20VC：Cerebras首席执行官谈数据中心的未来、令牌成本与内存｜我们并未陷入基础设施泡沫，Dario与埃隆在算力上达成了不利协议｜美国公司是否应向中国出售技术？为何大多数裁员都是“AI洗牌”——与安德鲁·费尔德曼对话

咱们换个角度想，这儿其实是个负面市场。

Let's try turn it around and say, there's a negative market here.

如果我每个月给你一千美元，让你家里用慢速网络，你也不会要。

If I gave you a thousand dollars a month to have slow Internet in your home, you wouldn't take it.

每个月一千美元。

A thousand dollars a month.

这说明，想用一项重要的技术慢慢来，根本行不通。

That's how impossible it is to engage with an important technology slowly.

那我们为什么觉得推理会不一样呢？

Why do we believe that inference will be any different?

我其实一直在推动他们。

I am kinda pushing them.

当你能用编码器大幅提升速度时，作为核心开发者，你根本不会说‘算了’。

When you power codecs and you're able to be so much faster, if you're core coder, you're not like, ah, bugger.

他们确实如此。

They are.

我觉得你必须得这样。

I think you have to be.

你必须得这样。

You have to be.

你能把产品卖给他们吗

Are you able to sell to

也包括他们吗？

them also?

再提醒我一次，别打断我。

Again, please tell me to sort off.

哦，不用。

Oh, no.

不用。

No.

不用。

No.

你瞧。

Look.

现在，我们正在处理硅谷历史上最大一笔交易之一。

Right now, we are digesting one of the largest deals in the history of Silicon Valley.

对吧？

Right?

你这话说得，哈里，真是让人无语。

You're like, for fuck's sake, Harry.

别开玩笑了。

Give me a break.

我才刚签了一笔两百亿美金的交易。

I've just signed a $20,000,000,000 deal.

你还想要更多？

You want more?

你知道的，

You know,

我们一路上的时候，有些投资者会问。

while we were on the road, some investors would ask.

他们会说，哎呀，你们的业务太集中了。

They'd say, oh, you're heavily concentrated.

你们的生意一大半都押在OpenAI身上。

You have a big portion of your business with OpenAI.

我说，一年前我跟你们谈过，那时候我跟G42签了个十亿美元的单子。

We say, I talked to you a year ago when I had a billion dollar deal with g forty two.

你们那时候也说，你们的业务太集中了。

And you said, you're heavily concentrated.

现在我回来，告诉你们我签了个两百多亿美元的单子，你们还是那句话。

You have a billion dollar deal with g I I come back to you in a year with a 20 plus billion dollar deal, and you tell me the same thing.

但这次客户不一样了。

But with a different customer as well.

客户确实换了。

With a different customer.

而且是不同的客户。

With a different customer.

我跟所有人都说，要服务好众多大客户，关键先拿下一个。

And I I tell everybody that the the way you get good and the way you have succeed with many customers of size is first you win one.

搞定大客户的第一步，是先拿下一个，边做边学，锻炼能力，调整供应链，摸清怎么跟大客户合作。

The way to catch big customers is first catch one and learn, build the muscle, change your supply chain, learn how to work with a large customer.

等你有了经验，下一个客户来找你的时候，才有机会赢下来。

Then you're in a position when the next one comes to have a chance to win.

而且，赢了之后，还能确保客户满意。

And what's more, chance to keep them happy once you won.

等你练出了这套本事，就能再去拿下下一个客户。

And then once you have that muscle, you're in a position to go out and win the next one.

这确实是个巨大的突破。

It is a huge deal.

实现这个目标最大的挑战是什么？

What are the biggest challenges in fulfilling it?

恕我直言，你

With the greatest of respect, do

晚上睡觉的时候会不会想，天啊，这也太多了吧。

you go to sleep at night going, oh, that is quite a lot.

说实话，我觉得发生的事是这样的，萨姆也说过这一点。

Look, I I think what has happened and Sam said this.

他说，人们第一次用GPT的时候，我记得好像是四点多。

He said the first time people used GPT, I think it was what, four something.

他们说，这太神奇了。

They said, oh, this is amazing.

可第二天就问了：怎么还没更快一点？

And the next day they're like, how come it's not faster?

在我们这个行业，人对新东西的适应速度，以及随之而来的更高要求，真是惊人。

The rate at which you get accustomed to something and then want better is amazing in our industry.

以前二十兆瓦就算很多了，后来一百兆瓦算多，再后来一吉瓦才算多。

And it used to be the case that 20 megawatts was a lot, and then a 100 megawatts was a lot, and then a gigawatt was a lot.

现在我们到处找几吉瓦级别的设施。

And now we're running around looking for multi gigawatt facilities.

搁以前，七百五十兆瓦简直让人难以想象。

And that's in any other time, 750 megawatts would have been a mind boggling amount.

可现在我们都觉得这很正常。

And now we're like, yeah.

这点规模我们拿得下。

We got that.

对吧？

Right?

对吧？

Right?

过去一两年里，我们整个行业的心态变化真是非同寻常。

It is the change in mentality over the last year or two for everybody in our industry has been sort of extraordinary.

五年前要是有人说我们要建几吉瓦的项目，比如Crusoe在做的，或者SoftBank Power那些公司干的事，大家都会觉得这人疯了。

Five years ago, if you'd have said, we're engaged in a multi gigawatt build out, or you think about what Crusoe is doing or think about some of the other cool companies, what SoftBank Power is doing, what some of these groups are doing, and you say, that'd be delusional five years ago.

现在呢，又来一个？

And right now it's like, oh, another one?

这倒说得通。

That makes sense.

咱们应该想办法让她在阿联酋部署五吉瓦的项目。

I mean, we should try and get her UAE stargated five gig.

没错。

Oh, Yeah.

没问题。

No problem.

这听起来挺合理的。

That seems reasonable.

我觉得这就是实际情况。

I think that's what's happened.

这是一种非常惊人的思维转变。

It's this extraordinary sort of change in thinking.

如果我们今天对吉瓦级别的大规模建设都无动于衷，五年后我们会变成什么样？

If we are nonchalant to multi gigawatt build outs today, what are we in five years time?

这真很难想象。

It's difficult to imagine.

顺便说一句，我觉得萨姆最厉害的地方就在这儿，可能埃隆也一样——当别人都想不下去的时候，他们还能继续想。

And by the way, that is exactly where I think Sam is the best in the world, maybe Elon, is where everybody else's brain shuts down.

