与Arm首席执行官Rene Haas一起重新定义芯片架构

从来没有哪个计算问题是不需要微处理器或者不能利用微处理器的。

There's no computing problem that's ever been invented that doesn't utilize and can't utilize the microprocessor.

它是一切的核心。

It is the heart of everything.

所有的路径都要经过它，围绕它，或者绕过它。

All roads lead through it, around it, past it.

总得有个东西来协调和决定那些令牌的去向。

Something has to do the orchestration, arbitration decision around where those tokens go.

这就是CPU的职责。

That's what CPUs do.

芯片设计可能需要24到36个月，具体取决于芯片的复杂程度等等。

Chip design can take anywhere from twenty four to thirty six months depending on the complexity of the chip, etcetera, etcetera.

实际设计并不是最耗时的部分。

The actual design is not the largest amount of time.

最耗时的是验证、确认和调试。

The largest amount of time is in the verification, the validation, the debug.

人工智能在这方面真的很擅长。

AI is really good at that.

如果我们把它关掉，就像回到90年代，你有了互联网，然后你说，只有下午两点到四点之间能上网。

And if we were to shut it off, it's like being in the 1990s, you've got internet and you're now saying, you know, only internet between the hours of two and four.

对。

Yeah.

之后，就去走廊那头的图书馆吧。

After that, go to the library that we have down the hall.

那会乱套的。

It'd be anarchy.

是的。

Yeah.

这已经是既成事实，无法阻止了。

The genie's out of the bottle and there's no stopping that.

嗨，听众朋友们。

Hi, listeners.

欢迎回到NoPriars。

Welcome back to NoPriars.

今天，我和Elad请到了Arm和软银集团国际的CEO Rene Haas。

Today, Elad and I are here with Rene Haas, the CEO of Arm and SoftBank Group International.

我们聊了Arm在芯片行业中的地位、芯片创新兴趣的回升、供应链的挑战、机器人技术的未来、能源问题、他在软银集团中的角色，以及他如何看待未来工作负载的变化和对Arm的影响。

We talk about the position of Arm within the chip industry, the resurgence of interest in chip innovation, the challenges of the supply chain, the future of robotics, energy, his place in the SoftBank Group, and how he sees workloads changing in the future and for Arm.

Rene，非常感谢你接受我们的采访。

Rene, thanks so much for doing this with us.

荣幸。

Pleasure.

恭喜你们在Hot Chips上的芯片展示，以及Arm取得的这些进展。

Congratulations on the chip presentation at hot chips and, you know, all of the progress that Arm has made.

我觉得科技行业和软件行业现在对芯片供应链有了极大的兴趣，大家也都想更深入地了解它。

I think there's enormous amount of interest from the technology industry and the software industry and just sort better understanding the chip supply chain recently.

对于不太熟悉的人来说，你能解释一下Arm在其中的位置吗？

For anybody who's not super familiar, can you explain Arm's position in it?

然后我们再聊一些更近期的话题。

And then we'll get into sort of more recent topics.

我们在芯片供应链中有两个角色。

So we have two positions in the chip supply chain.

我们的主要业务是授权IP，也就是CPU核心，这些核心用在智能手机、数据中心、汽车等各个领域。

Our primary business is licensing IP, the CPU core that finds its way into smartphones, data centers, automobiles, you name it.

我们的客户要么自己制造芯片，比如三星有自己的晶圆厂，要么大多数公司拿着芯片设计去找台积电流片。

Our customers are the ones who either build the chips themselves, a Samsung who's got their own fab, or the vast majority companies that take their chip designs and go to TSMC and get them taped out.

在这个领域里，Arm的特别之处在于我们覆盖的市场非常广泛，所以几乎能看到所有情况。

So in that world, and this is the cool thing about Arm, because we're so broad in terms of the markets that we serve, we kind of see everything.

我们对汽车、数据中心、智能手机的现状都有很清楚的了解。

We have a very good sense of what's going on in automotive, data center, smartphones.

所以我们可以从各个角度看到供应链的情况。

So we see the supply chain situation from all angles.

我们还在去年三月推出了第一款产品，就是你刚才在Hotchips提到的Arm AGI CPU。

We also introduced our first product last March, the one you just mentioned at Hotchips, the Arm AGI CPU.

所以现在我们自己也身处其中，也得想办法采购基板、晶圆和内存等等。

So now we're in that soup ourselves from the standpoint of we're also having to figure out how to buy substrates and buy wafers and buy memory, etcetera, etcetera.

所以我们在供应方面也是忙得不可开交。

So we're up to our waste and everything on the supply side.

为什么现在做出这个转变？

Why'd you make the move now?

我觉得Arm已经存在几十年了。

So Arm, I believe, existed for a few decades now.

之前的重点一直是IP，也就是设计芯片各组件的组合方式，然后授权给别人去制造并集成到他们的设计中。

The focus was always on IP, which is effectively designing the way the different chip components are put together, then you license that out to other people to actually manufacture and incorporate into their designs.

为什么你们决定开始自己制造一些CPU呢？

Why did you decide to start making some of your own CPUs?

是的，这是从早期我们只提供IP组件、CPU IP、GPU IP、系统IP等发展而来的。

Yeah, was an evolution from the early days of where we just supplied simply the IP components, the pieces, the CPU IP, the GPU IP, the system IP, etcetera, etcetera.

几年前，我们开始看到产品周期并没有变慢，而芯片制造时间却在延长。

A few years ago, what we were starting to see was that product cycle times aren't slowing down, chip manufacturing times are extending.

更快地推出解决方案变得越来越重要。

The ability to get solutions out faster was becoming more and more important.

所以我们从这些单独的组件转向了我们所说的计算子系统。

So we moved from these individual components into what we called compute subsystems.

几年前我们做路演的时候，我用了乐高的比喻，基本上就是我们提供蓝图，告诉你怎么把所有的东西拼在一起。

I used, when we did the road show a few years ago, I used the Lego analogy where essentially we're providing the blueprint on here's how you stitch it all together.

