Mistral CEO 谈构建欧洲 AI 基础设施

那么，我的问题是，你觉得什么早餐比较合适？我今天在伦敦车站站着吃了一个我只能形容为最疯狂的三明治。像佛卡夏面包，里面塞满了各种东西，味道很棒。问题是，它太 messy 了，站着等火车的时候吃最不合适。太 messy 了。我为什么等火车？因为我需要坐那趟火车来巴黎，这是第一次 Tech Download 在路上。我现在在巴黎，即将采访 Mistral 的 CEO Arthur Mensch。如果你不了解 Mistral，他们基本上是欧洲版的 OpenAI 和 Anthropic。他们制作各种 AI 模型，有些是开源或开放权重的，并且也进行商业化。他们非常专注于企业，也就是大企业将 AI 整合到业务中。他们的估值超过 130 亿美元。公司正在努力增加收入。在开始之前，我想提几个术语，因为我们会深入讨论。第一个是算力。你可能在谈论 AI 时听说过。它实际上指的是计算能力，以数据中心的形式运行芯片，为所有 AI 提供动力。现在场景就是这样。我很期待与 Mistral 的 CEO Arthur Mensch 进行对话。Arthur，非常感谢你参加 Tech Download。

So, my question is, what would you consider an appropriate breakfast food? I was standing in the station in London today, and I had what I can only describe as the most insane sandwich. It was like focaccia bread. It was filled with all sorts of things, and it was amazing. Problem was, it was so messy and the most inappropriate thing to eat while standing around waiting for a train. So messy. Why was I waiting for a train? Well, it's because I needed to take that train to come to Paris because this is the first ever Tech Download on the Road. I am here in Paris. I'm about to speak to Arthur Mensch, who's the CEO of Mistral. Now, if you don't know Mistral, they're basically Europe's answer to OpenAI and Anthropic. They make various AI models. Some of them are open-source or open weight, and they commercialize these as well. And they're very much focused on the enterprise, i.e., big businesses integrating AI into those businesses. Their valuation sits somewhere above $13 billion as well. And the company is racing to try to grow revenue. And before we get into this, there are a couple of terms I want to put on your radar because we're going to speak into them. We're going to get deep into those terms. The first one is compute. Now, you may have heard this when talking about AI. And effectively, what it refers to is computing power in the form of these data centers that are running chips that are powering and running all of this AI. So, let's set the scene. That's it right now. And I'm so excited to dig into this conversation with Arthur Mensch, the CEO of Mistral. Arthur, thanks so much for joining me on the Tech Download.

谢谢你邀请我。

Thank you for having me.

那么，Arthur，我想先从算力开始聊。我想为我们的观众和听众设定一下背景。你承诺投资 40 亿欧元在法国和瑞典建设数据中心。容量方面，目标是到 2027 年达到 200 兆瓦，到 2030 年达到 1 吉瓦。而且你现在宣布了一个专门用于推理的新站点。你能详细介绍一下这个新站点具体是什么，以及它如何融入你构建计算基础设施的更广泛计划吗？

So, Arthur, I want to kick off first by talking about compute. And I just want to set the context for our viewers and our listeners as well. You've committed, I think, 4 billion euros to investment in data centers across France and Sweden. I think capacity-wise, it was 200 megawatts was the capacity aimed by 2027, a gigawatt by 2030. And you've now announced a new site specifically for inferencing as well. Can you just run us through exactly what that new site entails and how it fits into your broader plans here around building out this computing infrastructure?