对吧？

Right?

当你在思考一百吉瓦、五百吉瓦的时候，这些人就是有种能力，不会被世界过去的样子束缚住。

When you're trying to think about a 100 gigawatts or 500 gigawatts, those guys, they have sort of this ability to not be constrained by the way the world has always been.

这种能力实在太了不起了。

And that is such an extraordinary power.

当你面对吉瓦级别的规模，比如五百吉瓦，能源会不会就成了决定成败的核心瓶颈？

When you have such scale as the multi gigawatt, you mentioned that the 500 gigawatt, does energy not just become the core crux and bottleneck that enables winners and losers?

我确实觉得萨姆、埃隆这些人说过，他们就是这么想的。

I certainly think that people like Sam and Elon and others have said that's what they believe.

归根结底，我们做的其实是把电能转化成智能。

That at the end, we're in the business of turning electricity into intelligence.

所以，真正的瓶颈在于电力。

Therefore, the limiting factor is electricity.

我不确定是不是完全同意，但以我们现在的视角来看，这个说法确实很有道理。

I don't know if I agree with that, but that is certainly a very reasonable view from from where we are.

那反对这种观点的最坏情况是什么？

What's the bad case against that?

有没有别的说法？

What's the alternative argument?

不一定是你自己的看法，但

It doesn't have to be yours, but

这种情况确实会发生。

it happens.

不会。

No.

不会。

No.

而是你碰上了别的障碍。

That that you bump into something else.

事实上，我们的模型不可能一直变得更聪明。

That what happens in fact is our models can't keep getting smarter.

你撞上了一个隐含的假设：模型会不断变聪明，而你会持续给它们提供更多能源。

That you hit something that has an assumption built in that the models keep getting smarter and you keep feeding them more energy.

最终，模型要么已经足够聪明，要么还在持续进步，值得继续投入能源。

And at the end of the day, the models are either smart enough or keep getting smarter so that it makes sense to keep feeding them energy.

这可能是对的。

That might be true.

我不确定。

I don't know.

你觉得我们能按需求建成足够的数据中心吗？现在即使已经获批，仍有百分之四十的数据中心因为地方市政的审批延误而无法推进，你认为我们真能解决这个问题吗？

Do you think we'll be able to build out data centers in the way that we need to and build out capacity in the way that we need to when we see that 40 out of a 100 data centers are now not being built out even post approval because of local municipalities permitting, like, disruption.

人工智能并不受欢迎。

AI is not popular.

哈里，我觉得这太滑稽了。

Harry, I think it's hilarious.

人们都说，天哪，数据中心又延迟了。

People say, oh my god, the data centers are late.

天哪，各种拖延都出来了。

Oh my god, there are delays.

你家厨房装好了吗？

Have you built a kitchen?

你的承包商也迟到了吧。

Your contractor was late.

对吧？

Right?

挑个家里最小的项目试试。

Pick a little tiny project in your home.

是按时按预算建好的吗？

Was it built on time and on budget?

不是。

No.

想象一下，你要建一个相当于五十个足球场那么大的项目，还得跟当地政府、电力公司还有受监管的行业打交道。

Now imagine building something the size of 50 football fields and requiring interaction with local municipalities and and power companies and regulated industries.

这类项目历史上就没几个能按时完成的。

These things aren't gonna be delivered on time historically.

人们一听到就震惊了，却从没想过自己家装修时的那些经历。

People's mind explodes, and they never think about their own experience in their own homes.

你的承包商每天都准时来吗？

Does your contractor show up every day?

不。

No.

他真能说到做到吗？

Does he do exactly what he says he's gonna do?

很少。

Rarely.

装修用的材料，比如瓷砖之类的，会不会有时候延误？

Do the materials, the tiles, or whatever you for your home, do they sometimes delayed?

会。

Yes.

建数据中心时，也会遇到同样的问题。

All that same thing happens when you build a data center.

发电机有时候会迟到。

The generators, sometimes they're late.

有时候甚至会在运输途中从卡车上掉下来。

Sometimes they fall off a truck, literally.

掉下来之后就会造成损坏。

They fall off a truck and damage is done to them.

变压器也会延迟到货吗？

Are the transformers late?

有时候变压器也延迟，我的意思是，大家一下子都慌了，说一切都晚了，还得应付当地的各种问题。

Sometimes the transformers I mean, everybody suddenly sort of throws their arms up and says, oh, everything's late, or they have to deal with localities.

任何搞过大项目的人，都知道这再正常不过了。

Anybody who's built anything big knows this is par for the course.

这本来就是搞建设的常态。

This is what building means.

所以你并不担心，当地社区会把数据中心看成是城市强加给他们的负担吗？

So you're not concerned then about bluntly local neighborhoods seeing data centers as a city think of

我们这个行业在与社区沟通方面，确实做得太差了。

our industry did a shitty job of engaging the community properly.

我觉得布拉德·史密斯之前发过一篇帖子，从一开始我们就该照着那样做。

And I think Brad Smith put out a post a while ago that should have been the way we all work from the get go.

数据中心可以是干净的，能创造就业，也能给社区带来好处，只要我们用心去做。

And it was these can be clean, they can make jobs, they can be good for communities, we can do this thoughtfully.

它们能带来成千上万个本地工作岗位。

These create thousands of local jobs.

成千上万的本地工作机会，意味着餐馆、午餐生意和酒店都会兴旺起来。

And thousands of local jobs means restaurants and lunches and hotels.

以前的做法呢，我不确定用‘偷偷摸摸’这个词准不准确，但确实是遮遮掩掩、不够透明，也没当好邻居。

And the way they were done was I don't know if sneaky is the right word, but sort of shielded and and wasn't open, and they weren't good neighbors.

现在我们完全没有理由不能做个好邻居。

Now there is no reason why we can't be good neighbors.

我们完全可以在融入社区的同时，让社区从中受益。

There's no reason why the we we can't add these to to communities and have the community benefit from it.

我们得好好想一想。

And we have to do some thinking.

对吧？

Right?

那些大型设备都已经在那儿了。

We have all the the heavy equipment out there.

干脆给当地学校建个足球场吧。

Build a football field for the for the local school.

建一所学校。

Build a school.

在教堂或犹太会堂为社区提供服务。

At a church or synagogue to the community.