市场对那个的需求非常疯狂。

Demand for that was insane.

我们发现，最初人们以为，人们不会想要这些子系统，因为那是芯片设计师的工作。

And what we were finding was we were, and we initially, people thought, well, people aren't gonna want these subsystems because that's what a chip designer does.

你们为什么要提供那个部分？

Why are you providing that piece?

但它节省了上市时间。

But it saved time to market.

而且在成本、速度等方面也节省了很多。

And it saved a whole lot of things in terms of cost, speed, etcetera, etcetera.

物理产品可以说是下一个飞跃。

The physical product was sort of the next leap, if you will.

有些客户会授权IP，他们有足够的能力基于Arm制造芯片。

And there are certain sets of customers that will license IP too, and they've got all the capability in the world to build chips based on Arm.

有很多公司想要基于Arm的产品。

There's a lot of companies who want to have product based on Arm.

并不是我们所有的客户都生产服务于这些市场的产品。

Not all of our customers build products that serve those markets.

所以Meta就是第一个例子。

So Meta was that first example.

他们想要一个通用的代理型CPU。

They wanted a general purpose agentic CPU.

当时没有人能提供给他们。

There wasn't anybody out who could give it to them.

他们来找我们说，嘿，我们为什么不一起做呢？

They came to us and said, Hey, why don't we do this together?

所以我们才涉足这个领域。

And that's how we got into it.

这对你们其他客户群体的反响如何？

How has that landed with the rest of your customer base?

我们非常谨慎地确保生态系统能接受这一点，因为我们做的是CPU IP，它的价值完全取决于生态系统的成熟度，包括芯片制造商和软件开发者等围绕它构建的各类伙伴。

So one of the things that we were very careful about was making sure the ecosystem was on board with this, because we do CPU IP, which is really only as good as the ecosystem, the ecosystem of chip people and the ecosystem of software folks and people who build around that.

所以我们几乎跟所有客户都沟通过，问他们对这一发展方向持什么看法。

So we talked to just about everybody who were customers and said, you know, how do you feel about this is the direction we're going?

出乎意料的是，我们收到的反对意见比预想的少得多。

And surprisingly, we got a lot less pushback than thought.

原因在于，市面上可用的软件越多，无论是专有的还是开源的，对整个生态系统和客户本身都有利。

And the reason for that was the more software that's available in the wild, whether it's proprietary and or open source, benefits the broader ecosystem and the customers themselves.

所以不管是英伟达、亚马逊、微软还是谷歌，这些所有基于Arm架构构建服务器芯片的公司，都对此表示支持。

So whether it was Nvidia, Amazon, Microsoft, Google, all people who build Arm based server chips, they were all on board.

我认为最有力的证明是，去年三月我们发布产品时，黄仁勋、罗尼·布克、Arm的代表以及詹姆斯·汉密尔顿等客户代表都出席了，他们都说恭喜，这是一件大好事。

And I think the ultimate proof point was when we announced the product last March, we had Jensen, we had Ronnie Bulker, we had Arm in, we had James Hamilton, all the folks from those customers I mentioned all saying, congratulations, it's a great thing.

所以整体来说还算顺利。

So it's been okay.

作为一家开始销售实体芯片的企业，你们目前处于学习周期的哪个阶段？这感觉需要掌握很多新能力。

Where are you in the learning cycle as a business now selling physical chips that feels like a lot of new capabilities?

对。

Yeah.

显然，为了交付产品，我们作为一家无晶圆厂半导体公司，

So obviously to deliver a product, we're a fabless semi company, right?

我们没有自己的晶圆厂，也打算建，但我们是这个生态的一部分。

We don't have a fab and we have no intention to build a fab, but we fit in that ecosystem.

但这意味着你需要有负责供应链运营的人。

But that means you need supply chain operations people.

正如我刚才说的，你得跟台积电、三星这些大厂打交道。

You need to work with, as I said, the TSMC's and Samsung's of the world.

你还需要找三星、美光、海力士这些公司合作，搞定内存的供应。

You need to work with the Samsung's and the Micron's, the SK hynix to get memory allocation.

然后在工程方面，你需要更多不同的能力。

And then on the engineering side, you need a lot more different capabilities.

你需要后端人员、布局人员、实现人员、调试实验室，还有硬件相关的工作，对吧？

You need backend people, layout people, implementation people, bring up labs, physical stuff, right?

我们以前并没有太多硬件相关的工作，这其实是原来业务的一个优势。

We didn't have a lot of physical stuff, which was kind of the beauty of the business, the original business.

我记得发现Arm的毛利率高达98.5%。

I remember discovering that Arm had a 98.5% gross margin.

对，挺漂亮的。

Yeah, kind of beautiful.

我

I

之前没怎么见过这种情况。

don't think I've seen that otherwise.

来这儿之前，我在英伟达工作。

I came from NVIDIA before I came over here.

我的职业生涯大部分时间都在芯片行业。

Most of my career was in the chip world.

我记得2013年刚来Arm的时候就想，没有库存，没有退货，没有废品，还有什么不喜欢的呢？

And I remember coming to Arm in 2013 and thinking, no inventory, no RMA, no scrap, what's not to like?

所以我们必须得补充很多这些能力。

So we had to add a lot of those capabilities.

我们的领导团队里有很多人就是从那个行业出来的。

We have a lot of people on the leadership team who've come from that world.

我的高管里有来自博通、高通、英伟达的。

I've got execs from Broadcom, Qualcomm, NVIDIA.

我现在在英伟达工作。

I work for NVIDIA.

所以我们的领导层在别的公司都有过这方面的经验。

So we have the leadership that's done this before in other companies.

所以我们也很快就上手了。

So we've been able to build up that muscle pretty quick.

你们是怎么推进AI应用的？

How have you approached AI adoption?

我们之前聊到，Open Editor今天发布了关于Jalapeno和他们设计的芯片的消息。

So we were speaking earlier that there's news from Open Editor today about Jalapeno, and the chip that they designed.

