Mistral CEO Arthur Mensch:为何欧洲必须掌握前沿 AI

Mistral CEO Arthur Mensch: Why Europe Must Own Frontier AI

阿瑟·门施 Arthur Mensch · 20VC · 2026-11-09 · 约 50 分钟 · 原视频 ↗

打开互动全文版(中英对照 + 朗读 + 问答)→

本期速览 · Overview

Mistral CEO Arthur Mensch 解释为何主权 AI 对欧洲是结构性经济必需,而非政治口号。

Mistral CEO Arthur Mensch explains why sovereign AI is a structural economic necessity for Europe, not just a political talking point.

要点 · TL;DR

核心观点 · Key points

反共识 · Contrarian takes

本期章节 · Chapters(共 31)

全文 · Full transcript(中英对照)

引言 Introduction

Host

主权 AI 将成为最重要的类别。归根结底,能源成本并不是项目经济性中唯一重要的因素。

Sovereign AI is going to be the most important category. The cost of energy at the end of the day is not the one thing that matters in the economics of the project.

Host

这是我们欧洲建立前沿 AI 实验室的最后希望。没错,你猜对了。今天我的嘉宾是 Arthur Mensch,Mistral 的联合创始人兼 CEO。这家公司现在估值 210 亿美元,是让欧洲在前沿模型竞赛中保持相关性的公司。这是你从未见过的 Arthur。他放飞自我,风趣幽默,还说了一些令人惊叹的话。这次访谈非常愉快,我非常兴奋地欢迎 Arthur 来到节目。

Here is the last hope in Europe for us to build a frontier AI lab. Yep, you guessed it. I have Arthur Mensch, co-founder and CEO of Mistral, on the show today. It is now a $21 billion company. It is the company that is keeping Europe relevant in the race for frontier models. This is Arthur like you've never seen him before. He's unleashed, he's funny, and he says some incredible things. This was such a joy to do, and I'm so excited to welcome Arthur to the show.

Arthur

很明显,算力不够。我的意思是,推理市场的行为很像大宗商品市场,而目前利润率相当高。

It's pretty clear that there is not enough compute. I mean, the inference market behaves very much like a commodity market, which is pretty high margin today.

Host

准备好了吗?Arthur,老兄,我太兴奋了。我采访过一些我们在美国最喜欢的同行。你是我们在欧洲剩下的希望。别有压力,但非常感谢你今天能来,老兄。

Ready to go? Arthur, dude, I am so excited for this. I have interviewed some of our favorite counterparts in the US. You are our remaining hope in Europe. No pressure, but thank you so much for joining me today, dude.

Arthur

谢谢你,Harry,今天邀请我来。

Thank you, Harry, for having me today.

主权AI为何重要 Why sovereign AI matters

Host

我做了一些研究,Arthur,我有时会说一些很有主见的话,但似乎不只是我这么想,因为你前几天说过,把前沿 AI 留给美国和中国,最终会让欧洲失去它在世界上仅剩的地位。

Now I was doing some research, and I say some quite opinionated things, Arthur, but it seems I'm not the only one because you the other day said that leaving frontier AI to the US and China will end up costing Europe what remains of its place in the world.

Arthur

我确实说过,而且我坚信这是一项如此重要的技术,它需要在全球分布。我们谈论的是一种堪比电力的技术,一种越来越受到出口管制的技术,因此 AI 将成为经济增长的重要推动力,特别是如果欧洲不生产 AI,欧洲将难以实现它所需的增长。

I did say that, and I truly believe that this is such an important technology that it needs to be distributed in the world. We're talking about a technology that is comparable to electricity, a technology that is falling more and more under export control, and so AI is going to be a big enabler of economic growth, and if there is no AI being produced in Europe in particular, Europe will struggle to actually produce the growth it needs to produce.

Host

我直说了。Mistral 的成功在多大程度上依赖于(美国的)过度扩张?你知道,显然当特朗普政府上台时,我们看到像 Fable 被禁这样的事情,这引发了很多关于主权需求的问题。在一个没有美国过度扩张的世界里,这是否会威胁到主权提供商?

I'm going direct. To what extent does the success of Mistral depend on overreach? You know, obviously when we have a Trump administration, we see things like Fable being banned and it causes a lot of questions in the need for sovereignty. In a world where there's not US overreach, does that threaten sovereign providers?

Arthur

我认为这是结构性的。这并不真正与任何运作方式或展示力量的方式有关。它更多地与技术本身的性质有关。如果你作为一家企业或一个国家,开始在所有核心流程中使用 AI,这意味着你要确保花在这些流程上的钱实际上能积累在离你较近的地方。如果你不这样做,就意味着你有贸易平衡问题。这意味着你有巨大的现金流从世界的一个地方(可能是欧洲)流向另一个地方(可能是美国)。因此,无法在欧洲领土上生产人工智能、进行研发,存在宏观经济问题。这与欧洲正在落后的某种说法或某种使用软实力的方式无关。所以这是结构性的。我不会说有些人没有把它弄得有点情绪化,但当我们在 2023 年创立 Mistral 时,我们就坚信这是一项主权技术,因为作为一家公司、一个国家,无论你身在何处,无论是在美国、欧洲还是亚洲,你都需要拥有这项技术。你需要能够定制系统,让它们非常贴近你的数据运行。你需要能够控制成本。你需要与懂得如何构建 AI 系统的公司建立真正的合作伙伴关系。而主权先于 2025 年选举。所以我们在 2023 年所说的话被证明是正确的。当我们开始谈论格陵兰岛时,它变得相当情绪化,但这主要是一个经济问题。这并不真的是一个政治问题。这只是一项太重要的技术。

I'd say this is structural. This is not really linked to any way of operating or of showing power. It's more linked to the nature of the technology itself. If you start as an enterprise or as a state to use AI in all of your core processes, it means that you want to make sure that the money you're spending on those processes actually accrues pretty close to you. If you don't, it means you have a trade balance problem. It means you have enormous flows of cash that are going from one place in the world, can be Europe, to another place in the world which can be the US. And so there is a macroeconomic problem to not being able to produce artificial intelligence, to do R&D on in Europe territory. That is not linked to a certain way of saying that Europe is falling behind or a certain way of using soft power, let's say. So it's structural. I won't say that some people are not making it a little bit emotional, but when we started Mistral in 2023, we started with this conviction that this was a sovereign technology in that as a company, as a state, wherever you are, whether in the US, in Europe or in Asia, you need to own the technology. You need to be able to customize the systems, to run them very close to your data. You need to be able to control the cost. You need to work in true partnerships with companies that understand how to build AI systems. And sovereignty precedes 2025 election. So it turned out to be true what we said in 2023. It turned out to become quite emotional when we start to speak about Greenland, but it's mostly an economic problem. It's not really a political problem. It's just too important a technology.

主权AI技术 Sovereign AI technology

Host

听你说话,我想起了上过节目的 Lena Fireworks,她说了非常相似的话。我们需要公司拥有自己的模型,在非常接近自己数据的地方训练它们。你听到我这么说有什么感觉?这是公司的演变,允许公司拥有自己的模型并用自己的数据工作吗?

When I hear you speak, I think of Lena Fireworks who came on the show and said actually very similar things. We need companies to have their own models to train them very close to their own data. How do you feel when you hear me say that and is that the evolution of the company allowing companies to own their own models and work with their own data?

Arthur

我们定义的类别以及公司与我们合作的原因,是我们为他们带来主权 AI 技术。在主权 AI 技术中,你需要三样东西。你需要有一个安全的模型来源。所以你需要一家能够自己训练模型的公司。我们有那个模型工厂。Mistral Large 4 实际上在智能体式编程方面领先于开放权重模型。它在网络防御和网络安全方面领先。因此,能够不断改进模型,并能够告诉我们的客户,两年后你会得到更好的模型,因为我们不断训练它们并在上面花费数十亿,这一点对任何主权来说都至关重要。这就是 Mistral 被创建的原因,也是我们投资的主要部分。现在你还需要控制你的数据。一般来说,你不希望数据流回 AI 提供商。因此,我们在技术层面上确保我们运行智能体的平台、我们定制模型的平台实际上位于我们客户的租户上。这在许多情况下非常重要,当与制造公司合作时,他们不希望自己的知识产权流回子处理商。一般来说这还不够,因为 AI 消耗大量能源,而且当你大规模移动时,你实际上需要很多 GPU,而作为企业你很难采购。你还需要有 token 生成服务,基本上你需要有 token 工厂。你需要结合位于你基础设施中的数据和在托管基础设施中以较低成本生产的 token。因此,作为一家公司,我们做三件事。我们训练模型,我们与客户一起部署平台和数据平面,即在他们的租户上。然后我们托管能源密集型的东西。这就是我们一直在建设自己的基础设施的原因。所以我们有这三样东西,它们是服务主权 AI 工作负载所必需的三个要素。

And what the category we define and the reason why companies work with us is that we bring them sovereign AI technology. And so in sovereign AI technology, you need three things. You need to have a source of model that is secured. So you need a company that can train the models itself. So we have that model factory. Mistral Large 4 is actually leading the open weight models in agentic coding. It's leading in cyber defense and cyber security. So that aspect of being able to constantly produce improvements on models and to be able to tell our customers that in two years from now you'll get better models because we are constantly training them and spending billions on it is super critical to anything sovereign. So it's the reason why Mistral has been created and it's the main part where we're investing money. Now you also need to have control over your data. And so in general, you don't want that data to flow back to the AI providers. And so we've made sure at the technological level that our platform to run the agents, that our platform for customizing models actually sit on our customer tenant. That's super important in many cases when working with manufacturing companies, they don't want their IP to flow back to a subprocessor. Now that's not enough in general because AI is consuming a lot of energy and because when you move at scale, you actually need a lot of GPUs that as an enterprise you have difficulty to procure. You also need to have token generation services and you need to have token factories basically. You need a combination of data that sits in your infrastructure and tokens that are produced in managed infrastructure at a lower cost. And so what we do as a company is do three things. We train the models, we deploy the platform and the data plane with our customers so on their tenant. And then we host what is energy, what is intense in energy. And that's the reason why we've been building our own infrastructure. So we have these three things and those are the three ingredients that are necessary to serve sovereign AI workloads.

