the firm level, we have to understand that what Nvidia sells is not just a chip.

如果你像我这样的老一辈人，最好的描述方式就是回忆个人电脑早期时代。

And if you're an old person like me, the best way to describe it is in the old PC days.

那时被称为'Wintel联盟'。

It was called WinTel.

那是使用Windows操作系统及所有运行在Windows上应用的英特尔芯片。

It was an Intel chip using the Windows operating system and all of the apps that ran on Windows.

所以如果没有这些，真的很难提高生产力。

So if you didn't have that, it's really hard to be productive.

但我认为中国大多数关键参与者，尤其是那些将赢得未来的企业，实际上都建立在NVIDIA的技术栈上。

But I think most of the key players in China, again the ones that are going to win the future, have really built on the NVIDIA stack.

他们使用NVIDIA芯片、CUDA、所有软件库等等这些东西。

So they NVIDIA chips, UDA, all the software libraries, all of these things.

再想想这有多困难。

And again, think about how hard it is.

一旦你学会了使用Windows，再回到DOS时代，你会记得所有命令行，然后必须去学习一个全新的系统。

Once you learn to use windows, go back to DOS and you remember all the command lines and you have to go learn something completely new.

那真的非常非常困难。

That's really, really difficult.

我认为在效率和速度方面，答案已经很明确了。

And I think in terms of just efficiency and speed, I think it's clear.

我认为所有中国科技公司，如果没有政治考量的话，都会选择英伟达。

I think all of the Chinese tech companies that they had, there were no political considerations, would go with Nvidia.

这也是像摩尔线程这样刚上市公司的机会，他们采取的策略是通过宣称你仍可使用CUDA和英伟达的其他功能，只需更换芯片，来降低转型难度。

And this is the opportunity as well, I think, for companies like MoreThreads, which just IPO ed, taking an approach of making it easy to transition by saying you can still use CUDA, you can still use all these other things you had in Nvidia, but all you have to do is swap out the chip.

我认为我们还需观察，但这确实是个充满挑战的局面，因为有两个非常重要的目标需要平衡。

I think we have to see, but I think this is a challenging situation again because there's these two very very important goals that have to be balanced.

再从英伟达的角度看，作为上市公司，他们需要增长故事，希望提高利润和利润率，我们拭目以待。

And then again from Nvidia's perspective, obviously a publicly listed company, they need a growth story, they want to increase profits, ideally increase margins, and we have to see.

我认为只要能在中国销售的产品，他们肯定乐意卖，而且更倾向于销售高利润、高价位的产品。

I think whatever they can sell in China, I'm sure they'll be happy to sell and they'd much prefer to sell the higher margin, higher priced items.

当然，这又受到政治因素的制约。

And again, this is constrained by politics.

是的。

Yes.

当然，H200芯片如你所说有很多优势。

Of course, the h 200 chips got a lot of advantages like you mentioned.

但这25%的额外费用或关税税是否会影响

But will this 25% extra fee or levy from Tariff tax.

没错，会影响买家兴趣或影响

Right, affect buyer interest or affect

嗯，我们还得观察。

Well, we have to see.

再想想关税问题，传统微观经济学理论认为关税由消费者承担，但英伟达若表示愿意承担部分费用以保持价格竞争力，我想这并不令人意外。

And again, we think about it like a tariff, conventional microeconomic theory says that tariffs are paid by the customer, but you know it wouldn't be surprising I think if Nvidia said look we'll cover a part of it, you pay a part of it to make it more cost competitive.

我们都知道这里面有各种操作空间，对吧？

We also know that there are all kinds of games you can play, right?

比如在发票上做文章，将部分金额归为服务费而非芯片本身，以最小化15%或25%的税率影响。

In terms of what you call it invoicing, how much of it is services, how much is the actual chip to minimize that whatever that percentage is 15 or 25%.

因此我认为这对英伟达而言更多是战术决策，他们需要在战略层面平衡——既要安抚中国市场，又要维系美国政府在这个竞争性相互依存关系中的满意度。

So I think this is really going to be more of a tactical decision for Nvidia and again you know they have to balance, I think against strategic appeasement on the Chinese side while keeping The US happy again with this you know, this idea of government interdependence under rivalry.