他们声称产品上市时间非常短，部分原因是利用AI工具来加速芯片设计。

Their claim is it was a very fast time to market, and part of that was using AI tooling to sort of design chips faster.

你们在这方面应用得怎么样？

How much adoption have you seen there?

我知道其他公司也提到过，比如在亚马逊或其他地方采用形式验证来训练芯片。

I know other companies have also talked about things like adopting formal verification at Amazon or other places for their training chips.

所以我觉得芯片行业在AI应用方面正在发展，我很好奇你们在Arm是怎么做的。

So I think the chip world is starting to evolve in terms of AI usage, and I'm just curious about how you've done that at Arm.

就我个人而言，我非常相信AI作为一种工具，能提高每个行业的生产力。

Personally, I'm a huge believer in AI as a utility that's going to help productivity for every single industry.

对于能迅速起步的公司来说，这将会是一个巨大的平衡器。

It is gonna be the great leveler in terms of companies that can get started super quickly.

而且对于各行各业，无论是医疗、基础设施、机器人，每个行业都会把人工智能当作一种公用事业来使用，就是这样。

And for industries, whether it's healthcare, infrastructure, robotics, every industry is gonna use artificial intelligence as utility and stop.

所以，既然我对此深信不疑，那我们在Arm内部当然也会大量使用它。

So since I'm such a believer in this, of course we use it very heavily inside of Arm.

在非工程方面，我们到处都在用；但在工程方面，我们看到了巨大的、巨大的收益。

On the non engineering side, we're using it all over the place, but on the engineering side, we've seen huge, huge benefit.

你提到了验证。

You mentioned verification.

芯片设计可能需要二十四到三十六个月的时间，具体取决于芯片的复杂程度等等。

Chip design can take anywhere from twenty four to thirty six months, depending on the complexity of the chip, etcetera, etcetera.

实际的架构设计、RTL生成，也就是架构的映射，并不是最耗时的部分。

The actual design of the architecture, the RTL generation, if you will, the mapping of the architecture is not the largest amount of time.

最耗时的是验证、确认、调试、文档编写等等。

The largest amount of time is in the verification, the validation, the debug, the documentation, etcetera, etcetera.

而人工智能在这方面非常擅长。

AI is really good at that.

而且我想说，现在Arm内部大概有80%到90%的工程师每天都在用。

And I would say we probably have 80 to 90% of engineers today in sidearm who use it on a daily basis.

如果我们要关掉它，我常跟人打比方说，这就像在90年代，你有了互联网，然后现在有人跟你说，只能下午两点到四点之间上网。

And if we were to shut it off, my analogy I give to people, it's like being in the 1990s, you've got internet, and you're now saying, only internet between the hours of two and four.

其他时间，去走廊那头的图书馆，那里有所有书，你可以去查资料。

After that, go to the library that we have down the hall, that's got all the books that you can go up and look all this information up.

那肯定要乱套了。

People, there'd be anarchy.

所以这已经是既成事实了，对吧？

So the genie's out of the bottle, right?

挡也挡不住了。

And there's no stopping that.

不过现在有些方面，工具还不太成熟。

Now there's certain things that the tools are still not that mature of.

其中一个就是RTL生成，还有物理设计和实现要做到最顶尖水平这块。

One of them is really around RTL generation, and then physical design and implementation in best of class.

这主要是因为模型是基于公开可用的数据训练的，而很多信息都是相当专有的。

And that's simply because the models, they're trained on what's available publicly, and a lot of the information is quite proprietary.

话虽如此，生态系统之间以及行业内的每个人之间都有巨大的机会来改善这一点。

That being said, there's massive opportunity between the ecosystems and everyone in the industry to make that better.

只会越来越好。

It's only gonna get better.

你有没有在微调模型来尝试解决这个差距，考虑到你拥有的信息……

Have you been fine tuning models to try and address that gap, given the information that you

……我们确实在和模型制造商合作解决这个问题。

want We've working with model makers around that, absolutely.

我认为这是一个非常大的机会。

And I think that's a big, big opportunity.

在Arm，我引以为傲的一点是，鉴于我们的业务，也就是核心IP业务，我们可能拥有最丰富的IP组合，不仅包括IP本身，还有关键的是文档、测试平台，以及如何构建IP。

And one of the things that I'm proud of at Arm is given our business, our core IP business, we probably have the richest IP portfolio, both in terms of not only the IP, and this is the killer, the documentation, the test benches, how you build the IP.

我以前在芯片公司工作过，他们说过，嘿，我们为什么不把我们的IP授权出去呢？

I've worked for chip companies in the past that have said, Hey, why don't we license this IP that we've got?

因为它的价值真的非常高。

Because it's really, really valuable.

然后你就会发现，等等，居然没有任何文档说明。

And then you get into, Wait a minute, there's no documentation.

也没有任何关于如何使用的解释。

There's no explanation on how

根本没人能用得上这东西。

No one's ever gonna be able to use this.

根本没法用。

It's unusable.

也没法做测试。

It's untestable.

既然没法用也测不了，那实际上就没法用来训练模型。

And if it's unusable and untestable, it's actually untrainable.

没法训练的话，自然也就没法用在人工智能领域了。

And if it's untrainable, it's not usable for AI.

我觉得我们的商业模式有一些天然的优势，能让我们在工具越来越好的时候真正利用它们。

I think we have some built in advantages based on our business model that'll allow us to really be able to take advantage of the tools as they get better.

真让人兴奋。

So exciting.

你觉得，如果推演一下——这个问题有点不确定——

How much do you think, if you were to extrapolate out and this is a little bit of an uncertain question.

但如果推演到两三年后，AI工具都成熟了，还能针对你提到的设计某些方面微调模型，你觉得那个24到36个月的周期会不会缩短到一年甚至六个月？

But if you extrapolate out two, three years, and all the tooling is likely to come in AI and the ability to fine tune models against some aspects of the design that you mentioned, do you think that twenty four to thirty six month cycle shrinks to a year to six months?

你觉得它会保持大致不变吗？

Do think it stays roughly where it's at?