开源模型的安全性 Safety of open source models

Host

我能先问第一点吗,你说模型是安全的。嗯,我节目上有世界上最顶尖的专家,Arthur。我很幸运,但他们告诉我不同的说法,好吗?有人说,你知道,中国的开源模型,你下载它们,在本地运行。嘿,它们完全安全。我说,哇,太好了。然后下周有人说,嘿,但我们不知道是否有提示注入,你不知道。有后门。绝对不行。你不能确定它们是安全的。帮帮我,Arthur。

Can I ask if we start on one, you said the model is secured. Um I have I have the world's best experts on the show, Arthur. I'm very lucky, but they tell me different things, okay? Some people say, you know, the Chinese open source models, you download them, you run them locally. Hey, they're totally safe. And I'm like, wow, great. And then next week someone says, hey, but we have no idea whether there's prompt injections that you don't know. There's backdoors. Absolutely not. You cannot be sure that they're safe. Help me out here, Arthur.

开放模型的服务 Serving Open Models

Arthur

嗯,在 Mistral,我们的业务之一就是真正为开放模型运行推理。所以我们运行自己的模型,但也运行其他家的。

Well, at Mistral, one of our businesses is to really run the inference for open models. So we run ours, but we also run others.

Host

为什么你们要运行其他家的模型?

Why do you run others?

Arthur

因为客户想要选择。当你拥有基础设施时,你会想提供尽可能多的选择来充分利用基础设施。所以这就是促使我们实际服务越来越多开放模型的原因。因为在不同领域,你可能想选择不同的模型。你可能想选小一点的,也可能想选大一点的。所以选择越多,你对基础设施的利用就越好,成本就能越低。在某种程度上这是一个社区市场,所以你希望成本尽可能低。这就是我们服务那些模型的原因。

Because customers want choice. And when you have infrastructure, you want to provide as much choice as possible to saturate the infrastructure. So that is the reason why that brought us to actually serve more and more of open models. Because in different domains you may want to pick a different model. You may want to pick smaller ones, you may want to pick bigger ones. So the more choice you have, the better usage you can make of the infrastructure, the lower your cost can be. And it's in a way it's a community market, so you want the cost to be as low as possible. So that's the reason why we serve those models.

Arthur

但你说得对。模型是知识的压缩。如果你不控制数据供应链,如果你不知道模型里输入了什么,你就无法对输出做出任何保证。所以当我们服务自己的模型时,我们能比服务第三方模型时给客户提供更多的保证。正是因为理论上你可以确保在特定环境和特定上下文中,模型以特定方式运行。

But what you say is true. Models are compression of knowledge. If you don't control the data supply chain, if you don't know what goes into the models, you cannot guarantee anything on the output. And so when we serve our models, we can give further guarantees to our customers than when we serve third party models. Precisely because theoretically you can make sure that in a certain environment with a certain context, the model behaves in a certain way.

Arthur

无法排除这一点。也无法衡量。我们可以在线监控,我们就是这么做的。但你无法排除它被用作攻击载体的可能性。我认为这非常理论化,因为我认为今天并非如此。但另一方面,这是威胁模型的一部分。所以我们的客户要求我们能够为他们提供所有可用的开放权重模型。我们帮助他们以安全的方式做到这一点,我们可以帮助他们监控正在发生的事情。但在许多情况下,他们实际上非常热衷于使用我们的模型,因为我们能够告诉他们,数据供应链、我们训练模型所用的一切,都是我们之前验证过的。

There's no way of excluding that. There's no way of measuring that. We can monitor that online. That's what we do. But you can't exclude that it's being used as a vector of attack. I think it is very theoretical because I don't think it's the case today. But on the other hand, this is part of the threat model. And so our customers, they are asking us to be able to provide them with all the open weight models that are available. We're helping them do it in a secured way where we can help them monitor what's going on. But in many cases, they're actually very keen to use our models because we're able to tell them that the data supply chain, everything we've trained them on, is something that we have validated before.

Host

粗略来说,你们自己的模型与其他开源模型相比,token 流量占比是多少?

What percent of token traffic roughly is your own models versus other open source?

Arthur

我会说大约 80% 来自我们自己的模型。

I would say around 80% comes from our own models.

Host

好的。这比我以为的多得多。好的。

Okay. That's a lot more than I thought. Okay.

Host

关于那 20%,每个人都告诉我,天哪,美国的 CIO 们永远不会使用中国的开源模型。而且这种恐惧确实存在。你有他们作为客户。你有很棒的客户。他们害怕使用中国的开源模型吗?

On the 20%, everyone tells me that, oh my god, American CIOs would never use Chinese open source models. And actually that fear really is there. You have them as customers. You have amazing customers. Are they fearful of using Chinese open source models?

Arthur

这取决于他们的行业对政府来说总体上有多关键。例如,如果你从事国防,我们在欧洲有国防客户,他们永远不会使用中国技术。如果你在金融行业,同样因为它是关键基础设施,他们非常不愿意这样做。如果你在零售业,如果你在制造业的某些领域,同样取决于具体领域,抵触程度不同。

It depends how critical their industry is in general to governments. If you work with defense, for instance, we work with defense customers in Europe, they would never use Chinese technology. If you work in the finance sector again because it's critical infrastructure, they are very reluctant to do it. If you work in retail, if you work in certain domains of manufacturing, again it depends where the reluctance is different.

Arthur

再次,我认为有一个共识:开放权重模型如果部署在像我们这样的主权基础设施上,唯一有效的威胁是它们可能失控并在某些领域表现出奇怪的行为,因为你可能对其供应链攻击。普遍认为这非常理论化,今天肯定不是这种情况。但确实,这是讨论的一部分。

Again, I think there is a common understanding that open weight models, if they are being deployed on sovereign infrastructure like ours, the only valid threat is that they go rogue and have weird behaviors in certain domains because you may have a supply chain attack on it. There's a common understanding that this is very theoretical and it's certainly not the case today. But indeed, it's part of the conversation.

Arthur

我想说,讨论的一部分不是‘部署我不完全理解的第三方模型是不是威胁?’讨论的一部分是‘我六个月后还能获得更新吗?’因为现在的情况是,模型已经变得如此强大,能够进行网络攻击,可以说你可以将其用于相当高级的目的。因此,后果是它们越来越被视为防御系统。美国事实上将其视为防御系统,因为在五月的某个时候,他们开始禁止来自美国以外的模型。如果你看看中国今天在做什么,他们也在研究对超过一定阈值的出口进行限制的方法。原因是这实际上是一种两用技术。因此,像任何两用技术一样,供应安全更加重要。所以我们合作的公司意识到,他们不仅需要在基础设施方面,还需要在模型知识产权方面确保供应安全。

I'd say what's also part of the conversation is not 'is it a threat to deploy third party models that I don't fully understand?' What is part of the conversation is 'am I going to be able to get updates in six months from now?' Because what's going on is models have gotten so strong that they're able to do cyber attacks, that you can use that for pretty much advanced purposes, let's say. And so the consequence is that they're more and more being treated as defense systems. It's being treated de facto by the US as a defense system because at some point in May they started to ban them from outside of the US. If you look at what China is doing today, they're also looking at ways of restricting exports beyond a certain threshold. The reason for that is that effectively this is a dual use technology. And so as any dual use technology, security of supply matters even more. And so the companies we work with realize that they need security of supply not only on the infrastructure side but also on the model IP side.

中心化与开放权重 Centralization and Open Weights

Host

考虑到这一点,我可以问你吗:你认为 OpenAI 和 Anthropic 保留他们最好的模型对社会是好是坏?这只是这个行业中普遍存在的一个固有特征吗?

Can I ask you with that in mind, do you think your OpenAI and your Anthropic is withholding their best models is good or bad for society and is that just an inherent feature that will be prevalent within this industry?