所以他们必须如履薄冰。

So, you know, they have to walk a tightrope.

再谈谈其他方面的影响。

Then a little bit on this other impact.

实际上，我认为鲍威尔教授已经提到过一些，在美国实施更严格禁令期间，中国企业确实加速了本土研发。

Actually, I think Professor Powell mentioned a little bit on this, but during the period of stricter US bans, Chinese firms actually accelerated local research and development.

现在限制有所放宽，这是否会在某种程度上减缓中国对高端国产芯片的追求，周博士请先回答。

Now that restrictions are easing, will this slow, to any extent, China's drive for high end homegrown chips to Doctor.

周博士，请先发言。

Zhou first, please.

嗯，是的，我认为对一些尚未确定技术路线的公司来说，这确实是可能性之一。

Well, yeah, I would say that is part of the possibilities for some companies who are really want to, you know, or they haven't decided what kind of technology they want to follow.

因为如果采用已经非常成熟的技术，就能跳过研发中最艰难的部分，这确实会更轻松。

So it's easier because if you are going to use already very good technology, it will be easier for you to skip some of this most hard part of the research.

但实际上，想要往更高更深层次发展并不容易。

But actually, that is not so easy for you to go deeper and higher.

就像专家们提到的，对人工智能领域那些核心企业、巨头公司来说，他们必须考虑未来生态、整体商业环境等因素。

So for many companies, like the experts have mentioned about some of those core companies, the bigger companies, the giant, the players in the artificial intelligence, they will think about that, the future, the ecosystems, about the whole businesses and the environment.

对这些公司而言，他们并非频繁地在不同技术路线间摇摆不定。

For those companies, they were not just trying to shift frequently from one side to another.

我想补充另一个考量因素，特别是对小公司使用Edge 200芯片的可能性而言，关键在于成本问题。

I wanna say another factor we have to consider about the possibilities for the companies, especially the small companies, use the chips of the Edge 200 is about the cost.

Edge 200芯片的价格非常昂贵。

The price of the Edge 200 is very expensive.

可能在早些年，小企业很难负担购买这些芯片的费用，更不用说达到支持其人工智能计算模型所需的规模。

Maybe in the previous years, it is very difficult for smaller companies to afford buy the chips and to put them in the, or, you know, a scale large enough to support their artificial intelligence, the model of computation.

所以实际上这对他们来说仍然非常困难。

So actually it's still very difficult for them.

但现在我认为既然Namita需要承担额外成本，或许最终会分摊部分使用成本。

But now I think that since Namita have to pay more for the extra cost, maybe it will end some more cost of using that.

因此对于那些试图转换技术路线的公司来说，适应这种局面相当困难。

So for the companies who are trying to shift from one side to another, it is very difficult for them to adapt to the situation.

但Mimeta已经基于代码系统建立了非常重要的生态系统来支持这些芯片的使用。

But Mimeta has already built a very important ecosystem based on the coda, on the code based system to support the use of these chips.

因此对于企业来说，我认为其中一些公司的立场非常明确。

So for companies, I think that some of them are very clear.

他们并非仅仅试图转向艾瑞达，尽管购买X-two 100对他们来说或许可行，但对小型企业而言，这仍然非常困难。

They were not just trying to ship to the Erida as although maybe it's possible for them to buy the X-two 100, but for smaller companies, it is still very difficult for them to do that.

没错。

Right.

鲍威尔教授，您是否认为这有点像一把双刃剑？

Professor Powell, do you think it's kind of like a double edged sword here?

您认为这对中国来说是好事还是坏事？

Do you think it's a good thing or a bad thing for China?

说实话，我并不太担心这个问题。

Look, wouldn't be too worried about it, to be honest with you.

主要原因是总体风险评估将起决定性作用。

And the main reason is that the overarching risk assessment will dominate.

这项风险评估既涉及安全问题，最终也关乎地缘政治的半衰期。

The risk assessment touches on both questions of security and ultimately in terms of the geopolitical half life.

当然，我们还面临一个交叉现实：这类技术和硬件的实际使用寿命相对较短。

We also, of course, have the intersecting reality that this kind of technology, this kind of hardware, has a use by life that's actually relatively short as well.