我有点好奇，这到底会怎么影响这些周期和上市时间，因为这对整个行业的节奏会有巨大的影响。

I'm a little bit curious about how does that really impact these cycles and time to market, because that has pre dramatic ramifications in terms of the clock speed of the entire industry.

我不知道是不是两三年内，但五年以上，能不能从想法直接到GDS2文件——就是发给晶圆厂去制造的那个文件——对于某些设计来说？

I don't know if it's in two to three years away, but five plus years, can you go from idea to a GDS two file, with GDS two file being the file that you actually send to the fab to go get built for certain designs?

很有可能。

Quite possible.

这样一来，整个设计环节就省掉了。

So it takes that whole design piece out of the way.

相对于验证来说，整个环节也省掉了。

It takes that whole piece out of the way relative to the verification.

所以我觉得对于比较直接的设计来说，完全有可能。

So I think for the more straightforward designs, quite possible.

现在，如果你进入工具里说，给我设计一个比Verruvan快10%、便宜20%、效率高30%的模型，你不可能按一下按钮就马上实现。

Now, if you go into the tool and say, design me something that's 10% faster than Verruvan, 20% cheaper, and 30% more efficient on this model, you're not gonna be able to get press a button and have it happen right away.

但我认为在五到十年内，我们的行业也会看到芯片设计方式上的巨大变化。

But I think in five to ten years, you know, our industry as well, we're gonna see some amazing differences relative to how chips are designed.

这怎么改变你对业务的看法？我相信你对此有一些预测，但现在有很多大玩家和新玩家都想要有自己的芯片设计，这怎么改变你的看法？

How does it change I'm sure you had some prediction of this, but how does it change the way you look at the business given there's just a big diversity of large players and new players that all, you know, want to have their own chip designs now?

而且，像Vera和Graviton这样的公司，他们都用Arm。

And, you know, the Veras and the Gravitrons of the world, they all use Arm.

这对他们来说是一个很大的进步。

It's a big step up for them.

但客户群确实变得非常多样化，对吧？

But it's a big diversification of the customer base, right?

这只会是好事。

That can be only good.

哦，当然。

Oh, absolutely.

我觉得关键点在于，回到我们之前讨论的供应链问题，就是要理解供应链的影响，以及这一切如何被构建并最终整合到终端产品中。

I think what's going to matter, back to the earlier discussion we had on supply chain, it's understanding the supply chain impacts, how all of that gets built and put into ultimate end products.

我认为这将成为未来越来越重要的能力，因为说起来是一回事，做起来是另一回事。

I think that's gonna become a much more important muscle as we go forward, because it's one thing to say it another way.

现在有很多非常优秀的年轻公司在做AI芯片，知名公司获得大量资金，设计创新，等等等等，但它们所面向的行业资本需求巨大。

There's a lot of really great young companies today doing AI chips, well known companies getting tons of funding, innovative designs, etcetera, etcetera, selling into an industry where the capital requirements are just massive.

与内存供应商的关系极其关键，或者与基板供应商的关系。

And the relationships with memory vendors is incredibly critical, or the relationship with substrate vendors.

所以公司将不得不更加——

So companies are going to have to be much more-

更多机会接触到三纳米生产线、六纳米生产线、先进封装线。

More access to a three nanometer line, a six nanometer line, advanced packaging line.

对，都是这些。

All that.

我觉得这种情况不会在十二个月内结束。

I think that's not going to stop in twelve months.

也不会在二十四个月内结束。

It's not gonna stop in twenty four months.

我觉得我们会处在这种受限的环境中至少三到五年。

I think we're gonna be in this constrained environment for three to five years, at least.

只要Transformer架构还是AI训练和推理的主要方式，它本质上就是计算密集型和内存密集型的。

So long as the transformer is the unit of energy relative to how you generate AI training and AI inference by design, it's very compute intensive, it's very memory intensive.

所以想想看，这会带来对供应链管理能力的大量需求，这又回到了那些在芯片设计上有好点子的公司，他们需要很多其他资源才能获得资金、晶圆，以及你刚才提到的所有东西。

So if you think about that, that's gonna drive a lot of demand on having supply chain acumen, which then goes back to people who've got great ideas on chip design, they're going to have to have need a lot of other things just to be able to get access to capital, wafers, everything you just talked about.

我们刚刚经历了一连串的事情，这些都是AI行业计算能力提升的瓶颈。

We've just had a cascading series of things that have been the bottleneck to more compute for the AI industry.

所以，你知道的，大概两年前吧，我觉得瓶颈是在封装和封装相关的东西上。

So, you know, two years ago or so, I think it was like packaging and packaging related items.

而现在呢，大家开始讨论内存等问题，这些因素在某种程度上限制了某些系统的大规模构建。

And then eventually now, people want to talk about how it's memory and things like that that are in some sense limiting to certain systems being built at sufficient scale.

你觉得下一个即将到来的瓶颈会是什么？

Do you have a view of what is the next sort of bottleneck that's coming?

我觉得数据中心建设会成为瓶颈。

I think building out the data centers is going to be a bottleneck.

说到扩建，如果你看看现在正在进行的所有项目，没几个是提前完工且比预期更省人力的。

And when I say building out, if you look at all the projects that are being done today, not a lot of them are ahead of schedule and needing less labor than they thought.

再加上，美国各地传来不少风声，说要放缓数据中心建设或者加限制。我觉得这些基建方面的阻力，跟目前的大环境结合起来看，可能反而是件好事。因为如果基建不是个阻碍的话，晶圆和内存的产能不足才会成为真正的瓶颈。

And then when you layer on top of that, a lot of buzz that's coming from different parts of the country in The United States here, relative to slowing down data center development or putting restrictions around it, I think that infrastructure build out could be a headwind just relative to everything going on, which may be, end quote, okay, because if infrastructure build out was not a headwind, I think capacity for wafers, capacity for memory, that probably would be a headwind.