对吧？

Right?

我们完全可以以极低的成本成为好邻居。

We can be good neighbors at very, very low cost.

我们完全可以自给自足。

We can pay our own way.

别试图利用电力公司过去三十年分摊输电线路成本的漏洞，把负担转嫁给社区。

Don't try and use sort of loopholes in the way power companies have historically amortized the cost of new power lines over thirty years and push that on the community.

这完全是胡说八道。

That's BS.

我们理应自己承担费用。

We ought to pay our own way.

我们该好好照顾周围的邻居。

We ought to look after our neighbors.

如果我们这么做了，那些接纳这种理念的社区一定会受益匪浅。

When we do that, I think that the neighborhoods that that embrace this will benefit enormously.

但我们没做好。

But we didn't do a great job.

我的意思是，整个行业都没做好。

I mean, we didn't do a great job as an industry at all.

这感觉就像哥伦比亚的贩毒集团那样。

It feels like kind of the cartels in Colombia, were.

他们建教堂、盖学校，生意做得好，利润又高，所以能靠这些来脱身。

It's like they built the churches and they built the schools and they had such good businesses and such high margins that they could kind of get away with it because of that.

不过话说回来，这也不错。

And hey, great.

我不觉得这样是对的。

I don't think that's right.

我觉得这些地方其实有个没被利用的资源。

I I I think these localities have have a resource that that isn't being used.

它们有电力资源。

They have power.

很多土地很便宜，因为没人想要。

Many of these land is cheap because nobody wants it.

我们不会去纽约都会区那些热门地段。

We are not going to parts of Metro New York.

人人都想抢那块地。

Everybody wants that chunk of land.

我们要去的是地价低迷、但靠近电力设施的地方。

You're going to places where the the price of land is depressed, you are near power resources.

我的观点是，我们应当做个好邻居，也要保持透明。

And my position is that we ought to be good neighbors and we ought to be transparent.

我们应当自己承担成本。

We ought to pay our own way.

在我看来，这一点其实没什么好争议的。

Now this seems not to be very controversial in my mind.

我觉得最清晰简洁的描述，就是微软提出的那个方案。

I think the best and most concise description is is what Microsoft has put forward.

我们从一开始就应该这么做。

And we should have been doing that from the get go.

大多数社区都乐于见到邻居自己承担所有费用。

Most communities are comfortable when their neighbors pay all their own way.

只有当一些团体试图转嫁成本，或者不支付他们所用资源的全部费用时，才会出问题。

And it's only when groups tried to shift costs or not pay for the full resources they're using.

我们的数据中心根本不需要消耗大量水资源。

Our data centers don't need to use a ton of water.

它们可以循环利用水。

They can recycle it.

完全可以实现闭环运行。

You can have a closed loop.

我们能自己承担费用。

We can pay our own way.

我们可以升级变电站。

We can upgrade substations.

我们可以升级电网，并全额承担费用。

We can upgrade grids and pay for it in its entirety.

我们不该把这笔负担转嫁给当地社区。

We shouldn't be pawning that off on local communities.

你担心AI这个品牌会出问题吗？

Do you worry about AI as a brand?

你看到Matter今天大规模裁员了。

And you see matter lay off a huge amount of people today.

看到扎克伯格凌晨四点发邮件，眼看着有人丢掉工作，确实让人揪心。

It's challenging to see four AM emails from Zuck and, you know, jobs being lost.

是的。

Yeah.

我确实为此担心。

I do worry about it.

这些人都是有家庭的。

Those are those are people, and they have families.

我觉得，哈里，大概有两种看法。

And I think there are sort of two views, Harry.

到目前为止，大部分裁员其实都是打着AI的旗号。

I I I think to date, most of the layoffs were AI washed.

根本原因是我们在疫情期间招人太冒进了。

They were because we did boneheaded hiring during COVID.

真正的原因是，这些年来生产力提升的成果，现在才开始显现。

It is actually because a great deal of productivity gains has been have occurred over the years that we're just now harvesting.

能够跨部门收集信息、整合并集中呈现，正在重新定义中层管理的含义。

The ability to gather information from across the organization to synthesize it and put it in one place is now changing what it means to be middle management.

信息收集和汇报这类角色，正在被取消。

The role of information gatherers and presenters is being eliminated.

我们能够自动化这些岗位。

The ability for us to automate roles.

这些还都不是人工智能做的。

None of this is AI yet.

在我看来，绝大多数裁员其实都是因为这个原因。

That is really 90%, 95% of what the in my view, what the the terminations have been about.

把它们归到人工智能底下，确实挺容易的。

It's easy to to to put them under the umbrella of AI.

现在，人工智能才刚开始对企业产生实质性影响。

Now, AI is starting just now to have meaningful enterprise impact.

如果你的工程团队都看不出怎么利用效率高得多的工程师，那恐怕很难撑下去。

But if you are an engineering organization that can't see how to take advantage of vastly more productive engineers, I don't think you're long for this world.

我想让工程师们做的事，多出我们现有工程师人数的五十倍都不止。

I mean, the list of things I want our engineers to do is 50 times as much as we have engineers.

对吧？

Right?

随着效率提升，我们能做的工作更多，自然会招更多工程师。

As we get more productive, we do more things, we're gonna hire more engineers.

我们不会减少工程师的招聘人数。

We're not gonna hire less engineers.

我想问一下，你有没有注意到贝索夫说，他每年在Anthropic上投入三亿美元，相当于开发者薪资的3.8%？

Can I ask you, saw Benioff say that he spends 300,000,000 a year on Anthropic, which equates to about 3.8% of developer salaries on Anthropic?

要让这些公司的估值合理，这个比例得达到20%才行。

To make it justify the valuations that we're seeing for these companies, it needs to be 20%.

你对从3.8%涨到20%这个趋势有担心吗？

Do you have any concern in that movement from 3.8% to 20%?

没有。

No.

我的看法是，虽然我没仔细算过，但你看看我们给硬件工程师的薪酬，再看看他们用的EDA工具成本，我觉得实际比例可能更接近15%到20%，而不是2%到3%。

I mean, I think if you look at I've never done this in any detail, but if you look at what we pay hardware engineers and you look at what the tools, which we the EDA tools they use, I bet you're much closer to 15% or 20% than 2% or 3%.

