注意：本在线文稿仅提供中英文对照，仅供参考；
如需所有版本（英文+中英文）和格式（PDF+Word）的跟读文稿及对应的音频（高码WAV）视频（高清MP4）下载（年度或永久更新，请按需选择），请至小店按需选择1-5号购买：
In a move that could mark a potential breakthrough for China's AI ambitions,
此举可能标志着中国人工智能雄心的潜在突破，
Huawei has unveiled new hardware that it says could deliver world class computing power,
华为发布了一款新硬件，称其可提供世界级的计算能力，
without having to rely on the advanced chips of American tech giant NVIDIA.
而无需依赖美国科技巨头英伟达的先进芯片。
The Chinese telecoms equipment company said it had developed
这家中国电信设备公司表示，它已经开发出
what it called the world's most powerful Super Node computing cluster
其称之为全球最强大的超级节点计算集群
using local chip technology.
采用的是本土芯片技术。
And Henry Yin joins me with the details.
Henry Yin 将为我们带来详细报道。
Yes, Grace.
是的，Grace。
Well, that's the Atlas 950 and Atlas 960.
嗯，那就是Atlas 950和Atlas 960。
And Huawei says they can link thousands of its chips together
华为表示，它们可以将数千颗自家芯片连接在一起
and giving the systems much greater computing power.
从而为系统提供更强大的计算能力。
It's also rolling out new variants of its Ascent AI chip.
它还在推出其昇腾AI芯片的新型号。
Next year, it will launch the Ascent nine fifty, and in 2027,
明年将推出昇腾950，2027年，
the nine sixty followed by the nine seventy in 2028.
将推出960，然后在2028年推出970。
Well, the chip will be powered by the firm's own high bandwidth memory.
嗯，该芯片将由公司自研的高带宽内存提供支持。
NVIDIA CEO, Jensen Huang,
英伟达首席执行官黄仁勋，
says Washington and Beijing have larger agendas to work out
表示华盛顿和北京有更大的议程需要解决
as the tech giant is caught in the intensifying competition between the two countries.
因为这家科技巨头正陷入两国日益激烈的竞争之中。
Mister Huang was asked in London about a Financial Times report
黄先生在伦敦被问及《金融时报》的一篇报道
saying China's Internet regulator ordered top tech firms to stop buying
报道称**互联网监管机构**顶级科技公司停止购买
NVIDIA's AI chips and cancel existing orders.
英伟达的AI芯片并取消现有订单。
Some of these named companies reportedly include TikTok parent ByteDance
据报道，这些被点名的公司包括TikTok的母公司字节跳动
and ecommerce giant Alibaba.
以及电商巨头阿里巴巴。
He said he was disappointed about the directive,
他表示对该指令感到失望，
but that he will remain patient while the two countries iron out their issues.
但他会保持耐心，等待两国解决他们之间的问题。
He added that NVIDIA can only be in service if a country wants them to be.
他补充说，只有当一个国家需要英伟达时，英伟达才能为其服务。
And NVIDIA has been navigating tensions between The US and China
英伟达一直在美国和中国之间的紧张关系中斡旋
as it tries to meet growing demand for their AI chips.
同时努力满足对其AI芯片日益增长的需求。
Shares of the company fell 2.6% overnight.
该公司股价隔夜下跌了2.6%。
Now for more, we are joined by Alex Capri, a senior lecturer with NUS Business School.
Alex, thank you so much for joining us this afternoon.
Alex，非常感谢您今天下午加入我们。
So Huawei has unveiled its new AI chip technology
那么，华为已经发布了其新的人工智能芯片技术
aiming to challenge NVIDIA's market dominance.
旨在挑战英伟达的市场主导地位。
So how significant is this development,
那么，这一进展有多重要，
particularly when considering
特别是考虑到
that cluster based computing like Super Node remains distinct from traditional AI chips?
像超节点这样的基于集群的计算仍然有别于传统的人工智能芯片？
Right.
是的。
So what we're talking about here is not taking one NVIDIA chip
所以我们这里讨论的不是拿一颗英伟达芯片
and taking one Ascend chip and having that Ascend chip perform at the same level as the NVIDIA chip.
并且让单颗昇腾芯片的性能达到与英伟达芯片相当的水平。
What, uh, Huawei is doing here is they are clustering.
华为在这里所做的，是集群化。
Right?
对吧？
They're taking thousands of Ascend chips, as you mentioned,
正如你所说，他们将数千颗昇腾芯片，
and they're cobbling them together in such a way that they can maximize performance.