所以，我们的新站点是一个高可用性站点。我们将用它来服务我们的客户。我们的客户需要 token，目前实际上供应短缺。因此，我们一直在努力在 2026 年为他们建设更多容量，到 2027 年将大幅增加。我们为什么这么做？因为我们这个行业本质上就是把电子转化为 token。要做到这一点，你需要训练模型。我们的一切都建立在开源模型之上，这对客户来说更容易，我认为也更健康。但一旦有了模型，你需要把它们放到 GPU 上。我们通过几个月前收购 Koyeb 积累了高效服务 GPU 并将其转化为 token 生成器的专业知识，并通过基础设施投资来实现。欧洲在基础设施建设方面落后了。所以我们正在投资以缩小差距。这就是新站点，位于法国。所以，我应该说，这是低碳 token。它将用于服务我们的客户，包括我们的 studio 产品、公有云服务和私有云服务。

So, the new site that we have is a high availability site. So, we're going to be using it to serve our customers. Our customers are in need of tokens. It's actually in short supply at the moment. And so, we've been at work building more capacity for them in 2026 and they will be much more in 2027. So, why do we do that? Because inherently the business we're in is about transforming electrons into tokens. In order to do that, you need to train the models. We're building everything on top of open-source models, which makes it easier and I think more healthy for our customers. But then once you have the models, you need to put them on GPUs. And we've built that expertise on serving efficiently GPUs and turning them into token generators through the acquisition of Koyeb that we've talked about together a couple of months ago. And through the investments that we are doing on the infrastructure. Europe is lagging behind when it comes to the build-out of infrastructure. And so, we are investing to close that gap. So, that's a new site. It's based in France. So, it's low-carbon tokens, I should say. And it's going to be there to serve our customers and to both with our studio offering and public cloud services and private cloud services.

还有其他人采取了从别处租用容量的方式。现在建设非常昂贵的基础设施有什么必要呢？

And there are others who have taken the approach of perhaps renting capacity from elsewhere. What's the need here to build out what is very expensive infrastructure at this point?

嗯，你知道，我们有很多客户把他们对 token 的依赖视为一个关键的供应链问题。在欧洲，我们是一家欧洲公司。作为欧洲公司，当我们建设欧洲的容量时，我们可以提供完全由我们控制的 API。因此，与其他供应商脱钩对我们的国防客户和制造业客户来说非常重要。这就是我们决定投资的原因：从产品角度来看，在欧洲拥有全栈极具吸引力。更重要的是，因为我们构建的软件栈可以适配任何硬件，我们也将它带到亚洲。我们宣布了与 Singtel 的合作，实际上是将我们的公有云平台放在他们正在构建的硬件之上，使他们成为全栈 AI 提供商。我们致力于构建有韧性的全栈 AI 提供商，在数字服务方面不依赖外国实体。因此，当我们拥有算力时，这项投资对我们很重要，因为它使我们能够提供大型产品。现在，我们也租用算力。我们与世界其他地区的超大规模提供商合作，但在欧洲，我们的客户要求我们真正拥有全栈。

Well, you know, we have a few many customers, I should say, that look at their dependency to tokens as a critical supply chain problem. And one of the things that we have in Europe is that we're a European company, I should say. And as a European company, when we build European capacity, we can actually serve APIs that are fully under our control. And so that decoupling from other providers is actually quite important for our defense customers, for our manufacturing customers. So that's the reason why we really decided to invest is that from a product perspective, owning the full stack in Europe is extremely attractive. And what's more is that because we build that software stack that can fit on any hardware, we are also bringing it to Asia. We've announced with Singtel the partnership where we're actually bringing our public cloud platform on top of the hardware they're building so that they become a full stack AI provider. So we're all about building full stack AI providers that are resilient, that are not dependent on foreign entities when it comes to the digital services. And so that investment when we own the compute is important for us because that allows to have a big product offering. Now, we also rent compute. We work with hyperscalers in other parts of the world, but in Europe our customers are asking us to actually own the full stack.

那么，请给我讲讲你在算力建设方面的策略。显然，你自己也在使用。你在自己建设的一些算力上训练模型，也从别人那里租用等等。你现在建设的算力，是专门为你合作的一些客户准备的，比如 ASML，或者你提到的 Singtel？还是说，你也可以把它租给其他不一定与你合作的客户？我指的是其他实验室。

So just walk me through the strategy in terms of the compute that you're building. Obviously you're using it for yourselves. You've trained your models on some of the compute you've built, renting it from others, etc. The compute you're building now, is this specifically for some of the customers you work with, you know, ASML for example, or Singtel as you mentioned? Or is the view also you know, you could rent this out to other kind of customers who aren't necessarily you're working with? I'm talking other labs for example.