过去，软件工程师用的工具成本很低，而硬件工程师用的EDA工具却贵得惊人。

What's happened is historically software engineers use very low cost tools, and hardware engineers used extremely expensive EDA tools.

这挺有意思的，对吧？

And so that's interesting, isn't it?

毕竟硬件出bug的代价太高了，所以我们早就习惯了用很多昂贵的工具。

I mean, the cost of bugs in hardware is so high that we became accustomed to using many expensive tools.

而软件领域，我们通常是靠堆人来解决问题，而不是靠工具。

In software, we threw people at the problem rather than tools.

随着AI变得越来越高效，我完全不觉得软件工程师每人每年花五万到十万美金在token上会有什么问题。

As AI becomes more productive, I certainly don't see a problem where software engineers are using 50 or 100,000 a year each in tokens.

全球有四千七百万名软件工程师。

There are 47,000,000 software engineers in the world.

光是软件工程师用token这一项，每年就是五万亿美元的开销。

I mean, that's $5,000,000,000,000 just in software engineering token use.

我们提到了硬件工程师，也提到了软件工程师。

We mentioned hardware engineers, we mentioned software engineers.

你觉得现在还不存在，但三到五年后会变得非常普遍的角色是什么？

What role does not exist today that you think will be incredibly commonplace in three to five years?

在过去二十五到三十年里，我亲身经历了几次技术变革，这些变革催生了以前根本不存在的岗位。

I've been a part of, over the past twenty five or thirty years, several technical transformations that produced jobs in companies that didn't exist.

在九十年代中期之前，首席信息官这个职位根本不存在。

Prior to the mid nineties, the role of CIO didn't exist.

随着思科的崛起，首席信息官才逐渐成为一个正式的职位。

CIO arose as a role with Cisco, with their sort of rise to dominance.

九十年代中期以前，企业网络的规模微不足道。

Prior to the mid nineties, the amount of enterprise networking was de minimis.

那时候通常有个职位，叫电信基础设施的副总裁。

There was a role that was often VP at telco infrastructure.

这个岗位现在已经消失了。

That job is gone.

我们不再有传统的电话系统。

We don't have a phone system.

事实上，办公桌上根本没人放电话了。

In fact, we don't have phones on people's desks.

他们用手机给我打电话。

They call me on my cell phone.

那个职位彻底消失了。

That job disappeared completely.

没了。

Gone.

以前生产PBX设备的公司，比如Rome那些，现在业务都萎缩得差不多了。

And companies that built the PBXs like Rome and all these other because that business has shrunk to nothing.

后来，二月份 Palo Alto Networks 这些安全公司崛起的时候，发生了什么？

Now later, what happened in the February with the rise of Palo Alto Networks and these other security?

首席安全官（CSO）这个职位以前根本不存在。

The role of CSO never existed prior to that.

接下来你会看到，一些新职位会因企业对AI的治理需求而兴起。

And what you're gonna see is the rise of roles that reflect the governance of AI in companies.

现在有些公司已经设立了首席AI官。

Now some companies have chief AI officers.

我不确定是不是这么回事。

I don't know if that's what it is.

但随着这些技术在公司运营中变得越来越重要，一些前所未有的新岗位就出现了。

But as these technologies become important in company's life, new jobs emerge, jobs that never existed before.

以前不存在的组织，现在也冒出来了。

New organizations exist where there were none.

而一些旧的岗位则慢慢消失了。

Previous ones disappear.

我觉得人力资源的角色发生了根本性的变化。

I think the role of HR changes fundamentally.

以前那些只负责回答问题、提供福利信息的人力资源职能，已经没用了。

The part of a r HR that just that answered questions, that provided information about benefits, that disappears.

AI能回答你所有的问题。

AIs can can answer all your questions.

它们不仅能更快地回应，还能给出更准确、更有深度的答案。

They can provide better answers, faster answers, more thoughtful answers.

它变成了完全不同的东西。

It becomes something different.

人员管理也变得不一样了。

The management of people becomes something different.

我觉得组织里还有很多其他部分都会发生根本性变化，因为AI现在能回答那些以前由人来答的问题。

I think there are all sorts of other parts of organizations that have fundamental changes because AI can answer the questions that they used to answer.

你觉得企业采用AI最大的障碍是数据结构和数据清洁度吗？

Do you agree the biggest inhibitor to enterprise adoption of AI is data structure structure and data cleanliness preventing No.

不是。

No.

最大的障碍其实是律师。

The biggest are are lawyers.

不是。

No.

真的吗？

Really.

我觉得，安全团队和律师们，因为不懂这项技术，总是直接说不行。

I think the security apparatus and the lawyers who, when they don't understand the technology, say, no.

这事干不了。

We can't do it.

他们的工作就是说不。

They're in the saying no business.

创业者的工作则是把事情做成。

Entrepreneurs are in the getting it done business.

这背后有原因——你的安全团队和律师，他们干的活儿，人人都能怪到他们头上。

There's a reason for this, that your security apparatus and your lawyers I mean, every they're they're in jobs that everyone just blames them.

没功劳，没功劳，出了问题，全怪他们。

No credit, no credit, failure, blame.

没功劳，没功劳，出了问题，全怪他们。

No credit, no credit, failure, blame.

我的意思是，这就是他们的日常，太残酷了。

I mean, that that's their life, and it's brutal.

这日子真难熬。

It's brutal.

你讲得真有说服力。

You're selling it so well.

给所有打算当律师的人听。

For any of you aspiring lawyer.

如果你在听，那就去当CISO吧。

If you're listening, go into being a CISO.

那可是真够艰难的。

It's brutally hard.

有一年我们啥事都没发生。

We had a year where nothing happened.

干得漂亮。

Well done.

那才是他们的梦想。

That is their dream.

我的意思是，只要电话一整天没响，他们就觉得天要塌了。

I mean, every day their phone doesn't ring, they're it's like, woah.

又撑过了一天。

Made it to another day.

但一遇到新技术，因为他们收益结构决定了他们得尽量规避风险。

But when confronted with new technology, because their payoff structure is such that they're in the business of trying to avoid risk.

所以他们就成了推动新事物落地的绊脚石，这种现象到处都能看到。

They are a drag on adoption of new things, and you see this across the board.

律师们根本没合同可依。

Lawyers don't have a contract for it.

也没有先例可循。

There's no precedent.