以一种可以最大化性能的方式将它们组合在一起。
And so I think what this tells us is when it comes to pushing AI innovation, you know,
因此我认为，这告诉我们的是，在推动人工智能创新方面，
when it comes to other, uh, applications of semiconductor compute power,
当涉及到半导体算力的其他应用时，
China does not necessarily have to come up with something that's as good on a piece by piece basis.
中国不一定非要拿出在单体性能上同样出色的产品。
It has to be good enough to perform whatever the end game is.
它只需足够好，能够实现最终目标即可。
So I think, you know, that's a distinct possibility here.
所以我认为，这在这里是一个明显的可能性。
Um, we'll have to, uh, wait and see the outcomes.
嗯，我们必须拭目以待结果。
Mhmm.
嗯哼。
Well, in the meantime,
嗯，与此同时，
Huawei has also introduced a series of AI chips including the Atlas 950, 960,
华为还推出了一系列AI芯片，包括Atlas 950、960，
and 970.
和970。
So are you convinced that these chips have the potential
那么，您是否相信这些芯片有潜力
to rival NVIDIA's established technology in the AI space?
在人工智能领域与英伟达的成熟技术相抗衡？
Well, again, uh, you know, we have to wait and see.
嗯，同样地，我们必须拭目以待。
We have to wait and see how the the application of all these chips, you know,
我们必须看看所有这些芯片的应用情况，
functioning together, these Ascend chips, uh, perform.
这些昇腾芯片协同工作时的表现如何。
So it's early days now.
所以现在还为时过早。
Right?
对吧？
We still have to see we still have to see to what extent, certainly,
我们仍然需要看看，在何种程度上，当然，
when it comes to producing Ascend chips, you know,
在生产昇腾芯片方面，
whether or not HiSilicon is the designer or SMIC as the as the, you know, the fab,
无论海思是设计方，还是中芯国际作为代工厂，
whether they're still gonna be reliant on Western technology, like deep ultraviolet,
他们是否仍将依赖西方技术，比如深紫外光刻技术，
you know, graphics technology and that sort of thing.
你知道的，光刻技术之类的。
Right?
对吧？
So we can't just look at this in a vacuum.
所以我们不能孤立地看待这个问题。
We have to look at the broader strategic supply chain issues as well.
我们还必须关注更广泛的战略性供应链问题。
And that is really gonna be a matter of time before we can see this play out.
而这真的需要时间才能看到结果。
Mhmm.
嗯哼。
And this week, other major Chinese tech giants such as Baidu, SMIC,
本周，其他中国科技巨头如百度、中芯国际，
and Alibaba have also announced new initiatives aimed at achieving AI self sufficiency.
以及阿里巴巴也宣布了旨在实现人工智能自给自足的新举措
So is this, you know, purely coincidental,
那么，这是否纯属巧合，
and what impact could it have on The ongoing US China trade discussions?
它又会对正在进行的中美贸易谈判产生什么影响？
Well,
嗯，
I think this is reflective of the broader trends
我认为这反映了更广泛的趋势
when it comes to very strategic supply chains,
谈到非常具有战略意义的供应链时，
uh, and obviously chips rank right up there at the top.
呃，很明显，芯片就排在最顶端。
And that is that self sufficiency is becoming more and more of a premium
这就是说，自给自足正变得越来越重要
for companies.
对公司而言。
So the really big tech companies are looking to vertically integrate if they can.
所以，真正的大型科技公司都在寻求垂直整合，如果他们能做到的话。
You know, we see more tech companies looking to design their chips.
你知道，我们看到越来越多的科技公司希望自己设计芯片。
But as as chip supply chains become, um, much,
但随着芯片供应链变得……嗯，更加
much more vulnerable to geopolitical pressures,
更容易受到地缘政治压力的影响，
the benefits of having a very,
拥有一个非常……
very resilient supply chain obviously is super important, right,
非常有弹性的供应链显然是超级重要的，对吧？
for any big tech company like a Tencent or a ByteDance or, you know,
对于任何大型科技公司，比如腾讯或字节跳动，或者你知道
or an Alibaba.
或者阿里巴巴。
So I think we're going to see more and more attempts
所以我认为我们会看到越来越多的尝试
by the big megatech companies in The States,
来自美国的大型科技巨头，
in China, looking to bring more and more of the value chain in house.
和中国的，他们希望将越来越多的价值链收归内部。
Mhmm.
嗯。
And also according to that Financial Times report,
此外，根据《金融时报》的那篇报道，
Chinese authorities reportedly urged local tech companies
据报道，中国当局敦促本土科技公司
to reduce reliance on NVIDIA's chips.