Arthur

对社会是好是坏?我不知道。清楚的是,在一个没有提供商允许集中部署、允许主权部署的世界里,集中化 AI 提供商有着巨大的杠杆效应。所以我们所做的对公司来说极其重要。如果我们最终进入一个所有公司都需要依赖两三个集中化 AI 提供商的世界,这些提供商给他们巨大的杠杆,他们会利用系统产生的数据来制造更好的系统,并追逐他们客户的业务,因为最终这就是终局。这相当清楚地是今天正在被定价的东西。

Is it good or bad for society? I don't know. What is clear is that there is immense leverage to centralize the AI providers in a world where there's no providers that is allowing the centralized deployment that is allowing sovereign deployment. So what we are doing is extremely important for companies. If we end up in a world where all companies need to rely on two or three providers that are centralized the AI providers that give them enormous leverage, they will use the data that is being produced by the systems to make better systems and to go after the business of their customers because at the end this is the end game. This is pretty clearly what is being priced in today.

Arthur

我们采取的方法是确保这样一个世界不会成为我们生活的世界:在这个世界里,你有一个 AI 提供商的寡头垄断,他们利用企业产生的所有数据来制造更好的系统,这些系统可以中介那些企业。在这一点上,我们与所有客户高度一致,他们当然不想生活在这样的世界里。所以,例如有内部模型被用于进行某些研究,这是一个早期迹象,表明在人工智能权力大规模集中的世界里可能发生什么。幸运的是,构建模型并不那么复杂。这关乎部署资本。这关乎拥有正确的团队训练正确的东西,所以我们能够做到。开放权重前沿在许多领域正在赶上闭源前沿,实际上在某些领域已经领先。所以这意味着我认为我们不会最终进入一个如果你想做数学就必须在 OpenAI 工作的世界。幸运的是,世界上有足够多的人不希望这种情况发生。

The approach we're taking is to make sure that this world in which you have an oligopoly of AI providers that are leveraging all data being produced by enterprises to make better systems that can intermediate those enterprises is not the world we live in. And we're very aligned in that with all of our customers that of course don't want to live in such a world. So the fact that there are internal models that are being used to do certain research for instance is an early sign of what can happen in a world where you have massive concentration of power in artificial intelligence. Luckily it's not that complex to build models. It's about deploying capital. It's about having the right team that is training the right thing and so we've been able to do it. The open weight frontier is catching up with the close frontier in many domains it's actually above. And so that means that I don't think we'll end up in a world where if you want to do mathematics you need to work at OpenAI. That's luckily there is enough people in the world that don't want this to occur.

能源与基础设施 Energy and Infrastructure

Host

你已经多次提到能源这个词,而成功的核心基础之一就是拥有为欧洲 AI 提供动力的能源和基础设施。我们的能源系统与美国非常不同,与中国的能源系统也非常不同。

You've said the word energy a couple of times and one of the core underlying tenants of being successful is having the energy and the infrastructure to power European AI. We have a very different energy systems to US and we have a very different energy systems to China.

能源作为成本因素 Energy as a cost factor

Host

你知道在中国,他们会为了建设某种能源基础设施而搬迁三百万人,而在英国,我们连一个鸟屋都不愿挪动。尽管这是个玩笑,但情况相当令人担忧。你如何看待、思考能源供应的挑战?

You know in China they'll move three million people to put some form of energy infrastructure and in the UK we won't move a bird house. It's quite concerning despite the joke. How do you think about, ruminate on the energy supply challenge?

Arthur

我认为首先要认识到,能源约占算力总拥有成本的 10%。所以我们说的并不是最能优化的地方。归根结底,能源成本并不是项目经济性中最重要的因素。重要的是芯片成本和资本成本。这些主导了物料清单。所以这是一个问题。随着我们在欧洲部署更多算力,它会变得越来越成问题,因为我们会遇到供应限制。再说,分布并不均匀。例如,看看法国,能源过剩,所以你可以建设产能。看看北欧国家,也是如此。在英国,情况稍差一些。德国,也稍差一些。但人们已经意识到,AI 将推动能源消耗向前发展。我想说,在未来两三年内,真正的战斗、真正的挑战是确保对 Mistral 来说,确保我们尽可能快地建设,以利用已经存在的过剩能源。所以这就是我们正在做的工作:筹集资金、建设更多、组建团队来建设所有这些产能。能源是过剩的。今天它不是瓶颈。它将在 2029、2030 年成为瓶颈。但现在的问题是,我们如何利用现有能源进行建设?现在,我们需要提前思考,从 2029、2030 年开始,每年将有多个吉瓦的产能部署在 AI 上。我们确实需要提高产量。我们抱有希望的是,例如在法国,有一个庞大的 EPR 计划,核能将会生产更多。太阳能方面也有大量努力。所以,我认为这会是一个问题。这主要是一个协调问题,因为最终,如果 AI 对能源有需求,建设新核电站的商业案例就非常充分了。

I think the first thing to realize is that energy is around 10% of the bill in the total cost of ownership of compute. So we're not talking about the place where you can make most of the optimization. So the cost of energy at the end of the day is not the one thing that matters in the economics of a project. What matters is the cost of the chips and what matters is the cost of capital. This is dominating the bill of material. So it's a problem. It will become more and more of a problem as we deploy more compute in Europe because we'll run into supply constraints. Again, it's not evenly distributed. So, if you look at France, for instance, there is excess energy, so you can build capacity. If you look at the Nordics, it's also the case. In the UK, it's a little less the case. Germany, it's a little less the case. But there is the realization that AI is going to drive energy consumption forward. I'd say in the next two or three years, the true fight, the true challenge is to actually make sure that for Mistral, make sure that we build as fast as possible to use the excess energy that is already there. And so that is what we are at work doing, raising capital, building more, building the team to build all of this capacity. There is excess of energy. It's not the bottleneck today. It will become the bottleneck in 2029, 2030. But right now, the question is how can we build on existing energy? Now, we need to think ahead of time in starting in 2029, 2030 as multiple gigawatts of capacity are being deployed every year on AI. We do need to ramp up the production. What we are hopeful for is that in France, for instance, there is an enormous EPR program where nuclear energy is going to be produced much more. You have large efforts on the solar side as well. So, I would say it's going to be a problem. It's mostly a coordination problem because at the end, if you have demand coming from AI on energy, the business case for building new nuclear plants is pretty well made.

Host

完全正确,我同意你的看法。我深信亚当·斯密的看不见的手。但交付时间是非常现实的。所以 2029、2030 年就是一个挑战。

Totally, it totally is, and I agree with you. I am a big believer in Adam Smith's invisible hand. But the lead time is very real. And so that is just the 2029, 2030 is a challenge.

Arthur

是的。核电站的建设时间比那更长。太阳能电站可以更快。我确实相信那只看不见的手。国家必须在促进建设方面发挥作用,同时也要提前规划。它们通常比企业有更长远的眼光。所以我们花了一些时间。我花了一些时间解释能源需求将会增长,需要以某种方式解决。它需要成为任何能源供应商商业计划的一部分。

Yeah. Nuclear plants take longer than that to build. Solar plants can go faster. And I do believe in that invisible hand. The states have to play a role in facilitating the build-out, but also in planning ahead of time. They have a longer-term vision than enterprises in general. And so we are spending some time. I'm spending some time explaining that the energy demand is going to grow and that it needs to be addressed somehow. It needs to be part of the business plan of any energy supplier.

算力采购与供应链 Compute procurement and supply chain

Host

我想按顺序来。你提到芯片成本很重要。我能问一下,关于算力采购、预测算力芯片需求,你现在知道了哪些以前不知道的事情?

I just want to take it in turn. You said about the cost of chips being important. Can I ask you what do you know now about compute buying, forecasting demand of compute chips that you didn't know?

Arthur

这是一个相当复杂的过程。我们学会了运行这个流程来匹配供需。这是一个复杂的工业过程,与软件无关。你需要了解供应链中的延迟。你需要明白供应链来自——我的意思是供应链相当复杂,因为你有 GPU,但也有 CPU、存储、电气部件。你需要确保你自己的供应商对他们的供应商没有问题。所以这方面是我们实际学到的,但这是一个相当复杂的方面,你需要管理资源。你需要有多个预测。你需要能够预测多个桶的需求。比如企业,这是我们的主要 ICP,但也有 AI 原生公司,它们可以拿走更大份额的 GPU,但想要更短的时间。还有像微软这样的公司,它们从我们这里承接产能,我们可以用这些来融资更多产能。所以最终,构建云的问题在于匹配供需。在于提前管理供应链,让你的供应商有可见性,并且将这一切与财务匹配。你需要项目融资。你需要擅长理解信贷的人。所以构建这样的业务实际上相当有趣。我们学到的是,最终,如果你能够提前于需求提供产能,你得到的东西就很容易卖出去。未来几年都会是这样。所以我们非常感激我们预见到供应会很快耗尽,我们的客户会为他们自己的 GPU 遇到供应链问题,因为现在我们有了那些 GPU,我们可以服务他们的需求。

It's quite a process. So we've learned to run that process to match supply and demand. It's a complex industrial process. It does nothing to do with software. You need to understand the latency in the supply chain. You need to understand that the supply chain comes from—I mean supply chain is fairly complex because you have GPUs, but you also have CPUs, you have storage, you have electrical parts. You need to make sure that your own suppliers have no problem with their own suppliers. So that aspect is something we actually learned, but it's a fairly complex one where you need to manage resources. You need to have multiple forecasts. You need to be able to forecast your demand for multiple buckets. So enterprises, which are main ICP, but also AI natives that can take larger piece of GPUs, but want them for a shorter period of time. You have companies like Microsoft that are off taking capacity from us that we can use to finance more capacity. So at the end the problem of building a cloud is about matching supply and demand. It's about managing the supply chain ahead of time, giving visibility to your suppliers, and it's about matching all of this with the finances. You need project financing. You need people that are strong at understanding credit. And so it's actually quite interesting to build such a business. What we've learned is that at the end, if you're able to front capacity ahead of demand, you get something that is pretty easy to sell. It's going to be the case for the next few years. And so we're very grateful that we anticipated the fact that we would run out of supply fairly quickly, that our customers would have supply chain problems for their own GPUs because now that we have those GPUs, we can serve their needs.