他们的行业本来就是靠回头找先例的。

That's in a business of backward looking precedent.

想让律师坐立不安？

You wanna make a lawyer uncomfortable?

我知道你女朋友是律师。

I know your girlfriend's a lawyer.

你问她去处理一个没有先例的领域试试。

Ask her to work in an area with no precedent.

他们根本不知道该怎么下手。

They don't know what to do.

他们接受的全部训练就是看别人以前是怎么做的。

Their whole training is about what has everybody else done before?

我们该怎么整合这个？

How do we synthesize this?

我们该怎么在这些规则里做事？

How do we work within those rules?

我觉得现在AI在企业里的大规模应用，主要受限于安全和法律问题。

I think the wide scale adoption and use of AI in organizations is today limited by security and legal.

等他们拍板了，我们才好动手。

Once they agree, we need to do this.

这是我们接下来要遵循的规则。

Here are the rules we will use.

这里面有巨大的效率提升空间。

There's a huge amount of productivity to be gained.

你立刻就会受到当初选择如何管理和整合数据的限制。

Then you are immediately constrained by the way you chose to husband and marshal data.

你多年来如何组织数据的方式，也会成为制约因素。

The way you chose to organize data over years.

像梅奥诊所这样的机构，花了三十年时间来系统化数据管理。

And so companies like organizations like Mayo Clinic that have been on a thirty year quest to organize data.

他们现在占据了明显的优势。

They are at a huge advantage.

像葛兰素史克这样的公司也是如此。

Same with companies like GalaxoSmithKline.

而那些在数据组织上不够严谨、不够用心的公司，则处于劣势。

And other companies who who haven't perhaps been as disciplined, as thoughtful about the organization of their data are at a disadvantage.

在安全和资源分配方面，你觉得各个行业会不会像法律行业那样突然醒悟，发现‘天哪，我们真得用AI了’？

On the security and the provisioning side, do you think we will see industries tip like legal has done where the biggest firms in the world are now going, oh, shit.

我们真的需要人工智能。

We need AI.

客户都在说，我们必须用AI，比如Harvey或者Agora。

Our clients are saying we need AI, Harvey or Agora.

我就不具体说哪个了，但反正就这两个选择。

And I'm not gonna get into which one, but like, there's two options.

就这么突然发生了。

Boom.

你觉得所有行业都会这么迅速转变，还是会慢慢接受，觉得这已经是新常态了？

Do you think all industries will follow the tipping, or do you think most will follow the slow agreement that it's the new normal?

现在的情况是，行业领头羊已经率先行动了。

What's happening is the leaders are tipping.

对吧？

Right?

而且我觉得，就连詹森都讲过，他当时在公司内部跟法务团队为是否使用 Cursor 这类工具争执不下。

And I I think even Jensen told the story that that he was battling with his own internal lawyers around the use of, I think it was cursor.

最后，他直接拍板了。

And finally, he just decreed.

我们就这么干了。

We're gonna we're gonna do it.

我觉得，到了某个阶段，领导者都会权衡：一边是效率的提升，另一边是那些看不见的潜在风险。

I think at some point, leaders weigh the productivity gains against the unseen boogeyman of brisk.

而问题在于，这些看不见的风险，有时候还真不是瞎担心。

And the problem with unseen boogeyman is sometimes they're actually real.

对吧？

Right?

不常有，但确实会发生。

Not often, but sometimes.

这才是真正的难题。

That's the problem.

他在《约翰·威克》里管他们叫什么？

What does he call them in in John Wick?

巴巴亚加？

Baba Yaga?

约翰·威克就是那个被派去杀巴巴亚加的人？

John Wick is the guy you send to kill Baba Yaga?

说实话，我觉得我虽然不算年轻了，但状态还是能跟得上。

I'm just I you know, I think for me, I'm not that young anymore, but I'm definitely capable of ex as you good.

我知道你六十多岁了，但看起来真不错。

I know you're in your sixties, but you look good.

是的。

Yeah.

这其实就是一套面部保湿的流程。

And it's it's a it's a facial moisturizing routine.

我们之前提到过安全、权限、法律，还有中间所有那些环节。

We mentioned security, permissioning, legal, everything in between.

一旦是开源的，他们就更紧张了。

They get even more freaking nervous when it's open source.

他们直接慌了。

They shit the bed.

你怎么看这个问题？

How do you think about that?

我看到越来越多公司，尤其是

I see more and more companies, especially in

硅谷的公司，不断在前沿领域突破，同时尽量

the Valley, really push the boundaries on with Frontier and then try

借助开源技术，毕竟成本优势摆在那儿。

and get as close as possible with open source given the cost advantages.

这会是未来的趋势吗？

Is that the future?

这又意味着什么？

And what does that mean?

你看，作为整个生态体系，我们在开源的法律问题上确实取得了实实在在的进展。

Look, I I think we, as an ecosystem, have made real progress in sort of the the legal gunk around open source.

但结果却是复杂得让人头疼。

But the result has been a complexity that hurts your head.

如果你真想掉进一个没底的坑，那就去跟律师聊一聊开源软件吧。

If you ever want to dive down a rat hole that has sort of no bottom, begin a discussion with lawyers about open source software.

你往下钻得越深，只会越觉得无聊，而且永远看不到头。

And there's no end to the depth and the boredom with which you will suffer as you head down this hole.

更糟的是，一些最出色的开源模型其实是国内公司做的，而且水平真的特别高。

This is made doubly worse by some of the best open source models were made by Chinese companies, and they are exceptionally good models.

比如 Kimikade 二号、DeepSeq、Quen、GLM，这些模型都极其出色。

Kimikade two, DeepSeq, Quen, GLM, these are extraordinarily good models.

虽然还没完全赶上闭源模型，但已经是非常顶尖的了。

They're not quite as good as the closed source models, but they're exceptionally good models.

我觉得这背后其实是大家在权衡：到底要不要省钱。

And I I think that is a case of people trying to decide whether it makes sense to to to to save money.

我们用起来特别顺手，运行速度也快得惊人。

They have been easy for us to adopt, to demonstrate extraordinary speed on.

这确实是个难题。

It's a hard problem.

真别羡慕那些整天琢磨这些事的法务和安全团队。

I mean, don't envy the the the the legal team and and the security groups that are thinking about these things.

现实是，这股浪潮太猛了，需求也太旺盛。

The truth is the tidal wave is so big and the demand is so high.