所以我觉得你会看到各种各样的因素在制约这个行业的发展。稍微展开一下，我之前参加过好几个论坛，总有人问我AI泡沫什么时候会破。

So I think you're gonna see a number of different governors, if you will, not governors of states, but different things that are going to throttle the growth of this, which just expounding for a second, I had been on a bunch of panels and I of get a questions about AI bubble and when's it going to stop?

咱们先不谈估值泡沫，那是股市指数层面的事。我指的是供需层面的泡沫，也就是我们是不是供过于求了？

And there's setting aside the valuation bubbles, which is a stock market index component, the bubble in terms of, are we oversupply to demand?

差得远呢。

Not even close.

而且我觉得，这主要是因为需求实在太大了，这些模型的运作方式就是这样。再加上基础设施的建设、晶圆的获取、内存的获取，所有这些因素加在一起。

And I think again, that's because the demand is insatiable, just given the way these models work, And infrastructure build out, access to wafers, access to memory, all of that's combining.

你提到了这一点，我觉得现在很多公司都在学习，在这样一个时代，要么需要消耗大量的算力，要么需要投入大量的资本支出，或者是在做像开发CPU IP这样的大型技术项目，战略性地利用股权结构和资本获取渠道非常重要。

You mentioned that, and I think a lot of companies are learning today, that strategic use of the cap table access to capital in an era where you either need to consume a lot of compute or you need to put a lot of CapEx into the ground, or you're just doing big technical projects like coming up with CPU IP.

你经营着软银集团国际。

You run SoftBank Group International.

你有这样一个大股东。

You have this one dominant shareholder.

从外部来看，Arm本身就像一台漂亮的现金流机器，对吧？

Arm itself as a business is just like a beautiful cash flow machine from the outside, right?

我相信你肯定经常思考软银的杠杆作用，以及如何好好利用它。

I'm sure you think a lot about like the leverage of SoftBank and how to use that well.

根据你所说的，对于在这些资本密集型行业里打拼的企业家，你有什么建议？

Like what advice do you have for entrepreneurs navigating these CapEx intensive industries from what you said?

对。

Yeah.

所以我们在Arm有一个好处，虽然上市了，但有一个非常大的单一股东。

So one of the benefits we have at Arm, publicly traded, yes, but a very, very large single shareholder.

所以我经常和我的大股东非正式会面，讨论这些事情。

So I have lots of informal investor meetings with my chief shareholder all the time about this.

我们有一个很大的优势，就是可以和软银进行很多互利合作，帮助Arm推进其计划，有软银作为我们最大的股东，我们期待在这方面进行创新。

We have a big advantage in that there's a lot of things symbiotically we can do together that can help Arm advance its initiatives by having SoftBank as our largest shareholder that we look to be very, very innovative around.

对于年轻公司，我想说，尽早建立战略伙伴关系，无论是与供应链上的人、私募股权公司还是银行本身。

To your point in terms of young companies, I would say strategic partnerships incredibly early, whether it's with people inside the supply chain, people in private equity, the banks themselves.

现在情况不同了。

It's a different game now.

一方面，半导体行业有点复苏了，因为现在出现了一波半导体初创公司。

On one hand, semis are kind of back because you now have a wave of semiconductor startups.

有很长一段时间，这个行业都没有投资。

There was a long time where that was just not happening, investment in the industry.

现在虽然有很多，但对他们来说，能不能获得资本支持才是关键。

Now we've got a lot, but access to capital is going to be the gate for them in terms of how they get through that.

所以我觉得，他们需要更创造性地与整个生态合作，这一点会非常非常关键。

So I think getting much more creative in terms of how they work with the ecosystem is going be super, super key.

而在软银，这正是我们非常看重的一点。

And we at SoftBank, that's one the things we look at very strategically.

我们会把一些公司纳入投资组合，帮助他们，提供资金支持，或者，你知道，如果你想想软银，我们刚宣布了软银Neo，也就是我们想成为一家新云服务商的计划。

Companies that we can bring in into the portfolio that we can help, that we can provide a combination of either the backstop, you know, and or if you think about SoftBank, we just announced we being SoftBank, SoftBank Neo, which is our intent to become a neo cloud.

在这样的背景下，我们可以成为这些年轻公司的家，他们拥有芯片技术，但在其他情况下，他们可能得去微软或谷歌争取设计订单。

And in that world, we could become a home for these young companies who have chip technology that in other worlds they'd have to go up and figure out how to get a design win at Microsoft or Google.

我们可以为他们提供很多有趣的途径。

We can provide a lot of interesting avenues for that.

你能多谈谈你在软银负责的投资组合吗？

Can you talk a little bit more about the portfolio of things that fall under your purview at SoftBank?

我知道你提到了Arm，还有更广泛的业务范围。

I know it mentioned there's Arm, and then there's this sort of broader suite of things.

那我很想多听听你还负责哪些方面，而且我们也有一些具体的问题想问问这些方面。

So love hear more about what else you're responsible for, and we had some specific questions for some of those as well.

是的，你可以这样理解：软银集团总部在日本，由孙正义领导，旗下有很多不同的运营公司。

Yeah, so the way to think about it is SoftBank Group, which is headed in Japan, that is Masa, have a lot of different operating companies underneath them.

其中最大的一家是软银KK，基本上就是软银移动。

One of the largest ones is SoftBank KK, which is essentially SoftBank Mobile.

在美国这边，软银集团国际有很多投资活动在进行。

Inside The US, there's a lot of investment activity that's going on with SoftBank Group International.

还有软银愿景基金。

There's SoftBank Vision Fund.

但越来越多的是，我们在软银这边尝试做的很多策略，都是在帮助孙正义在他日本的股东大会上公开谈到的那些战略，就是围绕机器人、OpenAI、基础设施和Arm。

But increasingly a lot of the strategies that we're trying to do around SoftBank is helping the strategies that Masa talked about publicly at his shareholder meeting in Japan, which is around robotics, OpenAI, infrastructure, and Arm.