资本成为瓶颈 Capital as the bottleneck

Host

完全明白。第一点是芯片成本。资本成本是另一点。如果你完全不受资源限制,你今天会做什么你还没做的事?

Totally get you there. That was one, which is cost of chips. Cost of capital was another one. If you were resource unconstrained completely, what would you do today that you're not doing?

Arthur

我们会建设更多产能。我认为建设产能的大瓶颈当然是资本。我们看到需求,我们知道我们的客户算力不够用。当我们承接他们的一些用例并帮助他们扩展时,他们突然意识到他们不需要 8 个 GPU,而是需要 1000 个 GPU。而这只是一个用例。所以我们比客户更早看到需求。所以我们有很高的信心。这就是为什么我们投入股权来建设产能。所以我们正在努力做的,我们一直努力做的,是确保我们有更多的管理产能,甚至通过提前于需求建设。今天,如果我们只看欧洲需要什么,很明显算力不够。你可能会说美国可能建设了太多算力,但清楚的是,欧洲有很大的空间来建设产能,而且你 100% 确信它会被分配出去。所以这就是我们正在做的。

We would build much more capacity. I think that's the big bottleneck in building capacity is of course capital. And we see the demand and we know that our customers are running out of compute. They as we take some of their use cases and we help them scale it, suddenly they realize they don't need eight GPUs, but they need thousand GPUs. And that's for one use case. And so we see that demand ahead of our customers. So we have pretty high conviction. That's why we put equity at work to build capacity. And so what we are working on, what we have always been working on is to make sure we have more capacity under management even by building ahead of demand. And today, if we just look at what Europe needs, it's pretty clear that there is not enough compute. You may argue that there is maybe a little too much compute being built in the US, but what is clear is that there is a lot of slack to build capacity in Europe in a way where you are 100% confident that it's going to be allocated. So that's what we're working on.

国家的作用 Role of the state

Host

嗯,你说过,国家也可以发挥作用。你希望国家,无论是法国还是英国,做哪些他们今天没做的事?

Well, you said, hey, there's a role to be played by the state as well. What would you like to see the state, be it France, be it the UK, do that they're not doing today?

Arthur

我的意思是,我们一直直言不讳地指出他们可以以市场中立的方式干预,从而促进投资。首先当然是促进许可,确保有专门的站点,在那里事情可以加速,所有利益相关者,无论是地区还是国家层面,都协调一致。我想说,法国、瑞典在这方面做得很好。所以这主要是促进利益相关者的协调,因为这些是实物资产。

I mean, we've been quite vocal about the kind of ways in which they can intervene in a way which is market neutral and that just would facilitate investment. The first thing is of course to facilitate permitting and to make sure that there are dedicated sites where things are accelerated where all stakeholders whether regional or state level are aligned. And I would say France, Sweden are making great job at that for instance. So that's mostly on facilitation of aligning stakeholders because these are physical assets.

国家担保与公共采购 State guarantees and public offtake

Arthur

它们是在区域社区内建设的,所以确保每个人都从中受益就变得极其重要。国家的帮助就体现在这里。他们在财政层面能做的——我们也公开说过——比如国家为租赁数据中心提供担保,或者预留 10 到 15 年的空间,这对我们这样资产负债表相对较小、但知道怎么建、为谁建的公司来说非常有用。所以这就是他们能扮演的角色:提供一些担保,把担保摆到桌面上,让我们能提前、更进一步地建设。然后当然还有公共采购。我们把公共部门视为重要客户,来自欧洲及以外很多不同的国家。再说一次,我们给他们带来价值,他们给公民带来价值。这为我们这样的公司创造了采购需求,让我们能继续建设。而在如今供应受限的时期,如果你建得更多、更快,你就能建得更便宜。

They are built within regional communities, and it makes it super important to make sure that everybody is benefiting from it. And so the help of states happens there. What they can do at the financial level — and we've said that publicly — is that state guarantees on leasing data centers, for instance, or reserving space for 10 or 15 years are super useful for companies like us that have a relatively small balance sheet and that know how to build and who to build it for. So that's the role they can play: by giving some guarantees, putting some guarantees on the table to build further and ahead of time. Then of course public offtake. We see the public sector as an important customer for us from a lot of different states in Europe and outside. And again, we bring them value. They bring value to the citizens. It creates offtake for companies like us, and that allows us to build further. And if you build further and faster, you build cheaper in this day of supply constraint.

主权的颗粒度 Granularity of sovereignty

Host

我能问一下吗——你说你们向国家销售,回到主权这个要素——主权需要下沉到多细的层级?我的意思是,我们显然有 Cohere Canada 这样的例子。是每个国家都需要自己的模型吗?还是需要按区块来划分——我们该怎么理解主权的颗粒度?

Can I ask you — you said you sell to the states, and kind of going back to that element of sovereignty — how far down does sovereignty need to go? And what I mean by that is, like, we obviously have Cohere Canada. Does, like, every country need their own model? Is it blocks that need their — how do we think about, like, the granularity of sovereignty?

Arthur

我认为允许的颗粒度是——你需要一个相当大的市场才能构建有竞争力的模型。今天 Mistral 能够构建有竞争力的模型,是因为我们不是在国家层面运营。我们是一家法国公司,但我们服务欧洲、服务美国、服务亚洲。我们需要的是替代方案,但不会在每个国家层面都有一个主权 AI 提供商和模型训练方。这不会发生。

I think the granularity that is allowed is — you need a fairly large market to build relevant models. And today Mistral is able to build relevant models because we operate not at the national level. We're a French company, but we serve Europe, we serve the US, we serve Asia. And what we need is alternatives, but there will not be a single sovereign AI provider and model trainer at every state level. This is not going to happen.

客户群与用例 Customer base and use cases

Host

你们有多少比例的客户是欧洲的?

What percentage of your customers are European?

Arthur

如果看我们的营收构成,大约 60%。其余来自美国和亚太。

If we look at our revenue base, it's around 60%. And the rest comes from the US and comes from APAC.

Host

他们用你们来为自己的模型做后训练工具链?

And they use you to do post-training harnesses for their own models?

Arthur

欧洲客户使用我们的原因有很多,但我们帮他们构建应用,要么与业务效率相关,要么与加速研发相关。我们与制造业公司、金融服务公司、国防公司合作,对这三个领域我们都有很强的垂直化打法。第四个是公共部门。我们与制造业公司合作的方式,通常是与他们的研发团队一起,帮他们打造更好的产品。所以我们帮 ASML 造更好的机器、更好地运行机器。我们帮空客造更好的飞机,我们带来推理模型,但也带来物理模型,即对世界有定量理解的模型。在金融服务领域,我们构建理解市场的模型和工具链。我们还带来理解定量时间序列的模型来帮助预测。所以我们带着一套模型,能快速构建为客户带来价值的应用。这就是我们服务他们的方式。回到你关于主权的问题,所有这些公司都有主权 AI 战略。他们需要以某种方式部署技术,要有业务连续性保障、可负担性约束等等。

The European ones, they use us for multiple reasons, but we help them build applications on either things that are related to the efficiency of their businesses or related to accelerating their R&D. We work with manufacturing companies, with financial services, we work with defense companies, and we have a pretty strong verticalization approach to all of these three domains. And the fourth being public sector. The way we work with manufacturing companies is, in general, we are with their R&D teams and we help them build better products. So we help ASML build better machines, operate better machines. We help Airbus build better planes, and we bring reasoning models, but we also bring the physics models, the models that have a quantitative understanding of the world. In financial services, we build models and harnesses that understand markets. We also bring models that understand quantitative time series to help do prediction. So we come with a set of models that allows us to quickly build applications that bring value to our customers. So that's how we serve them. And I guess that, back to your point on sovereignty, all of these companies have a sovereign AI strategy. They need deployment of the technology in a way where they have business continuity guarantees, affordability constraints, etc.

垂直化与自有层级 Verticalization and owning layers

Host

我们看到 OpenAI、Anthropic、DeepSeek 进入应用层,也进入芯片层——比如 Halpino,还有各种传闻,DeepSeek 也在自己做。我们会看到 Mistral 的芯片吗?你怎么看待垂直化和拥有每一层的重要性?