很多事根本来不及处理，就被冲过去了。

They often just get washed over.

那你觉得我们该不该把芯片卖给他们？

Do you think we should be selling chips to China as a result?

不该。

No.

咱们先撇开所有人的私心，哪怕我这会儿说的，其实违背了我自己的利益。

I I think let's remove all of us that are self interested, even though I'm arguing against my self interest.

对吧？

Right?

如果你把我和詹森、丽莎，还有整个芯片行业的人全都拿掉，然后问一个问题：如果我们把尖端技术卖给了搞安全业务的人，我们卖尖端技术给中国，他们的军队会不会用？

If you remove me and you remove Jensen, you remove Lisa, and you remove everybody in the chip industry, and and you say, if we sell to to somebody in the security business, and you ask this question, if we sell leading edge technology to China, will their military use it?

所有人都会说会。

Everybody says yes.

这一点上没人有异议。

There is no debate on that point.

他们的军队肯定会用。

Their military will use it.

你再问第二个问题：如果我们把尖端技术卖出去，他们的政府会不会通过本国产业，用这些技术来获得对我们的竞争优势？

You ask a second question, which is if you sell our leading edge technology, will their government use it through their industry to compete with us in an advantaged way?

答案还是肯定的。

The answer is also yes.

到这儿我就停了。

That's where I stop.

安全领域的人，还有芯片行业以外的人，都一致认同这两点是事实。

There's complete agreement that those two things are true by everybody in the security business and outside of the chip business.

当然，你可以说，把他们留在我们的体系内，才是应对这个问题的最佳方式。

Now, you can say that keeping them in our ecosystem is the best way to manage that problem.

这是一种说法。

That's one argument.

这种说法确实有道理。

And there's some merit to that.

阻止他们建立自己的完整体系，这对我们是有利的。

There is keeping them from building their own their own ecosystem is something that's in our interest.

这一点也有很强的现实依据。

There's some real merit in that.

我本人不认同这两种观点，但它们确实存在，而且有充分的依据。

I don't agree with either of those arguments, but they're real arguments and they have real merit.

至少在今天，他们是我们在产业上的对手。

They are, at least, today, our industrial adversary.

你环游世界，看到他们产业政策的成果——比如大幅降低太阳能成本、降低锂电池成本，还有在汽车行业的表现。你到处都能看到中国车，而美国车却越来越少，他们确实是我们产业上的对手。

As you travel the world and you see the results of some of their industrial policy, for example, the driving down the cost of solar, driving down the cost of lithium batteries, the results it's had in their auto industry, and the fact that you you travel the world and you see Chinese cars and with fewer and fewer American cars, they're an industrial adversary.

我不喜欢这一点。

I don't love that.

多年来，我一直在和百度、腾讯、滴滴这些公司里杰出的创业者合作，他们一点不比硅谷的人差。

I, for years, did business with extraordinary entrepreneurs there at Baidu and at Tencent and Didi and all these companies, and they're every bit as good as anybody in Silicon Valley.

我真希望世界不是这样，他们不是我们的产业对手，而是能和我们联手解决真正的问题。

And I would love a world in which they weren't an industrial adversary, and instead we were working together to solve real problems.

世界现在是什么样，就是什么样。

The state of the world is the state of the world.

对我来说，如果这是美国产业的问题——我们卖的芯片少了，也不卖给他们，我完全能接受。

For me, if it's an industry, as American industry, we sold fewer chips and we didn't sell them to China, I'm just fine with that.

有人会反驳说，就像你讲的，如果我们不卖给他们，他们就会自己搞出能力，而且会做得非常出色，到时候我们就控制不住了。

People would argue back and say exactly as you said that if if we don't sell to them, they'll build their own capabilities and they'll get very good at it, and then we won't control it.

那你为什么不觉得这个说法有道理呢？

Why do you not think that's a credible argument?

我觉得芯片行业绕不开台积电，而台积电又得靠ASML或者三星。

I think the chip industry requires you to go through TSMC, and TSMC requires you to go through ASML or Samsung.

这些环节其实都是可以有效管控的关键节点。

There are reasonable choke points to manage those challenges.

不管谁怎么看，我觉得连反对我的人也会同意，不该把最前沿的技术卖给他们。

I think the strategy in any case is even those I think who who disagree with me would would suggest that you don't wanna sell them your cutting edge technology.

我们得让他们落后一步。

You wanna keep them down rev.

我可不只想让他们落后一步，我得让他们彻底跟不上。

I'd like to keep my industrial adversaries more than down rev.

考虑到台湾对大陆的脆弱性，把台积电这样的能力和企业迁到美国本土，有多重要？

With that, how important is it that we onshore TSMC like capabilities and companies given Taiwan's vulnerability to China?

我们在美国的问题在于，缺乏长远的政策规划。

We have problems in The US in long range policy.

我们需要能跨越多届政府、持续生效的政策。

Policy that endures more than a single administration.

对吧？

Right?

我们在建设跨行政区的必要基础设施方面存在困难。

We have problems building infrastructure that is clearly needed and crosses municipality lines.

我们来看看中国做得特别好的一些方面。

Let's look at things China has done extremely well.

他们的电力基础设施非常出色。

Their power infrastructure is extraordinary.

而在美国，我们要是运气好，用的还是上世纪五十年代的技术，这真的很糟糕。

And in The US, we are a patchwork of nineteen fifties technology if we're lucky, and that's really bad.

有些事情我们确实做得不够好。

There are things we don't do well.

其中之一就是没有长远考虑像不在美国投资晶圆厂这样的决策后果。

One of them is thinking about long term consequences of decisions like not investing in fabs in The US.

我的意思是，我们失去的不只是晶圆厂。

I mean, we didn't just lose the fabs.

我们还失去了整个配套生态。

We lost the surrounding ecosystem.

我们失去了封装技术方面的专长。

We lost the packaging expertise.

我们失去了一整套相关的战略性岗位和产业，重新夺回这些至关重要。

We lost a whole set of surrounding strategic jobs and industry, and it is extraordinarily important we get it back.

我十几年来一直这么说，不是《芯片法案》，也不是补贴英特尔。

And I've been saying that for a decade and a half, that not the chips act, not subsidizing Intel.

关键是我们得在美国建起最先进的晶圆厂，还要围绕它们布局最先进的封装技术。

It's important that that we have cutting edge fabs in The US and that we surround them with cutting edge packaging technologies.