所以老实说，我可能关注的事情很多，主要是帮助孙正义真正实现那个愿景的执行。

So I've probably got my eyeballs on a lot of stuff, to be honest with you, in terms of helping Masa really realize the execution of that vision.

所以是的，我在领导Ampere、Graphcore，还有另一家叫Stack AV的公司，这家公司是做自动驾驶相关业务的。

So yes, I'm leading the direction of Ampere and Graphcore and another company called Stack AV that's doing things around autonomous.

但也许更好的理解方式，Elad，是我在很多Moss参与的讨论中都在场，帮他制定策略，更重要的是帮他执行。

But maybe a better way to think about it, Elad, is that I'm kind of in the room on a lot of discussions that Moss is having and helping him sort of formulate that strategy and more importantly help execute it.

作为软银集团的一员，或者与所有这些不同的公司合作，甚至包括SP Energy和更广泛的生态系统，这如何改变了你对Arm能做什么的看法？

How has being part of the SoftBank Group or working with all these different companies, even SP Energy and the broader ecosystem changed your point of view on what you can do with Arm?

嗯，这给我们提供了一个巨大的全局视角，相对于整个行业的发展方向，无论是基础设施、资本还是能源方面。

Well, one thing it does, it gives us a huge bird's eye view relative to where the broader industry is going, whether it's around infrastructure, whether it's around capital, whether it's around energy.

但你也想象得到，它可能为我们的产品提供一个家，对吧？

But also you can imagine it could provide a home for our products, right?

所以它不一定是唯一的家，但绝对可以是一个家，这也是很大的帮助。

So it doesn't need to be the home, but it certainly can be a home, which is also a big help.

当我们考虑软银涉及的垂直领域，比如机器人、能源、数据中心基础设施，再看看Arm的产品，我们目前只公布了Arm AGI CPU。

When we think about the verticals that SoftBank's involved with robotics, energy, data center infrastructure, and then you look at the products that Arm has, the only one we've announced so far is the Arm AGI CPU.

你可以开始把这些点连起来，说，天哪，可能有一些非常有趣的机会，对Arm来说是个机会，但这不一定意味着我们要进入广泛的商业芯片业务。

You can start to connect the dots and say, gosh, there could be some very interesting opportunities that could be an opportunity for Arm, which necessarily doesn't mean that we're getting into the broad merchant ship business.

我们可能只是为软银做产品。

We could be just doing products simply back for SoftBank.

我们在更通用的机器人领域已经认真投入了几年了，对吧，跟大语言模型差不多十年的发展相比。

We're a couple years into serious efforts in more generalized robotics at this point, right, if you compare it to like about a decade for LLMs.

越来越多的公司展示了令人感兴趣的演示结果，在任务和环境的泛化、更强的鲁棒性，甚至上下文学习方面，但大规模部署还没到时候。

There's increasingly interesting demo results from companies on generalization of task and environment, more robustness, maybe even in context learning, but not like wide scale deployment quite yet.

对。

Yep.

首先，你同意这种描述吗？

First, would you agree with that characterization?

完全同意。

100%.

你对机器人市场有什么预测？Arm在那里有什么机会？

What predictions do you have about the robotics market and any opportunity for Arm there?

哦，总的来说，我认为无论是人形机器人还是可以重新训练来完成特定任务的专用机器，市场都会非常巨大，对吧？

Oh, well, broadly speaking, I think whether it's humanoids or dedicated machines to do certain level of tasks that can be retrained is gonna be enormous, right?

机器人1.0，也就是为特定目的而建的行业。

The robotics 1.0, which is a purpose built industry.

以前，机器人是为特定任务设计的机械，软件也是针对那个任务优化的。

You had piece of mechanics designed to do a certain task, and the software that was optimized for that task.

如果一条全新的汽车生产线出现，或者换了不同的设备，而机器人不适合的话，就得把整条线拆掉重来，诸如此类。

If you had to, a brand new automobile line came up, or some different piece of equipment, if the robots weren't well suited for that, rip up the line, etcetera, etcetera.

所以你可以想象，那时候的门槛相当高。

So as you can imagine then the barrier was pretty high.

现在要走向一个机器人能通过学习或观察来学习的世界。

Getting to a world where the robots can learn just based upon either being trained or what they see.

然后再加上，你能不能设计出机械上足够通用的东西，能利用重新编程的优势？

And then when you then combine that with, can you design something mechanically general purpose enough that can take advantage of being reprogrammed?

再加上成本在下降，你一看就会想，天哪，还有什么它做不到的？

And then when you layer on top of that, the costs coming down, you look at it say, oh my gosh, what will it not be able to do?

这简直就像《杰森一家》里的场景，对吧？

So it's almost like something out of the Jetsons, right?

很多事最终都会由机器人来做，比如建筑、基础设施、服务、安保。

Where a lot of things will ultimately be done by robots, construction, infrastructure, service, security.

现在你在Instagram或者TikTok上能看到很多机器人在参加奥运会比赛之类的视频。

Right now you see a lot of stuff on Instagram or TikTok of Olympic races with robots, etcetera, etcetera.

我觉得没人会对机器人体育联赛感兴趣。

I don't think anyone's gonna have any interest in watching a sports league of robots.

可能只有一些发烧友会关注这个，但更广泛的用途还是在那些最终可以被机器人轻松替代的人类劳动岗位上。

There may be an enthusiast class who might be interested in that, but the broader utility is going to be around a lot of human labor tasks that can ultimately easily be replaced by robots.

这一点毫无疑问。

There's no questions right A

大家关于机器人外形设计的各种设想，大致可以分成两到三类。

lot of the hypotheses people have about the form factor of robotics tends to split into two or three camps.

其中一类观点认为，机器人应该是人形或者大致符合人类的活动范围，因为现实世界中的很多设施和设备都已经按照这种标准设计好了，工具也是配套的。

One of the camps is that they're gonna be humanoid or roughly sort of the human footprint because so much of the physical world is already designed that way, and the tooling is designed that way.

这样一来，直接把机器人部署进去就行了。

And so you can just slot robots right in.