减少对英伟达芯片的依赖。
Given this push,
鉴于这一推动，
do you think China's AI capabilities are sufficiently advanced to achieve this goal,
您认为中国的人工智能能力是否足够先进以实现这一目标，
or it's purely, uh, you know, a political game there?
或者这纯粹是，呃，你知道的，一场政治游戏？
Well, look, when we look at AI capabilities, remember,
嗯，看，当我们审视人工智能能力时，要记住，
we have to look at the entire tech stack, Right?
我们必须看整个技术栈，对吧？
To get to that end game of producing superior performing AI,
为了达到生产卓越性能人工智能的最终目标，
you need the software and, you know, you need the semiconductors to do that.
你需要软件，而且，你知道，你需要半导体来实现这一点。
Right?
对吧？
So ultimately and really, this starts with the rare earths
所以归根结底，这始于稀土
and the critical minerals, which,
和关键矿物，而这些，
of course, China has a lot of processing capability.
当然，中国拥有强大的加工能力。
They dominate in that area.
他们在那个领域占据主导地位。
But when you get to the next level of the tech stack, right,
但当你进入技术栈的下一个层面时，对吧，
the semiconductor fabrication process,
半导体制造过程，
whether it's Huawei or, you know, any other company,
无论是华为还是，你知道的，任何其他公司，
if they still depend on critical choke point technology like, uh,
如果他们仍然依赖于关键的瓶颈技术，比如，呃，
deep ultraviolet technologies, right,
深紫外光刻技术，对吧，
which are completely dominated now by American and and Dutch companies in particular, um,
目前这个技术完全由美国和尤其是荷兰的公司主导，嗯，
that still has to be factored into this overall equation.
这仍然需要被纳入整体考量之中。
Mhmm.
嗯。
Well, beyond the race for chip development,
那么，除了芯片开发竞赛之外，
what are the key obstacles China faces now in its quest for AI self sufficiency?
中国在寻求人工智能自给自足的道路上现在面临哪些主要障碍？
Well, look, you know, when we look at, again, the road to to AI, um,
嗯，看，你知道，当我们再次审视通往人工智能的道路时，嗯，
leading edge innovation, there are so many players and it's such a big ecosystem.
前沿创新，这里有太多的参与者，这是一个如此庞大的生态系统。
Ecosystem.
生态系统。
So, uh, you know, this isn't just unique to China.
所以，呃，你知道，这并非中国独有的情况。
But, uh, you know, as the world fragments, right,
但是，呃，你知道，随着世界变得碎片化，对吧，
as global value chains continue to to to ring fence and to regionalize and to localize,
随着全球价值链不断地被圈定、区域化和本地化，
that is, of course, disruptive to, you know, a hyperscaled, you know,
这当然会颠覆，你知道的，一个超大规模的，你知道的，
massive economy of scale kind of global value chain.
具有巨大规模经济的全球价值链。
So for China,
所以对中国来说，
the challenge is going to be to to try and bring as much of
挑战将是尝试将尽可能多的
that value chain within their own sort of indigenous ecosystem,
价值链纳入他们自己的本土生态系统，
which for chips is gonna be very, very difficult even for The United States, which is,
这对芯片来说将非常非常困难，即使对美国也是如此，它，
you know,
你知道，
dominant in the design and in the in the tooling
在设计和工具方面占据主导地位，
and a lot of the fabless aspects of chip of chip production,
以及芯片生产的许多无晶圆厂方面，
even for The United States,
即使对美国来说，
which has close relationships with the other key countries
它与其他关键国家（和地区）有着密切的关系，
like Japan, South Korea, Taiwan,
比如日本、韩国、中国台湾省，
The Netherlands.
荷兰。
Um, even for The US, it's gonna be difficult.
嗯，即使对美国来说，这也很困难。
Right?
对吧？
So a fragmented system,
所以一个碎片化的系统，
a fragmented global chain a global value chain system makes it more challenging for any country
一个碎片化的全球链、一个全球价值链系统，使得任何国家
to try and excel in any any given field.
想要在任何特定领域脱颖而出都变得更具挑战性。
Yes.
是的。
Indeed.
的确如此。
Long way to go amid those uncertainties.
在这些不确定性中，还有很长的路要走。
Thank you so much for your analysis.
非常感谢您的分析。
Alex Capri, senior lecturer with NUS Business School.
Alex Capri，新加坡国立大学商学院高级讲师。