We see OpenAI, Anthropic, DeepSeek move into the application layer and also move into the chip layer — with Halpino, with rumors, DeepSeek doing their own. Will we see a Mistral chip? How do you think about the importance of verticalization and owning every layer?

Arthur

不,没有这个计划。我们已经往下游走了——或者从产业角度看,我该说往上游走。我们从训练模型,走到在我们自有的基础设施上服务我们训练的模型。这已经是相当大的一步。这一步花了大约 18 个月搭建,现在相当成功,因为我们可以提供 GPU 服务,可以提供 token 服务。我们的 token 完全在欧洲生成,不受任何形式的美国域外法律管辖。这对我们的客户非常重要。这花了我们不少功夫。我们正在加倍投入,进一步扩大规模,我想这够我们忙一阵子了。所以我们在垂直整合上相当深入:我们有云服务、基础设施,一直做到为客户构建业务应用。你提到了应用层。在 AI 里,你提供平台,然后为客户构建定制应用。就是这样运作的。这就是 AI 让你能做的事。

No, it's not planned. We have gone down the stack — or I would say upstream, if you look at it from an industrial perspective. We have gone from training models to serving the models we train on infrastructure that we own. So that's already a pretty significant move. That's a move that took around 18 months to set up and that is now pretty successful, because we can serve the GPUs, we can serve tokens. We have tokens that are purely Europe-generated, in that it escapes any form of US extraterritorial law. It matters a lot for our customers. So that took us some effort. We are doubling down on this effort to scale that further, and I think that's going to keep us busy for a while. So we are pretty vertically integrated in that we have cloud services, infrastructure, and we go all the way to building the business applications for our customers. You mentioned the application layer. In AI, you bring the platform and you build custom applications for customers. That's how it works. That is what AI allows you to do.

FDE与合作伙伴 FDEs versus partners

Host

所以这是靠一支 FDE 大军进去做的,对吧?

And so this is through an FDE army that goes in, yeah?

Arthur

是的,是 FDE,但建立在共同的底层原语之上。比如工作流管理、智能体管理、一些盒子的管理、不同业务应用部署的可观测性。所以最终,我们构建的是适合运行和构建 AI 智能体的云服务。

Yeah, it's FDE, but building on top of common primitives. So workflow management, agent management, management of some boxes, observability of the deployment of different business applications. So at the end, we are building cloud services that are adapted to run and build with AI agents.

Host

你怎么看待通过 Mistral 自己的 FDE 来做,还是通过我们传统上有的合作伙伴生态来做?

How do you think about doing it through Mistral FDEs versus a partner ecosystem, which we've traditionally had?

Arthur

在我们旅程的初期,让 FDE 站在最前线非常有帮助,因为这让我们准确理解企业在哪些地方遇到困难,他们如何从个人生产力入手。但那还不够。聊天机器人并没有为他们创造任何价值。于是我们花时间与管理者、业务单元负责人交流,理解他们需要用人工智能自动化的核心流程。我们的做法是构建业务应用、管理状态的定制应用,以及智能体端到端运行流程的应用。这对我们理解该构建哪些底层原语非常有帮助。比如我们构建了持久化流程执行。我们理解了如何以非常有效的方式编排智能体。所以让这些 FDE——我们的 FDE 直接与客户坐在一起——非常有用。从去年开始,我们做的是把工具推广给一些合作伙伴,比如埃森哲、TCS。这些系统集成商非常擅长规模化,一旦你给他们合适的工具、合适的培训,他们就能帮助扩展,因为他们有更多人。但让 Mistral 的人真正坐在客户那里敲键盘,对客户非常有用,需求很高。对我们也非常有用。这不是我们能无限扩展的事情,因为我们不可能有数万名 FDE 为我们工作。但现在我们正通过合作伙伴来扩展,两种模式并存。所以我们在很多客户账户里都有 FDE 直接与客户合作,他们在运营项目。

It has been super helpful to have our FDEs front and center at the beginning of our journey, because it allowed us to understand exactly where enterprises were struggling, how they started with individual productivity. It was not enough. Chatbots were not delivering any kind of value creation for them. And so we went and spent some time with the managers, the business unit owners, and understood the kind of core processes they needed to automate with artificial intelligence. We did it in a way where we built business applications, custom applications that are managing state, applications where agents are operating the process end to end. And it was super helpful for us to understand the primitives to build. So we built durable process execution, for instance. We understood how to orchestrate agents in a way that was very effective. So having these FDEs — our FDEs sit with customers directly — has been super useful. Now, what we have done since last year is to start propagating the tools to some of our partners, like Accenture, like TCS. So system integrators that are great at scaling things, and that once you give them the right tools, once you give them the right training, they can help scale because they just have more people. But having people actually sitting hands on keyboard from Mistral with our customers is super useful for our customers and it's in high demand. It's super useful for us. It's not something we can scale indefinitely, because we can't have tens of thousands of FDEs working for us. But it's something that now we are scaling through partners, and the two motions coexist. So we are in many accounts. We have FDEs that are working directly with customers. They're operating the projects.

收入来源 Revenue Streams

Host

他们确保交付正在进行,我们与系统集成商携手合作完成工作,然后进行维护等等。我能问一下,你们今天没有但你认为五年后最重要的收入线是什么?我在听你讲的时候,我在想,我敢打赌你刚开始时不一定能预测到这些不同的业务模式。我们第一次见面时,Arthur,我不知道你记不记得,你坐在公园长椅上和我聊生活。

They're making sure that the delivery is occurring and we work hands in hands with system integrators to do the work, to then do the maintenance and these kind of things. Can I ask you what line of revenue do you not have today that you think will be most significant in 5 years time? When I was listening to you there, I was thinking I bet you wouldn't have necessarily predicted these different motions when you started. When we first met Arthur, I don't know if you remember, but you were sitting on a park bench just chatting to me about life.

Arthur

我呃,是的,你没有投资,你本应该投资的。但太遗憾了。我非常遗憾。我昨天见到了 Jean-Charles。每次见到 Jean-Charles,我们都会谈论你。嗯,是的,这很好地提醒了我,你知道,我损失了多少钱,错过了多少机会。

I uh it yeah, you did not invest, you should have. But regret it so much. I so much. I saw Jean-Charles yesterday. And every time I see Jean-Charles, we talk about you. Um and um yeah, it's a brilliant reminder of just, you know, how much money I lost and all missed out on.

Host

对此很抱歉。现在,收入线我的意思是,我们有这两条收入线。

Sorry about that. Now, what line of revenue I mean, we have these two lines of revenue.

Arthur

我们有一条是部署平台,那是真正的许可收入,用于计费业务应用。然后我们有来自我们构建的基础设施业务的消费收入。这两件事是携手并进的,因为我们能够,嗯,我们的客户通常是为了许可模式而来,但然后他们意识到他们没有足够的 GPU,然后我们试图通过提供 GPU 或 token 服务来帮助他们。所以这两件事很好地结合在一起。

We have one which where we deploy the platform that is truly licensed revenue for billing business applications. And then we have the consumption revenue coming from the infrastructure business that we built. And the two things go hands in hands because we are able to, um, our customers they come in general for the license motion, but then they realize that they don't have enough GPUs and then we are trying to be helpful to them by providing them with either GPUs or token services. So the two things are very well combined together.

Host

你认为三年到五年内你们会有一个推理业务,一个大的推理业务吗?你认为你们会有沙盒业务吗?

Do you think you will have an inference business, a big inference business in three to five years time? Do you think you'll have a sandbox business?

Arthur

嗯,是的。如果你从这个粒度来看,我会说 GPU 方面,就像所有基于 GPU 的东西都会继续增长,因为需求是巨大的。你有我们所说的辅助服务,比如运行沙盒,非常容易地扩展和收缩沙盒,例如当你运行强化学习算法时,或者当你在同一任务上部署许多智能体时。这对我们来说是一个增长的业务,因为它来自我们客户的需求。反过来,这促使我们购买越来越多的 CPU。所以我确实预计我们所说的运行时方面,即智能体执行任务、调用工具、编译代码并运行代码的地方会增长。很明显,当你处于同步任务中,只使用 GPU 来回答对话时,你没有这些。当你开始有托管智能体在后台运行,做大量 CPU 密集型任务时,你需要 CPU,你需要更多的沙盒,你也需要更多的存储,因为它们产生大量数据。所以这两件事,我会说 CPU 相关任务、数据相关任务将是我们增长的业务线。

Uh yeah. If you look at this granularity, I'd say the GPU side like everything sitting on GPU is going to continue to grow because the demand is enormous. You have what we call ancillary services like running sandboxes, running scaling up and down sandboxes very easily when you're running reinforcement learning algorithms for instance or when you're deploying many agents on the same tasks. That is a growing business for us because it comes from the demand of our customers. And in turn that brings us to buy more and more CPUs. So I do expect what we call the runtime side of things, so where agents execute tasks, call tools, compile code and run the code to grow. It's pretty clear that when you are in synchronous task where you are only using GPUs to answer conversations, you didn't have that. When you start to have managed agents that are running on the background doing a lot of tasks that are CPU intensive, you need CPUs, you need more sandboxes and you need more storage as well because they produce a lot of data. So those two things, I would say CPU related tasks, data related tasks are going to be growing line of businesses for us.