而另一些人则认为，未来会出现更多针对特定任务专门设计的外形。

Others view it as there's gonna be much more sort of specialized task specific form factors.

你有什么假设吗？

Do you have a hypothesis on

我觉得

I think

两者都有。

it's both.

对，我觉得两者都有。

Yeah, I think it's both.

有很多工作和任务都是围绕一个人六英尺高、手臂有一定长度等来优化的。

There are a lot of jobs and work tasks that are optimized around a person being six feet tall and having arms of a certain length, etcetera, etcetera.

但我认为两者都会有。

But I think it'll be both.

而且我觉得它们会变得很聪明，能学习。

And I think the fact that they're going to be smart and can learn.

回答你之前的问题，Arm 会无处不在。

And to answer your earlier question, Arm is going be everywhere.

从实时传感的角度来看，我们拥有大量的微处理器技术，这些技术将应用于手指，实现感知和传感功能。

We have a tremendous amount of technology from a real time sensing standpoint around microprocessors that will be out at the fingers that can do perception and sensing.

这些都将基于Arm架构。

That's all going to be Arm based.

如今，无论是NVIDIA还是高通的一些工作，你在人形机器人中看到的大多数“大脑”，都是基于Arm架构运行的。

Today, whether it's NVIDIA or some of the work that Qualcomm does, most of the brains, the brains that you see in the humanoids, those are all running on Arm today.

所以我认为，对我们来说，未来机器人行业将由Arm驱动。

So I think for us going forward, the robotic industry will be powered by Arm.

你看到什么早期迹象了吗？

Are you seeing any early indications?

我的意思是，考虑到Arm的普及以及许多不同类型的设备，你有一个很好的视角，可以比别人更早看到技术采用或转变发生在哪里。

I mean, have this great seat to your point where given the ubiquity of Arm and lot of these different types of devices, you can see the future before others in terms of where adoption is happening or where shifts are happening from a technology perspective.

有没有哪些特定的领域，你认为最有可能成为机器人的早期采用者，而你已经开始看到一些信号了？

Are there specific pockets that you think will be most likely the early adopters of robotics that you're starting to see some signal from?

我觉得现在还有点早，因为商业模式还没有真正确定下来。

I think it's still a little bit early because the business models have not been actually figured out.

机器人的成本太高了，对吧？

Cost of robots are so high, right?

因为机器人太贵了，直接买机器人回去用，这个模式确实挺难让人接受的，尤其是那些真正能替代人力的场景。

Because the cost of robots are so high, people buying the robots themselves, a tough model to sort of get people's heads around those that actually replace.

所以我觉得成本得降下来，商业模式也得经过实际验证才行。

So I think costs need to come down, and the business model need ultimately to vetted.

因为目前机器人的应用场景大多集中在汽车制造、某些手术机器人或者配送中心这些地方，对吧？

Because right now robotic footprints tend to be things like automotive, or certain surgical robots, or distribution centers, right?

目前有一些非常定制化的应用场景，我觉得这些主要是机器人销售方面的业务。

There's a few very sort of bespoke applications that I think are mostly robotic sales today.

所以我对这一点还挺好奇的。

And so that's why was a bit curious about

分销中心肯定算一个。

Distribution what centers for sure.

我的意思是，从配送环节来看，最终完全可以实现自动化，对吧？甚至一直延伸到最后的交付环节，对吧？

I mean, that can ultimately go completely automated, right, relative to And even ultimately to the delivery, right?

而且你可以想想，粗略来说，一辆有自动驾驶功能的卡车，某种程度上也算是一种机器人。

And you can question, to me loosely speaking, a truck that has an autonomous, is a robot of sorts.

所以在工厂自动化、配送和物流这些方面，那将会是最先被自动化的领域之一。

So around factory automation and delivery and distribution, that will be one of the very first to be automated.

毫无疑问。

No doubt.

现在关于供应链管控、使用限制，以及机器人和芯片、数据中心相关的政策或行政令，讨论越来越多了。

There is increasing debate and very quickly, like policy or EOs around supply chain controls and usage controls around both robotics and chips and data centers.

对吧？

Right?

抱歉。

Sorry.

我还得把出口管制也算进去。

I'm gonna throw export controls in there.

所以是四类管控。

So four types of controls.

这些管控措施都跟你有关。

All of these controls are relevant for you.

不管是从你终端客户的角度，还是作为一个相对较新的入局者——毕竟你们要自己掌控最终产品，并建立自己的供应链体系。

And now either from your end customer perspective or as a relatively new entrant to, you know, we're gonna own the end product and have a supply chain organization of your own.

比如，你对保护主义持什么态度？我知道你们不是美国公司，但在这里业务量很大。

Like, what's your stance on, you know, how protectionists I realize it's not an American company, but you do a lot of business here.

那么，在芯片制造、数据中心建设以及机器人领域，美国和西方应该采取多大程度的保护主义立场呢？

How protectionists The US or the West should be about manufacturing of chips, creation of data centers, robotics?

比如，你在这方面的整体立场是什么？

Like, are your overall stances here?

那么，我先以美国公民的身份来说几句。正如你提到的，Arm并不是一家美国公司。

So putting my American citizen hat on for a moment, and Arm, as you said, is not an American company.

它是英国的，

It is a UK,

没错。

correct.

我们的总部在英国，但我们在全球有很多员工。

Our HQ is in The UK, but we have a lot of employees.

我不敢说一半，但大概有30%的员工在美国，40%在英国，剩下30%在亚洲。

I wouldn't say half our employees, maybe 30% are in The US, think I 40% are in The UK and maybe 30% Asia.

所以虽然我们是一家全球性公司，但美国业务占了很大比重。

So we're a global company, but with a huge US footprint.

但作为一个美国公民，一个在半导体行业长大的人，我记得80年代美国在半导体领域是领先的，而日本公司那时候开始变得非常强势。

But as an American citizen and someone who grew up in semiconductors, and I remember in the 1980s when The US was the leader in semis and Japan Inc.