实际上有两类。一方面，当我们与企业客户合作时，他们并不真正关心 GPU 本身。他们关心的是在上面销售服务。所以是那些高价值服务，可以部署智能体，当然可以生成 token，然后使用 token 调用工具，记录系统产生的数据。我们利用这些数据训练更高效的新模型。因此，企业客户不想看到基础设施。他们想要无服务器产品，只想要一套全面的工具，组合起来可以构建业务应用。这就是他们需要的，也是我们满足他们需求的方式。另外，还有一些 AI 实验室和公司，他们可能在使用自己的数据飞轮、用自己的数据构建模型方面相当先进，他们可能实际上想要获得算力本身。

So there's really the two categories. So on one side when we work with enterprises among our customers, they don't really care about the GPUs themselves. What they care is to sell services on top. So the high value services that allows to deploy the agents, that allows to connect of course generate the tokens, but then use the tokens to call tools, record what the systems are producing in terms of data. We use that data to train new models that are going to be more efficient. And so enterprise customers, they don't want to see the infrastructure. They want to have serverless offerings and they just want like a comprehensive suite of tools that when stitched together allows to build business applications. So that's what they need and that's how we address their needs. Now you also have AI labs and companies that are maybe quite advanced in using their own data flywheel, using their own data to build their own models and those may actually want to get the compute itself.

所以我们确实通过托管 Kubernetes 向 AI 实验室和一些客户提供 GPU 即服务，这些客户的部分研发团队实际上在进行 AI 研究。所以这两样都有，但对企业而言，我们提供的是高价值服务和 token 生成。

And so we do offer GPU as a service so through managed Kubernetes to AI labs and to some of our customers that have part of their R&D teams that are actually doing AI research. So the two things are available but when it comes to enterprises, high value services and token generation is the thing that we provide.

大家好，我现在回到 CNBC 伦敦办公室的播客录音室。随着你们越来越多地收听 Tech Download，你们会听到我在各集节目中多次插入，因为这些时刻能帮助我补充一些对话的背景。我想谈的第一个术语是“tokens”。你们在这一集里听到了，未来很多集可能也会听到。Tokens 实际上是 AI 模型处理的基本数据单元。当你向聊天机器人（比如 Gemini、ChatGPT 或 Claude）输入查询时，它会被分解成 tokens，AI 处理这些 tokens，然后你得到输出。请求越长，使用的 tokens 就越多。这对 AI 公司来说有两个作用：一是衡量服务使用量，二是据此收费。对话的一部分涉及主权问题，对吧？在欧洲，以及建设欧洲基础设施。那么，有没有可能美国实验室，比如 OpenAI 和 Anthropic，成为你们算力的客户？

So everyone, I'm back in our podcast studio in CNBC's London offices now and you're going to hear me jumping in throughout episodes a couple of times as you listen more and more to the tech download because these are moments for me to help flesh out some of the contexts around the conversations I'm having. The first term I want to pick up on is something called tokens. So you've heard that in this episode, you're likely going to hear it in many more episodes going forward but tokens are effectively the basic data unit that AI models process. So whenever you put your query into a chatbot, something like Gemini or ChatGPT or Claude, it's broken down into tokens that the AI can then process and then of course you get your output. Then the chatbot gives you that output but also the longer the request, the more tokens that are effectively used and this is a way for AI companies to do a couple things. One is to measure how much their service is being used and then secondly to charge for that service as well. And part of the conversation and we'll get on to this is around sovereignty, right? Here in Europe and building out European infrastructure as well. So is there a world in which US labs, I'm talking OpenAI and Anthropic could be customers of yours for your compute?