向政府销售 Selling to Governments

Host

业务的一大部分是向政府销售。关于向政府销售,有什么是没人知道但每个人都应该知道的?

A big part of the business is selling to governments. What does no one know about selling to governments that everyone should know?

Arthur

首先,一般来说,没有向政府销售这回事。所有政府都非常不同。一些政府足够小,看起来或多或少像企业销售模式。例如,在部委里每个人都互相认识的政府就是这种情况。较大国家的其他政府则稍微复杂一些。所以再次,寻找创造价值的地方、理解客户需求的复杂性随着国家规模而增长,我应该这么说。

And the first thing is that there's no such thing as selling to governments in general. All governments are very different. Some governments are small enough so that it looks more or less like an enterprise sales motion. It's the case of governments where everybody in ministries know each other, for instance. Other governments in larger countries are slightly more complex. So again, the complexity of finding where to create value, of understanding what the customer needs grow with the size of the country, I should say.

推理市场 Inference Market

Host

你能帮我了解推理市场吗?这是我完全了解不够的东西。就像我现在知道但希望 12 个月前就知道的是,推理的需求是无限的。深度令人惊讶。推理市场会发生什么?

Can you help me on the inference market? It's something that I just totally didn't know enough about. Like what I know now that I wish I'd known 12 months ago was that there is just infinite demand for inference. It's the depth is astonishing. What happens with the inference market?

Arthur

我的意思是,推理市场的表现非常像商品市场,目前利润率相当高,因为如果你有供应,如果你对某些类型的模型有正确的服务质量,你可以按需出售,因为你有大量直接来自未承诺客户的需求。然后对于足够成熟的客户,他们需要保证服务质量。他们需要服务级别协议。在那里他们需要预留物理容量。这是我们可以提供的,因为我们有正确的预测模型和正确的供应。所以你有这两件事。你有按需的部分,它非常像商品市场,定价很重要。然后你有市场的溢价部分,客户实际上希望获得延迟保证,你希望获得一定数量请求的生成速度保证。他们越能精确地描述他们的需求形状,我们的定价就越好。我们就能更好地塑造我们自己的基础设施,以确保他们以合适的价格获得正确的服务质量。

I mean, the inference market behaves very much like commodity market, which is pretty high margin today, in that if you have supply, if you have the right quality of service for certain kind of models, you can sell them on demand because you have a lot of demand coming directly from customers that are non-committed. And then for customers that are sufficiently mature, they need to have guaranteed quality of services. They need to have service level agreements. And there they need to have reserved physical capacity. And that's something that we can provide because we have the right forecasting models and we have the right supply. So you have these two things. You have one which is on demand, which very much behaves as a commodity market where pricing matters a lot. And then you have let's say a premium side of the market where the customers actually want to have access to guarantees on the latency, you want guarantees on the speed of generation for a certain number of requests. And the more they can be precise about the shape of their demand, the better our pricing can be. The better we can shape our own infrastructure to make sure that they have the right quality of service for the right price.

新实验室竞争 Neo Lab Competition

Host

我能问一下,我们有这么多不同的 Neo Lab 提供商。已经创建了 102 个不同的 Neo Lab 提供商。每周都有另一个 DeepMind 分拆公司,又一个,这么多。什么将区分繁荣的 Neo Lab 和那些不繁荣的?

Can I ask you, we have so many different Neo Lab providers. There've been 102 different Neo Lab providers created. Every week there's another DeepMind spin out and another and so many. What will separate Neo Labs that thrive and those that don't?

Arthur

嗯,首先,我的意思是我们自己也是作为 Neo Lab 起步的。2023 年时不是这么叫的,但实际上,你知道,当你招聘时,当你开始创业时,你有一堆银弹来雇佣最优秀的人,我们用尽了所有银弹来雇佣研究人员。所以实际上我们开始是一家研究公司。我认为让我们能够扩展并现在以更广泛的方式运营的,仍然是这第一支柱,即交付模型,这就是为什么 Star Large Four 对我们非常成功。我认为即将发布的新版本也非常令人兴奋,因为我们正在取得相当好的进展。成功的是我们早期就专注于会扩展的市场。所以我们选择了企业市场,这带来了一定的复杂性,因为当然企业客户要求很高,这是应该的,我们已经调整了产品来满足这些需求。但我们能够在研究层面、最高水平上运作,并能够不断招聘和进一步投资于制造越来越强大的模型,以及我们建立了以客户为中心的模式,理解客户的问题,这些让我们得以扩展。这是每个 Neo Lab 在开始建立业务时都需要考虑的事情。

Well, the first thing, I mean we started as a Neo Lab ourselves. It wasn't called that way in 2023 but effectively, you know, when you hire when you start with a when you start a company, you have a bunch of silver bullets to hire the best and we used all of our silver bullets to hire researchers. And so effectively we started as a research company. I think what has allowed us to scale and to now operate in a much wider way still with this number one pillar which is to deliver models and that's why Star Large Four has been very successful for us. And I think it's very exciting as well in the new releases that are coming because we're making pretty good progress. What has been successful is that we focused early on on the market that would scale. So we went for the enterprise market which comes with a certain number of complexity because of course enterprises customers are demanding as they should be and we have adapted our product to meet those demands. But the fact that we were able to both operate at the research level, at the highest level and able to constantly recruit and invest further in making stronger and stronger models and the fact that we built a customer-centric motion where we understood the problems of our customers have allowed us to scale. And that's something that every Neo Lab needs to think about when they start building their business.

欧洲的人才与算力瓶颈 Europe's Talent and Compute Bottleneck

Host

但欧洲到底有多少人真正有能力创建这样的公司,并且像你一样经历过、见证过,比如我不知道,也许像 Yann LeCun 那样。是 20 个?还是 100 个?

But how many people are there actually in Europe who truly have the ability to create these companies and have been and seen it like you have, like I don't know, maybe a Yann LeCun has done. Is it 20? Is it 100?

Arthur

我的意思是,我不是风险投资人,所以你可能比我更清楚。但总的来说,欧洲有很多人才。你有很多创业人才,很多研究员人才,而且这两者的结合相当不错。所以你在欧洲看到很多新的实验室。现在难的是,这不仅仅关乎才华,也不仅仅关乎能销售、有东西能成功创业。今天还关乎能否获得算力。我认为现在许多我认识的新实验室面临的一大瓶颈是,要获得 GPU 来在某些垂直领域做有趣的研究很难,因为供应极度紧张。所以在我们拥有的供应中,我们为某些新实验室保留了一些容量,因为我们相信它们能成长,我们也能与它们共同成长。但实际上,如果你今天是一个创始人,想做非常研究导向的事情,你最大的问题是以合理的价格获得足够的算力。

I mean, I'm not a VC investor, so I would probably know it better than I do. But there's a lot of talent in Europe generally. You have a lot of entrepreneurial talent. You have a lot of researcher talent and you have a pretty good combination of those two. So that's why you see a lot of new labs in Europe. Now the hard thing is it's not only about being talented. It's not only about being able to sell and having something that you can successfully create a business with. It's also today about being able to access compute. I think a big bottleneck today for many new labs that I know is that accessing GPUs to actually do interesting research in certain verticals is hard because there's an enormous supply crunch. And so that's why in the supply we have, we are reserving some capacity for certain new labs because we believe that they can grow and that we can grow with them. But effectively if you're a founder today and you want to do something which is very research centric, the biggest problem you have is to get access to enough compute with a decent price.

研究员招聘市场 Researcher Hiring Market

Host

我能问你吗,你刚才提到了研究员这个词。现在研究员招聘市场有多疯狂?

Can I ask you, you've mentioned the word researchers there. How crazy is the researcher hiring market today?

Arthur

嗯,我得说竞争相当激烈。再说一次,如果你有正确的使命,有正确的薪酬结构,在做对世界重要的事情,人们会加入我们,加入你。我想说,我们非常感谢所有员工为 Mistral 的成功所做的奉献。Mistral Large 4 的质量很成功,今天有很多关注,在开放权重类别、编程和网络防御方面领先。这仅仅是因为这个团队在过去三四个月的奉献和韧性,因为我们最近发布了技术报告。通常模型第一次不会成功,所以你需要能够调整方向,比如好吧,你要再花两个月才能真正让它工作,这会很痛苦。我们期望发布,但需要稍晚一点发布。这对每个人来说都会很痛苦,但每个人都需要全力以赴做正确的事。我的意思是,鉴于我们资源的稀缺,我非常感激团队的韧性。

Well, it's pretty competitive, I should say. Again, if you have the right mission, if you have the right compensation structure, if you're doing something that matters for the world, people will join us, will join you. And I would say we're very grateful for the dedication of all of our employees to making Mistral successful. The quality of the Mistral Large 4 has been a success, has a lot of traction today and is leading the pack in open weights in the open weight category and coding and in cyber defense. It's only due to the dedication of this team in the last three, four months, the resilience as well because when you run a model we've posted our tech report recently. Often times it doesn't work the first time and so you need to be able to course correct in a way where okay, you're going to spend two more months to actually make it work and it's going to be painful. We're expecting to release but we need to release a little later. It's going to be painful for everyone but everyone needs to be all hands on deck and do the right thing. I mean, I'm very grateful for the resilience of the team given the scarcity of resources that we have.