他们在内存价格上非常激进，基本上抢走了很多市场份额。

Started to really get very, very aggressive in terms of memory pricing and essentially taking a lot of market share.

美国当时启动了一个叫Cematec的项目，目的就是重振美国半导体产业，我当时觉得这是正确的做法，而且当时围绕这个项目也有很多热情。

The US started something called Cematec back in the day, which is really around how to refortify the American semiconductor industry, which I thought at the time was the right move, and there was a lot of energy around that.

后来互联网兴起了，SaaS公司成了热门。

Internet hit, SaaS companies were all the rage.

大家就渐渐忘了半导体其实是一个具有战略意义的资产。

People kind of forgot about semi as being a strategically important asset.

但我认为，美国在本土拥有尽可能多的这类技术是至关重要的。

But I think it is critically important for The United States to have as much of that technology inside on US soil.

对英国我也会说同样的话，只是由于英国的规模，程度稍轻一些。

I And would say the same thing to UK, lesser just because of the scale of The UK.

但当你想到美国市场的规模，以及半导体对美国的重要性，无论是英特尔还是美光，我认为我们需要更多的美国晶圆厂。

But when you think about the size of The US market, the criticality of semiconductors to what The US does, whether it's Intel, whether it's Micron, I think we need more US fabs.

这对国家安全至关重要。

It's critical for national security.

这对供应链的多元化也至关重要。

It's also critical for diversification of supply chain.

所以我非常相信这是一个正确的战略。

So I'm a big believer in terms of that as a strategy.

我认为这真的非常非常关键。

I think it's really, really critical.

至于出口管制，我们将限制芯片出口，因为我们不想让中国赢得这场竞赛。

As far as the export controls go, and we're going to limit the chips because we don't want China to win the race.

我个人认为这是一场没有终点的比赛，不会有赢家。

Personal view is that it's an infinite game, I believe, first in terms of the race, that there's not going be a winner.

比赛不会结束。

The race is going to be over.

但可能会出现很多关键技术都不在美国的情况。

But you could get to a situation where a lot of the critical technologies are not US based.

那可不是什么好事，对吧？

And that's not going be a good thing, right?

因为有人说，成本会下降，商品会更便宜。

Because people say, well, the costs will go down and goods are cheaper.

但归根结底，我非常相信这一点，第一，出于国家安全和经济原因，你希望站在技术前沿，因为它推动创新，也推动生态系统的发展。

But ultimately, and I'm a big believer of this, number one, for both national security reasons and economic, you want to be at the forefront of technology because it drives innovation, but it also drives ecosystems.

想想20世纪50年代二战后的美国汽车工业，底特律是世界的中心，威斯康星、俄亥俄、伊利诺伊各地都有相关企业，比如凡士通、普利司通，虽然普利司通是日本的，但还有其他公司融入这个生态系统。

If you think about The US auto industry in the 1950s, post World War II, where Detroit was the center of the universe, you had spots across Wisconsin, Ohio, Illinois, whether it was Firestone or Bridgestone, or Bridgestone's Japanese, but other companies in that ecosystem that fit into it.

数据中心也是类似的道理。

Data centers are kind of the same way.

人们看着数据中心说，哦，那不过是个大仓库，停车场里就停着两辆车。

People look at data centers and say, Oh, it's a big Costco box, there's two cars in the parking lot.

而且所有东西都是自动化的，所以不产生就业。

And all of that is being driven automatically, so there's no jobs.

我完全不同意，因为你想，不管是能源、液冷，还是所有让数据中心变得更好的技术，这些都能在这里创造就业机会。

I call BS on that because if you think about whether it's around energy, liquid cooling, all of the things that make the data center better, those are all jobs that can be created and done here.

所以我觉得作为国家政策，投资美国本土，确保我们保持领先，这太重要了。

So I think as a national policy, it's incredibly important for us to be investing, A, in The United States, and B, making sure that we stay in the lead.

特别是在数据中心这块，那些试图阻止未来数据中心建设的行动，看起来协调得有点过头了。

On the data center side in particular, it seems like a lot of the actions that are being taken to try and prevent future data centers feel more coordinated than not.

我知道这被说成是草根运动，但感觉背后有组织在推动。

I know it's phrased as grassroots efforts, but it seems like there's some coordinated function there.

你觉得为什么某些地方会突然对数据中心有这种意想不到的反对声音？

Do you have a hypothesis in terms of why there's been this sudden unexpected outcry on data centers from certain corners?

我觉得可能是我们之前聊过的……

I think there is a We were maybe chatting about this a

之前提到过，人们担心AI意味着失业。

bit earlier, that there's a fear that AI means job loss.

而失业又意味着所有这些方面的影响。

And job loss means for all these things kind of implications.

所以我觉得不幸的是——而且我觉得

So I think unfortunately- And think

这种担忧有根据吗？

that fear is well grounded?

因为有时候看起来AI只是在创造就业机会。

Because sometimes it seems like it's only creating jobs.

不，我觉得这完全没有根据。

No, I don't think it's well grounded at all.

我记得电工工会特别说过，请不要禁止数据中心。

I think the Electricians Labor Union specifically said, please don't ban the data centers.

他们非常需要这些工作机会，就在最近。

We need these jobs very recently.

完全正确。

Completely.

而且这些工作能让人有成就感，这就是个很好的例子，对吧？

And these are jobs that make people That's a great example, right?

因为以前人们可能对电工这行有点偏见，对吧？

Because here's one where there may have been a stigma to being an electrician, right?

电工要么被看作是高学历的工作，要么被认为不需要博士学位也能干。

Electrician is either maybe viewed as a highly educated job or you don't need a PhD.

这是一份高技能的工作，需要大量的培训和认证。

It's a highly skilled job that requires a lot of training and certification.

而且做这类工作需要大量这样的人手。

And you need tons of them to do this kind of work.

这对数据中心来说非常、非常关键。

And that's very, very critical to data centers.

所以针对你的问题，我觉得部分反弹情绪其实源于恐惧。

So I think to your question, I think part of the backlash is just from fear.