这些是战略资产。

These are a strategic asset.

如果你能推动一项没有任何阻力的政策变革，你会选哪一项？

If I said that you have one policy change that you could usher through with no resistance, what would it be?

我会允许台积电和三星在二十年内免于所有地方性法规的约束，自由选择在美国任何地方建厂。

I would allow TSMC and Samsung to a twenty year period free from all local ordinances, all of them, to build fabs in their desired location in The US.

如果是亚利桑那州，那再好不过了。

If that's Arizona, that's great.

如果是得克萨斯州，也一样好。

If that's Texas, that's great.

二十年内，别搞地方性规定。

Twenty years, no local rules.

让他们建晶圆厂。

Allow them to build fabs.

就用你们在台湾那一套标准就行。

And I I would say that use the same rules you use in Taiwan.

别建些劣质设施。

Don't don't build garbage.

完全照搬你们在世界各地成功建厂时用的施工标准和规范。

Use exactly the same construction techniques and rules, etcetera, that you that you built fabs successfully elsewhere in the world.

地方性规定只会添乱，根本不是为了管金字塔这种事。

But local ordinances are disastrous and not intended to cover pyramids.

对吧？

Right?

晶圆厂就是现代的金字塔，哈里。

Fabs are modern pyramids, Harry.

在制造业里，晶圆厂绝对是人类创造的最了不起的东西。

They are the greatest things humans make in manufacture in the manufacturing world by far.

绝对是。

By far.

别的都比不上。

Nothing's close.

安德鲁，我能问你个问题吗？

Can I ask you, Andrew?

我坐在伦敦这里。

I sit here in London.

我该担心吗？

Should I be worried?

你们美国拥有全球最顶尖的前沿实验室。

And you have the best frontier labs in The US.

中国在开源技术和制造能力方面也极其出色。

You have amazing open source and amazing manufacturing capabilities in China.

那欧洲到底有什么优势呢？

What does Europe really have?

我们在模型领域已经落后了。

We've of failed on the model front.

舒尔先生是领头人，

Mr.

但遗憾的是，和其他地方比起来差距太大了。

Scholl is the leader, but sadly no nowhere near others.

我该担心吗？

Should I be worried?

你该担心的是这种趋势。

You should be worried at the pattern.

这种在多种技术领域都缺乏突破的模式。

The pattern of sort of lack of success across a range of technologies.

不只是领先的AI公司大多在美国，就连领先的芯片公司、软件公司也一样。

It's not just that the leading AI companies are are most of them are in The US, but the leading chip companies, but the leading software companies.

对吧？

Right?

当然，也有像SAP这样的例子，还有一些其他公司。

Of course, there's some examples, SAP and and some others.

但欧洲整体上形成了一种心态：先害怕新技术，再加以监管，然后征重税，这种态度严重打压了创业精神。

But there has emerged in Europe, a sort of be afraid of it, then regulate it, tax it sort of mentality that that works against entrepreneurship.

当然，这并不是整个欧洲的情况，剑桥、伦敦、斯德哥尔摩这些地方也有不少亮点，比如Lovable团队就在做非常有意思的事。

And and I think Europe this isn't true across the board, and clearly there are pockets outside of Cambridge and and in London and and Stockholm where the guys at Lovable are doing really interesting stuff.

其实反例并不少见。

And there are all sorts of counterexamples.

但总体来看，以欧洲的人口规模，本可以在各个行业取得大得多的创新成果，可这些机会却白白放过了。

But on the whole, given its population, the opportunity to do vastly better on the innovation front across industries is sitting there unexercised.

我觉得这确实让人担忧。

That I think is is a worry.

你认为五年后，你的业务中有多少会来自欧洲？

How much of your business do you think will be in Europe in five years' time?

此外，他们在采用新技术方面也一直比较滞后。

I think along with this, they have been slow to adopt new technologies.

他们不仅在发明新技术上慢了一拍，在采纳新技术上也同样迟缓。

Not only have they been sort of slower to invent new technologies, but they've been slower to adopt new technologies.

我认为最快采纳新技术的不会是欧洲，但在两到五年内，欧洲市场仍会占据相当可观的份额。

I think the fastest adoption will not be in Europe, but in the the the two and a half to three to five year range, it it will be a meaningful portion.

这和你的观察一致吗？

Is is that in line with your experience?

我的了解主要来自远方，靠的是定期访问和与客户交流。

I mean, my experience is from a long way away and from visiting regularly and talking to customers.

你的经验也是这样吗？

Is that your experience?

应用层？没有。

Application layer, no.

我觉得我们拥有世界上一些顶尖的公司，无论是十一实验室、Synthesia，还是DeepMind。

I think we have some of the world's best companies, whether you're eleven Labs or you're Synthesia or you're DeepMind.

在基础设施、芯片和模型方面，我百分之百肯定，毫无疑问。

I think a 100% on the infrastructure, the chip side, on the model side, unwaveringly so.

所以，总体上是的，虽然你刚才提到Lovable和热点时加了一点细微差别，但说得公平。

So yes, in large part with a little bit of nuance, you, to be fair, added there with Lovable and hotspots.

因此，在这一点上，我们的看法完全一致。

So I think we're totally aligned there in respect.

是的。

Yeah.

你确实做了扎实的工作来反驳这个观点，我向你和风险投资界的其他人致敬。

I think you've done real work to argue against that, and hats off to you and the others in the venture community.

我认为资本扮演着重要的角色。

I think capital plays an important role.

我觉得，允许失败的文化在欧洲并不常见。

I think a culture in which it's okay to fail plays a role that is not traditionally in Europe.

人们通常一辈子待在一家公司，这种模式催生了保守的心态。

Careers are at one company and are long, and that breeds a a conservatism.

硅谷最厉害的地方之一，就是你即使大胆尝试、最后失败了，也不会被人看不起。

I think one of the most powerful parts about Silicon Valley is the absence of a stigma if you try to do something extraordinary, crash and burn.

风投不会因此对你有偏见。

VCs don't hold it against you.

他们反而会问你学到了什么，而这种经历往往让你更受认可。

They ask you what you learned, and often it's great experience and a credit to you.

我不太清楚这种文化是怎么形成的，但它的存在是显而易见的。

I think that is something that I've not I don't understand its history, but it's clearly present.