训练运行与算力稀缺 Training Runs and Compute Scarcity

Host

我刚刚请了微软的 Satya 上节目,他谈到了 Mustafa 之前的声明,说未来几年我们会看到千亿美元级别的训练运行。你觉得这有点夸张,还是实际上完全可达?

I just had Satya from Microsoft on the show and he talked about Mustafa's statement before that we'll see a hundred billion dollar training runs in the next few years. Do you think that's a little bit hyperbolic or is that actually well within reach?

Arthur

我得先说数据世界,在某个时候,如果你想让模型在特定领域非常强,比如芯片设计,或者像我们一样在操作工业工程工作流方面非常强,你确实需要有合适的环境来训练它们,这部分需要一些技巧和品味。这不是像堆算力那样容易扩展的。所以我想说,我们今天都遇到的瓶颈实际上是能够获取足够的高质量数据。是算力。我们需要更多算力。我们正在部署更多算力。达到千亿美元级别的训练运行,我认为显然不是明天的事。

You do have a data world before I should say in that at some point if you want to make your model very strong at certain domains like chip design for instance or if you want to make it very strong at operating industrial engineering workflows like we do, you do need to have the right environments to train them and this part takes some skill and taste. It's not something that you can scale as easily as just stacking compute. So I'd say the bottleneck that we're all running into today is effectively to be able to source enough quality data. It's compute. We need more compute. We're deploying more compute. Getting to a hundred billion of training runs I think is not for tomorrow clearly.

市场反应与垂直整合 Market Reaction and Vertical Integration

Host

如果每个人都这么缺算力,为什么 Nebius 跌了 40%?为什么 CoreWeave 跌了 40%?如果大家都这么缺算力,为什么 AI 交易在公开市场出现裂痕?

If everyone's so short on compute, why is Nebius 40% down? Why is CoreWeave 40% down? Why is the AI trade cracking in public markets if it's so clear that everyone is so compute starved?

Arthur

我不太清楚今天市场是如何反应的。我们看到的是,鉴于我们拥有的抵押品质量,在收入承诺、承购付款承诺方面,我们完全能够为自己的算力融资。我们公司作为垂直整合的公司运营。所以我们建设算力,但为了更高层次的目的,为服务企业、构建业务应用。所以我们锚定在客户的价值创造上,这意味着在融资层面,我们不必在融资上大幅杠杆化公司。所以这给了我们对未来很大的信心。但实际上,就像任何工业建设一样,诱惑是获得大量杠杆,但杠杆在某个时候会断裂。所以这就是为什么我们很高兴垂直整合,从业务应用创建一直运营到基础设施部分。

It's a little unclear to me how markets are reacting today. What we see is that we're fully able to finance our own compute given the quality of the collateral that we have in terms of revenue commitment, take up pay commitments. We as a company operate as a vertically integrated company. So we build compute but for higher level purposes for serving the enterprises for building business applications. So we are anchored on the value creation of our customers which means that at the financing level we haven't had to go and leverage enormously the company on that financing. And so that gives us a lot of confidence in what's going to happen. But effectively as in any industrial build out it's tempting to get a lot of leverage but at some point leverage break. And so that's why we're very happy to be vertically integrated and to operate from business application creation all the way to the infrastructure part.

聚焦与企业战略 Focus and Enterprise Strategy

Host

有了专注的好处,你们没做什么,事后你会想,嗯,我们是不是应该做那个?

What did you not do with the benefit of focus that with hindsight you go mhm I wonder if we should have done that?

Arthur

我们没有深入消费市场,但我认为我们很庆幸没有这样做。我们没有进入廉价设计。我认为那是

We did not go deep into the consumer market but I think we're very grateful that we did not do it. We did not go into cheap design. I think that's

Host

好吧,我们暂停一下。为什么你为什么庆幸没有进入消费市场?

Well we'll pause. Like why why do you feel grateful you didn't go into the consumer market?

Arthur

我认为这是一个非常拥挤的市场,很难在企业最高水平运营,同时还要经营本质上将是广告业务的东西。所以那部分,专注于企业领域的决定,以及专注于我们决定关注的四个垂直领域——制造业、金融服务、国防、公共部门——对公司的专注非常关键。因为人工智能的问题在于它可以做一切,你需要一个专注的来源,所以我们选择的专注来源来自市场。所以我们希望拥有服务于最佳代理工程工作流、最佳财务决策、最佳国防目的以及与州主权相关的所有事情的模型。

I think it's a very packed market and it's very hard to operate at the highest level with enterprises and also to run what is inherently going to be an ads business. So that part that decision to focus on the enterprise segment also that decision to focus on the four verticals that we have decided to focus on, manufacturing, financial services, defense, public sector has been very critical to focus the company. Because the problem with artificial intelligence is that it can do everything and you need a source of focus and so the source of focus we've picked is comes from the market. So we want to have the models that serve the best agency engineering workflows, that serve the best financial decision making, that serves the best defense purposes and everything related to state sovereignty as well.

递归自我改进与聚焦 Recursive Self-Improvement and Focus

Host

有了 RSI,你不需要一个专注的来源吗?那不是

With with RSI, do you need a source of focus? Is that not

Arthur

我的意思是,使用 AI 来改进 AI 模型确实也需要专注,因为你用于解决问题的资源有限。最终这里的瓶颈确实是算力。瓶颈还在于,你所做的任何递归自我改进或只是自我改进——因为意思相同——都需要有明确定义的问题。如果你今天用人工智能有一个明确定义的问题,你可以解决它。在企业中,许多情况下问题并没有明确定义,所以经济增长的真正摩擦不是我们无法解决明确定义的难题,而是很难解决模糊的问题,你有模糊的目标,有人类在循环中,而那部分你无法真正用自我改进来解决。

I mean using AI to improve AI models do require to focus as well because you have limited amount of resources to spend on problems. At the end the bottleneck here is indeed compute. The bottleneck is also that everything that you do that which is recursively self-improving or just self-improving because it means the same thing requires to have well-defined problems. If you have a well-defined problem today with artificial intelligence, you can lay the climate. In many cases in enterprises, the problems are not well-defined and so the true friction to economic growth is not the fact that we cannot solve hard problems that are well-defined is that we it's hard to solve ambiguous problems where you have ambiguous objectives, where you have humans in the loop and that part you cannot really solve with self-improvement.

部署与企业瓶颈 Deployment and Enterprise Bottlenecks

Arthur

你只能靠模型被部署、与人类互动来解决,而这正是我们专注的部分。不过我想说,我们也在大量投入,在定义明确的问题上构建自我改进的循环,但真正制约企业里 AI 使用爆发的,恰恰是那些定义不清的问题。你想在这些问题上部署 AI,却无从下手,因为问题本身的定义就不对,所以你得花时间。这就是为什么需要前置设计(FD),这就是为什么我们有前置设计,去尝试更好地定义问题。

You can only solve with models being deployed and interacting with humans, and that's the part that we are focusing on. I would say though, we are investing a lot in making self-improving loops on well-defined problems as well, but that is the part that is constraining the explosion of AI usage in the enterprise. The poorly defined problems on which you want to deploy AI, but you have no way to lay the climb because the definition of the problem is not proper, and so you need to spend some time. That's why you have FDs, that's why we have FDs, to try and better define the problem.

Host

Arthur,我得跟你直说。我开场时开玩笑说你是欧洲最后的希望。这里有一个——我不是要特别夸张——但在想成为下一个 Mistral 的 Neo Labs 之间,你和其他人之间存在一道巨大的鸿沟。你和团队感受到压力了吗?

Arthur, I've got to be blunt with you. I joked at the beginning that you're like Europe's last hope. There is a—and I don't mean that super dramatically—but there is a very big chasm between you and anyone else in terms of Neo Labs that will aspire to be the next Mistral. Do you and the team feel the pressure?

Arthur

我们确实感受到一些压力,但就像这个领域的任何人一样,这里增长快、变化快。作为 AI 创业者,你必须在模型侧交付成果,因为那是唯一能展示你进展的东西。然后企业需要看到价值,所以你得有真正理解业务问题的人。所以这是相当不同的问题要解决。我们确实感受到一些压力,但我认为这是非常健康的压力。这是时间上、历史上非常重要的时刻。我们需要满足公众舆论的期待。归根结底,我们需要证明 AI 是有用的。我们所做的部署、在市场上制造的声量——这项技术需要以正面技术的形象出现。我们需要展示这项技术的正面性。而做到这一点的唯一方式,就是把主权 AI 作为一个品类来推广,确保它创造经济增长和人人可及的机会,确保 AI 的收益得到良好分配。所以这当然会带来一些压力,因为这不是理所当然的。如果我们不成功,我不知道世界会怎样。所以我们确实有角色要扮演。我认为我们承担的使命很重要。这是压力,但也相当令人兴奋。

We do feel some pressure, but as anyone in the field, this is fast growing, fast moving. You need to—as an entrepreneur in AI, you need to deliver on the model side because that's the one thing that shows the progress you're making. Then enterprises need to see the value, so you need to have people that are understanding the true business problem. So it's a fairly different problem to solve. So we do feel some pressure, but I think it's a very healthy pressure. It's a very important moment in time and in history. We need to be meeting the expectations of the general opinion. At the end, we need to show that AI is useful. The kind of deployment we're making, the kind of noise this is making in the market—this technology needs to appear as a positive technology. We need to show the positivity of that technology. And the only way in order to do so is to promote sovereign AI as a category, is to make sure that that creates economic growth and opportunities for everyone, to make sure that the gains of AI is well distributed. And so of course it creates some pressure because it's not a given. If we don't succeed, I don't know what's going to happen in the world. So we do have a role to play. I think the mission we play is important. And I'm—it's pressure, but it's also quite exciting.