这意味着就中国自身发展而言，其当前芯片性能仍低于其他地区的顶尖芯片。

Which means that insofar as China's own development, its current chip performance is below the top end available chips from elsewhere.

但这个性能差距正在以持续加速的速度缩小。

But the pace at which that gap is closing continues to accelerate.

因此无论A2100芯片能否提供支持，特别是对中国小型企业而言，它根本不会改变战略格局。

And so whatever the A2100s may or may not provide, particularly for smaller firms in China, it simply will not alter the strategic calculus.

另一个需要记住的关键点是，人们常常会过度关注整个生态系统中某个单一要素——虽然这些要素对实现我们所需的AI性能输出确实必要。

The other point to remember in all of this is that there is often a temptation to become obsessed with one element of the overall ecosystem that is actually necessary to deliver the AI performance outputs that we're looking for.

尽管英伟达芯片在纯粹浮点运算性能方面表现最佳，但当你考虑整个系统架构时——我指的是电力成本相对差异等因素。

And as much as the NVIDIA chips deliver the highest performance levels in so far as pure FLOPs are concerned, when you take into account the overall system architecture and I mean by that the relative difference in electricity costs.

比如华为昇腾架构所采用的软硬件协同设计，就能实现精细化调优和优化。

The hardware software co design that has been undertaken for instance with Huawei's ascend architecture enabling fine tuning and optimization.

这正好弥补了每瓦特低浮点运算性能的不足。

And that compensates for low FLOPS per watt.

我们拥有算法效率的优势，对吧？

We've got algorithmic efficiencies, right?

因此在软件层面，即使半导体或芯片性能稍逊，也能实现更高的性价比。

So really at a software level, getting much more bang for buck, even if your semiconductors or chips are inferior.

最终中国建立了垂直整合的国内供应链体系。综合来看，可以说中国芯片虽性能较弱，但数量更充足、价格更低廉，从整个生态系统角度评估，其实际表现优于美国芯片——无论是单独使用还是与美式生态系统对比。

And ultimately we've got vertically integrated domestic supply chains in China So when you bring all of that together what we probably have is an environment where you can say that the Chinese chip is weaker but it's certainly more abundant it's cheap and in the context of a whole of ecosystem perspective it actually outperforms the American chip when treated by itself and in comparison with US ecosystem.

所以归根结底，不能仅凭单个芯片来评判AI能力和性能。

So, you can't judge AI competency and performance in the end on a chip by chip basis.

必须从生态系统对抗生态系统的维度来审视这些问题。

You've actually got to look at these things from an ecosystem versus ecosystem basis.

目前美国能保持竞争优势的关键因素，恰恰在于其芯片性能位居世界前列——即便不是全球第一。

And the main thing that's keeping the Americans in the game at the moment is actually the fact that its chips are amongst the best in the world, if not the best in the world.

这弥补了其电力成本高昂的缺陷。

And that compensates for the fact that it has high electricity costs.

实际上他们并未实现硬件软件设计的协同优化。

It actually doesn't optimise hardware, software design.

归根结底，它还拥有一个很大程度上属于专有的商业模式，这种模式严重倾向于迎合华尔街的要求，需要以非常特定的方式实现盈利。

And ultimately, it also has a business model, which is in large part a proprietary business model that's heavily geared towards answering to the dictates of Wall Street, that needs to monetise in a very particular way.

在这种环境下，中国的环境更具竞争力。

And in that environment, the Chinese environment is more than competitive.

关于这一切我要说的最后一点，这涉及到它如何在整个经济中实际运作。

The last thing I'll say on all of this, and this goes to how it actually works across an economy.

这不仅仅是拥有出色的聊天界面。

It's not just about having great chat interfaces.

实际上关键在于基础技术——即结合大数据的快速高性能计算——能多快在真实工作流程中推动功能和运营效率的提升，无论是在政府行政管理、供应链还是生产管理系统中。

It's actually about how quickly the fundamental technology, which is rapid high powered compute with big data, can drive functional and operational efficiencies in real workflows, whether they're in government administration, or in supply chains or in production management systems.

而正是在这些领域，中国已经表现出色。

And it is in these areas where China is already excelling.