在我们开始激烈讨论之前，我能问你个问题吗？

Can I ask you before we do a quick fight?

我们开头提到了IPO。

We mentioned the IPO at the start.

在我看来，你选择在SpaceX、Anthropic或OpenAI上市之前，把IPO的时机把握得恰到好处。

You timed the IPO with the greatest of respects, in my mind, to absolute perfection before a SpaceX IPO, before Anthropic or OpenAI.

这是你有意为之的策略吗？

Was that strategic and deliberate with the

说真的，这到底是运气好，还是有别的原因？

greatest of respects, was it relative luck?

不是。

No.

让我跟你说清楚。

Let let let me share.

这完全是经过精心规划的。

It was 100% deliberate.

我们一年半前就尝试过上市，但因为碰上了CFIUS的审查，没成功。

We tried to go public a year and a half earlier, and we couldn't get it done because we bumped into CFIUS.

我们公司今年正好十周年。

We're ten years old.

我们多年来一直试图上市。

We tried to get public for years.

这完全是靠运气和毅力，一种永不放弃的执着。

It was 100% luck and grit, sort of a relentlessness and an unwillingness to fail.

但你们在规划时，有没有考虑过其他公司的上市时机，以及什么时候市场情绪和流动性会达到最高点？

But did you have in your mind the other IPOs and when liquidity would be best, excitement would be highest?

我们定日期的时候，根本不知道芯片股会迎来一波大涨，也不知道X AI和OpenAI他们那时候根本没法上市，这些我们全都没预料到。

Did we know when we set the date that chips would be on a run and that it was impossible for x AI and OpenAI and etcetera to get public before we didn't know any of that when we set the date.

但我们清楚的是，我们有机会成为整个市场上唯一一家纯粹的AI公司。

But what we did know was that we had a chance to be the first and only AI pure play in the entire market.

市场上就这一家，就是我们。

There's only one, and that's us.

我们有机会把一个非凡的增长故事带给那些一直被拒之门外的公众投资者。

We had a chance to bring an extraordinary growth story to public market investors who had been shut out.

我们一次又一次地尝试，这才是运气的由来，哈里。

And that we tried again and again, and that's how you get lucky, Harry.

靠的是聪明又勤奋的人，还有那种永不放弃的劲头。

It is smart, hardworking people, relentless work.

他们运气来了，偶尔也能抓住最完美的时机。

They get lucky, and occasionally, they they find the perfect time.

你有没有考虑过投资那些基于你们技术开发的公司？

Should you be investing in companies building on top of you?

你刚才提到一次又一次尝试的事。

You said about trying and trying again.

Jensen之前说过，他后悔没早点投资一些公司，现在终于决定布局英伟达周围的生态。

Jensen said before that he wishes there were companies he had invested in and about finally investing in the ecosystem around Nvidia.

你觉得Cerebras应该更积极地投资在你们之上构建的应用层吗？

Do you think Cerebras should be investing more aggressively in the application layer built on top of you?

我觉得这对我们来说，是个新出现的机会。

I think that's an opportunity that is newly available to us.

但可能不会用风投资金，或者传统的风投方式。

Probably not with venture dollars or traditional venture dollars.

对吧？

Right?

你得认真考虑你的投资人。

I I I think you have to think very carefully about your investors.

当你用风投资金时，问题就在于：我们应该投他们，还是我们的风投合伙人该投他们？

When you're using venture dollars, the question is should we be investing in them or should our venture partners be invent investing in them?

用公开市场的资金，要求就不同了，投资人的资源也不一样。

With public dollars, the mandate is different and your investors have different access.

因此，我们和客户、合作伙伴一起做些真正有意思的事情的机会变多了。

And so the opportunity for for us to do really interesting things with our our customers and our partners grows.

这包括收购公司、投资公司，还有各种合作模式，我们都得去探索。

That includes acquiring companies, that includes investing in companies, that includes different structures of partnerships, and we we have to explore them all.

你提到过多次尝试上市，还有那份坚持。

You mentioned the multiple times trying to go public and the persistence.

现在回过头看，关于上市这件事，你希望当初能早点知道什么？

What do you know now about going public that you wish you'd known when you were trying multiple times?

这和你当初想的一样吗？

Is it what you thought it would be?

不一样。

No.

听我说。

Look.

我觉得我们碰上了一个相当阻挠人的CFIUS审查问题。

I I think what happened was we bumped into a CFIUS challenge that was sort of obstructionist.

有些关于我们大客户的隐忧，一直没人说清楚，就在那儿悬着。

There were unnamed concerns that that never got articulated that that sort of lived in the ether about some of our our large customers.

后来政府换了人。

And then we got a new government.

那些顾虑一下子就没了，我们很快、很谨慎地推进，最终得到了一个公正的解决。

Those concerns disappeared, and we were able to move through it really quickly and thoughtfully with a a really fair resolution.

顺便说一句，这个解决方案其实我们一年前就提过了。

And by the way, a resolution that we had proposed a year earlier.

特朗普政府在促进商业方面真的一直更优吗？

Is the Trump administration unwaveringly better for business?

再说一遍，我现在身在英国。

Again, I sit here in The UK.

一直更优？这说法太绝对了。

Un unwaveringly better for business.

有些地方我认同。

There are things I agree with.

这个政府的政策，我有赞同的，也有不认同的，但要说一直更有利于商业，那倒未必。

There are things I disagree with in this administration, but unwaveringly better for business.

你得上场挥棒，天天都在推进才行。

You gotta be at bat taking swings, and you've gotta be building every day.

我们上市的时候，公司已经强大多了。

When we got public, we were a much stronger company.

我们的销售额也大了很多。

We had larger sales.

我们当时已经走到了路线图的更远一段。

We had we're further down our road map.

我们的客户也更优质了。

We had better customers.

你得把这两件事分开来看。

You you sort of have to separate.

我们上市不是因为那些嬉皮士，而是始终坚持在做业务。

We we didn't get public because of the hippies, but we kept building the business.

业务越做越好，这才给了我们再次尝试的机会。

The business got better and better and better, and that gave us the opportunity to try again.

我想对那些创业者、对你的听众要说的是，很多事都是你控制不了的。

That's I think the message to to your builders, to your audience who who who builds companies is a lot of stuff will happen that is not in your control.

对吧？

Right?

总会遇到低谷期。

There will be bad times.