当然。AI 实验室非常需要算力，我们有一些，有些实验室实际上在向我们要求大量算力。目前，我们需要优先分配访问权限，所以我们把算力给了一些 AI 实验室，但更重要的是，我们优先考虑那些使用量激增的客户，因为他们正在转向后台运行的系统以及生成更多 token 的智能体。因此，我们正在努力建设尽可能多的容量，以满足大量客户对不同用途算力的需求。

Absolutely. I mean the AI labs are in sore need of compute and we have some of it and some of them are actually asking us for a lot of compute. Today, we actually need to prioritize the access and so we're giving it to some of the AI labs, but more importantly we're prioritizing our customers that see a surge in usage as they're moving towards systems running on the background and agents that produce many more tokens. So, we're at work building as much capacity as we can to address the large amount of customers that want compute for different kinds of usage.

也许你收到了来自美国客户、AI 实验室、OpenAI 和 Anthropic 的咨询。

Maybe you are getting inquiries from US customers from AI labs, OpenAI, and Anthropic.

每个实验室都希望在欧洲进行推理，因为存在延迟问题等。所以这是一个用例。但说到训练，问题就不完全是地理问题了，而是可用性问题。随着我们拥有可用性，并且我们在这种硬件上有大量训练的成功案例，因为我们自己就在这种硬件上训练。所以，我们在今天向客户提供的训练软件上做了很多改进。

Every lab wants to have inference in Europe because you have latency problems, etc. So, that's one use case. But then, when it comes to training the question is not really a geographical question. It's an availability question. And as we have availability and as we have a lot of proof points on training on this kind of hardware because we train on this kind of hardware ourselves. So, we've made a lot of improvements on the training software that we expose to our customers today.

当我们看纯数字时，Arthur，我们看到你向数据中心投资了 40 亿欧元，而看看大洋彼岸的美国，超大规模企业今年在 AI 基础设施（数据中心、芯片等）上的支出超过 7000 亿或 8000 亿美元。是担心欧洲远远落后，还是更担心那些企业支出过多？

When we look at the pure numbers, Arthur, we see your 4 billion euro investment into data centers and we look over the pond to the US and we look at the hyperscalers spending north of 700 billion dollars or 800 billion, whatever the figure is this year in terms of AI infrastructure going into data centers, into chips, etc. Is there a concern that Europe is so far behind or is there actually more of a concern that those guys are overspending?

我认为这两种情况可能都有一定道理：欧洲企业正在采用这项技术，我们帮助他们采用并提高投资回报率，这样他们就能证明更多支出的合理性。另一方面，全球范围内的部署确实非常激进。欧洲拥有非常庞大且优质的电网。因此，能源可用性很高，建设数百兆瓦的数据中心也相当容易。这是我们的资产。现在，我们的客户告诉我们，AI 变得如此重要，以至于他们实际上需要考虑技术本身的来源。就像能源一样，你既进口能源，也自己生产能源。AI 现在看起来很像能源。你需要能源的可负担性。因此，在开源基础上构建也是进行更多定制、创建自己的模型以便在更小硬件上运行的方式。你需要供应的安全性。如果你的供应商受到外国实体的某些法律约束，你永远不知道会发生什么。因此，就你的韧性计划而言，如果你是一家全球公司的董事会成员，你肯定希望确保你的 token 可以来自世界不同地方。因此，我们能够为所有这些公司提供完全独立的 token 生成，这不仅在欧洲，而且在全世界都很有用。

Well, I think the two things may be slightly true in that European companies are adopting and we help them adopt the technology and increase the ROI so that they can justify more spending. And then, on the other side there's effectively pretty high aggressivity when it comes to deployment everywhere. What Europe has is a very big, very good grid. So, availability of energy is high and it's actually fairly easy to build data centers that are in the 100s of megawatts. So, that's an asset that we have. Now, what our customers are telling us is that AI is becoming so important that they actually need to think about where they source the technology itself. So, the same way in energy, you import energy sources, but you also produce your own energy. AI is really looking like energy at this point in time. You do need to have affordability of energy. So, building on an open-source foundation is the way also to do more customization to create your own models so that they can run on smaller hardware. You need security of supply. If your provider is actually under certain legal constraints that may come from foreign entities, you never know what can happen. And so, when it comes to your resilience plan if you're in the board of a company globally, well, you do want to make sure that your tokens may come from different places of the world. And so, the fact that we can provide fully independent token generation to all these companies is actually useful not only in Europe, but in the entire world.