所以你看，这是美国人的一种姿态。

So look, it's a gesture from the Americans.

这是试图让美国的人工智能泡沫再维持一段时间的努力。

It's an attempt to keep the American AI bubble afloat for a little bit longer.

这显然是釜山会议后地缘政治缓和的一部分。

It's clearly part of a geopolitical detente coming out of Busan.

但在战略层面，我认为这并不那么重要。

But at a strategic level, I don't think it matters that much.

对。

Right.

考虑到华盛顿方面或英伟达本身有一些既定目标，你认为这种转变能在多大程度上实现其目标？

And given that there are some intended goals by Washington, either by Washington or by NVIDIA itself, then how much or to what extent do you think the shift can achieve its goals?

比如这能为美国政府及英伟达带来多少收入或战略优势？

Like how much revenue or strategic leverage might this generate for the US government and for NVIDIA?

不知道鲍威尔教授或马克教授是否对此有所预估。

I don't know whether Professor Powell or Professor Mark, you've got any estimation there.

鲍威尔教授？

Professor Powell?

听着，我不认为这能给美国带来多少优势。

Look, I don't think it's going to give The United States much leverage.

别忘了，这些芯片很大程度上还依赖于上游的稀土原料供应。

Let's not forget that a lot of these chips are also dependent on upstream inputs of rare earths.

就杠杆效应而言，我怀疑实际能发挥的作用相当有限。

Insofar as leverage is concerned, I doubt that there is actually that much available.

这实际上传递的信号是试图缓和当前已存在的紧张关系。

What it signals is actually an attempt to take a little bit of heat out of the relationship that has been in place.

这并非坏事，但不会从根本上改变正在我们面前展开的战略格局。

That's not a bad thing, but it's not going to fundamentally change the strategic parameters that are now unfolding before us.

而从中国的视角来看，观察全球局势变化，并汲取过去十年的经验教训。

And from a Chinese perspective, watching what has been happening around the world, and of course taking on experiences over the course of the last ten years.

必须清醒认识到美国仍在推行遏制中国的政策。

It needs to be mindful that The United States continues to have a policy of containing China.

最近发布的《国家安全战略》文件再清楚不过地表明了这点，我们不应忘记。

That's no more clear than in the recently released National Security Strategy, let's not forget.

正因如此，这类短期商业机会最终都会经过风险评估的过滤。

And as a result of that, these sorts of, let's call them short term commercial opportunities, will ultimately be filtered through a risk lens.

虽然我认为不会对企业购买这些芯片设置巨大障碍，但购买这些芯片的企业主要会基于这样的理解行事：你不想围绕这些芯片设计整个架构，它们不过是个短期权宜之计。

And whilst I don't think that there are going to be massive barriers put into place for firms to purchase these chips, firms that do purchase these chips will largely be doing so on the understanding that you don't want to be designing entire architectures around them and they are nothing more than a short term stop gap opportunity.

聊天室以更轻松的方式剖析热点问题的观点和看法。

The chat lounge unpacks views and opinions on hot issues in a more casual way.

嗨，马克教授，从经济角度他们能实现这个目标吗？

Hi then, Professor Mark, economically can they meet the goal?

就像鲍威尔教授刚才提到的，给人工智能市场挤掉一些泡沫。

Like Professor Powell just mentioned, taking some bubble out of the AI market.

云南，我认为，有些人把这种情况比作罗夏墨迹测试，所以有人可能看到蝴蝶，另一个人可能看到别的东西。

Yunnan, I think this is, some people describe the situation as a Rorschach test, so one person can see a butterfly, someone else can see something else.

我想说这可能是三赢局面，因为在我看来这是个试点，是华盛顿正在尝试的事情。

I guess I would say that this could be win win win, because you know again, to me I would say this is a pilot, this is something that Washington is trying out.

两个战略竞争者能否在不损害美国认为重要利益的情况下继续保持联系？

Can two strategic competitors still connect without one party here in The US of losing some things that it considers important.

我认为这实际上是一种认知，再次表明美国承认它无法阻止、也无法减缓中国的发展。

And I think it's actually a recognition, again acceptance, that The US recognizes it cannot stop, it cannot slow down China.