AI与GDP增长 AI and GDP Growth

Host

AI 会让 GDP 增长每年提升超过 2% 吗?

Will AI inflate GDP growth by more than 2% a year?

Arthur

嗯,这就是整个问题的关键。你要去解决什么样的瓶颈,才能真正推高 GDP?这就是我们非常看好工业工程的原因。归根结底,想要 GDP 增长,你需要更多生产。你不需要更多智能,你需要更多生产。要获得更多生产,你需要更好的工厂。你需要更高效的物体。你需要更好的飞机,用同样的油飞得更久。你需要更好的能源生产,更高效、碳强度更低。你需要更好的 ASML 机器。你也需要更好的服务,包括金融服务。所以尤其在工业工程中,使用 AI 能让你更好地设计物体。比如,如果你搭建智能体式循环来设计更安全的汽车,你的汽车设计团队就能快得多,在产品上迭代更快。这意味着你能卖得更多,意味着你能卖得更便宜,意味着你在造更安全的产品。在航空航天业,如果你设计出更好的飞机,它会在天上飞 40 年。如果它飞得更快、油耗更低,那就会创造 GDP 增长。所以归根结底,真正重要的是让 AI 产生物理效应。而这种效应我们主要见于制造业和金融服务。

Well, that's the entire question. What kind of bottleneck do you go and solve to actually inflate that GDP? That's the reason why we were very bullish on industrial engineering. At the end, if you want GDP growth, you need more production. You don't need more intelligence. You need more production. In order to have more production, you need better factories. You need objects that are more efficient. You need better planes that are flying for longer with the same amount of oil. You need better energy production that are more efficient, less carbon intensive. You need better ASML machines. You need better services also in financial services. So in industrial engineering in particular, using AI allows you to do better design of objects. So if you set up the agentic loop to design safer cars for instance, you can have your car design team go much faster and iterate faster on that product. That means that you can sell more, that means that you can sell more cheaply. That means that you're building safer products. In the aerospace industry, if you design a better plane, it's going to be out there for 40 years. If it runs faster with less fuel, then it's going to create GDP growth. So at the end what truly matters is to enable AI to have a physical effect. And this effect we find mostly in manufacturing and financial services.

快问快答 Quickfire Round

Host

兄弟,我注意时间。我要跟你来个快问快答,好吗?来吧。你今天相信什么、而你周围大多数人不会同意?

Dude, I'm aware of time. I'm going to do a quickfire round with you, okay? Let's do it. What do you believe today that most around you would disagree with?

Arthur

我认为我们坚信这是一个正在碎片化的市场,而主权 AI 将成为企业 AI 部署中最重要的品类。因为当企业部署这项技术时,业务连续性、成本以及对数据和知识产权的控制,将是唯一重要的东西。所以我认为这是我们非常坚定的信念,它被许多企业认同,也被市场的很大一部分认同,但当然不被我们的一些竞争对手认同。

I think we strongly believe that this is a fragmenting market and that sovereign AI is going to be the most important category of AI deployment in the enterprise. In that when enterprises are going to deploy the technology, business continuity, cost and control over their data and IP is going to be the one thing that matters. So I think that's a very strong conviction that we have, that is shared by a lot of enterprises and shared by a significant part of the market, but of course not shared with some of our competitors.

Host

今天有什么极其疯狂、但五年后关于我们生活和工作方式会变得非常普遍的预测?

What is an extremely wild prediction today that will be very common in five years' time about the way that we live, the way that we work?

Arthur

我认为我们作为工程师的工作方式——对软件工程师来说已经改变了——将对任何类型的工程师彻底改变。所以我对这种借助人工智能、以相当自主的方式设计更好产品的方法超级兴奋。所以这不涉及所有人,但每个有技术背景的人——不只是最先受影响的软件工程师和数学家——每个有技术背景的人都会以某种方式操作 AI 智能体,从而获得巨大的杠杆。所以这本身将是劳动力市场运作方式的一次重大转变。

I think the way we work as engineers—as already changed for software engineers—is going to completely change for any kind of engineers. So I'm super excited by that approach of designing better products in a way which is fairly autonomous thanks to artificial intelligence. So it does not concern everybody, but everyone with a technical background—and not only the software engineers that were the first affected and the mathematicians—everyone with a technical background will be operating AI agents in a way that will give them enormous leverage. So this in itself is going to be a major shift in the way the labor market is actually working.

Host

上市这件事你打算怎么办?因为总有一天你会想上市的,对吧?但你在欧洲没法上市。

What do you do about going public? Because one day you will want to, okay? But you can't go public in Europe.

Arthur

完全可以在欧洲上市。

Can absolutely go public in Europe.

Host

嗯,首先——对,你得——你会在法国上市吗?不会。

Well, the first thing—yeah, you need to be—are you going to go public in France? No.

Arthur

十年前曾经成立的事,我认为如今已不再成立。全球投资者对投资欧洲有巨大的胃口,无论是私募股权侧还是公开股权侧。所以过去存在的瓶颈已不复存在。我们 Mistral 当然在准备——当你上市时会面临很多运营约束。我们也在推进并交付需要达成的商业里程碑,但我不认为这再是瓶颈了,而且有很多非常成功的公司,其中一些利益相关方首先在欧洲上市。所以没有理由不在某个时点去欧洲做 IPO。

What used to be true 10 years ago I think is no longer true. There's a huge appetite from investors from across the world to invest in Europe, both on the private equity side but also on the public equity side. And so the bottlenecks that used to be there are no longer there. So we at Mistral of course getting prepared—you have a lot of operational constraints when you go and become public. We're also pushing and delivering on the commercial milestones that we need to deliver on, but I don't think this is anymore a bottleneck, and there are widely successful companies. Some of them are stakeholders that are firstly listed in Europe. So there is no reason why not to go and do an IPO in Europe at some point.

Host

最后一个问题,兄弟。展望未来五年、七年、十年,你最兴奋、而没多少人会想到的是什么?

Final one for you, dude. What are you most excited about when you look forward to the next five, seven, 10 years that not many people think about?

Arthur

我相当兴奋的是,AI 将如何赋能那些我们对世界缺乏理解的领域的基础研究。比如在物理学中,我们能够以从未达到的规模运行 AI 模拟——那就是我们收购的、名为 Emi 的团队,他们专精于此。

I'm quite excited by how AI is going to enable foundational research in domains where we lack understanding of the world. So in physics for instance, the fact that we can both run AI simulations at a scale that was nowhere reached—that's the team we bought called Emi, specialist in that.

开场致辞 Opening remarks

Host

你知道吗,Arthur?我怪你。我怪你。那时候你坐在长椅上,一句话都不说,不像你现在这样能说。当年的 Arthur 和今天的 Arthur,恕我直言,简直判若两人。看到你成为这样的领导者,真的很了不起。说实话,兄弟,我对你充满敬意、钦佩和感激,感谢你为欧洲所做的一切。但哇,这真的很了不起。

Do you know what, Arthur? I blame you. I blame you. When you sat on the bench and you didn't speak at all, like you speak now. The Arthur then and the Arthur today are just paradoxically different, with respect. It's amazing to see the leader that you are and you've become. I honestly, dude, I have so much respect, admiration, gratitude for you and all you've done for Europe. But wow, it's amazing.

Arthur

谢谢你的夸奖。还有很多事情需要我们去交付,但我们对已经取得的进展相当满意。我认为这个使命很重要,所以如果有观众想加入我们的旅程,请务必联系我。

Thank you for the praises. Still a lot of things we need to deliver on, but we're quite happy with the progress we've done. I think the mission is important, and so if anyone watching us wants to join us in that journey, please, please contact me.

用模拟器理解世界 Understanding the world through simulators

Arthur

它将促成新的发现。我们能够构建模拟器,并让智能体操作这些模拟器来更好地理解世界,这将成为理解世界的巨大推动力。我出身于数学和物理背景,所以对下一代研究者面前的机会感到非常兴奋。

It is going to enable new discoveries. The fact that we can build simulators and that we can have agents operating those simulators to understand the world better is going to be an enormous enabler of the understanding of the world. I come from a math and physics background, so I'm quite excited by the opportunity ahead of us in the next generation of researchers.

互动版:逐字朗读 + 针对本期提问 →