关于支出过多这部分，你认为美国的一些超大规模企业目前是否支出过多？

And just on that overspending part of the equation, do you think some of the hyperscalers over in the US are spending too much at this point?

很难说。我们看到需求大幅增长。目前，实际上芯片不够，内存也不够。欧洲电力充足，但美国电力不足。所以，目前需求远高于供给。现在，我们都以略微不同的方式预测需求的增长。我们与客户合作，确保他们每花一欧元在 token 上，实际上能得到大约两欧元的回报。因为如果不是这样，整个事情就会崩溃。他们将拥有过多的算力，因为到了某个时候，每个人都会问：‘我花了运营支出的 10% 在 AI 上，它是否带来了超过 10% 的增长？’因此，归根结底，在评估我们是否支出过多或不足时，真正重要的是企业是否有效地采用这项技术来构建实际用例。目前有效并推动需求的是编码。如果你是一名开发者，它会带来很多生产力。但最终，如果你想对实体经济产生影响，你需要将技术应用到现实世界的物体、制造业等。因此，这也是我们真正在投资的方向，确保 AI 系统影响到工程师，不仅是软件工程师，还有工业工程师。

It's hard to tell. What we see is there's a very big increase in demand. And today, there's actually not enough chips, not enough memory. There is enough electricity in Europe, not enough electricity in the US. So, today the demand is actually way above the supply. Now, we all anticipate the rise in demand in slightly different ways. We're at work with our customers to make sure that when they spend a euro in tokens, they actually get like 2 euros in return. Because if that's not the case, the entire thing is going to collapse. They will have too much compute because at some point everybody is going to ask, 'I'm spending 10% of my OPEX in AI. Is it bringing me more than 10% in growth?' So, at the end of the day, what really matters when it comes to estimating whether we're overspending or underspending is whether the enterprises are effectively adopting the technology to build real-world use cases. What is working today and what is driving the demand is coding. If you're a developer, it brings you a lot of productivity. But at the end of the day, if you want to have an impact on the real economy, you do need to bring the technology to real-world objects, to manufacturing, etc. And so, that is also something that we're really investing in, making sure that AI systems are affecting engineers, but not only the software engineers, the industrial engineers as well.

是的，我认为投资回报率部分非常重要，未来将是一个大焦点。我想稍后再谈这个。我想再花几分钟谈谈基础设施这部分，特别是主权问题。

Yeah, that return on investment piece is incredibly important, I think, going forward and such a big focus. I want to get onto that in a minute. Just want to spend a couple more minutes on this infrastructure piece of the equation here and particularly around sovereignty.

我想接着你最近对法国立法者的一些评论，谈谈欧洲目前的需求。你警告说，欧洲只有两年的时间窗口来建设独立的 AI 基础设施，否则就有可能失去控制权，被一些美国科技巨头掌控。我记得这些评论被翻译成，欧洲可能面临沦为美国附庸的风险。请谈谈你对 AI 基础设施的担忧，以及你认为欧洲需要更自主、更独立的基础设施的原因。

And you I want to pick up on some comments you made recently to some of the lawmakers here in France around the requirements for Europe right now. And you warned Europe has a two-year window to build independent AI infrastructure or risk losing control to some of these American tech giants. I think the comments were translated as Europe could risk becoming a vassal state to the US. Just lay out some of your thoughts around what your concerns are when it comes to AI infrastructure and the need for Europe to have a more sovereign and independent infrastructure.

这不仅是欧洲的需求。我认为这是每个想要战略自主的国家的需求。主要原因是经济上的，这项技术将占预算的 10%左右，相当于工资的 10%。所以，如果你看全球工资总额，大约是 50 万亿美元。那么，我们谈论的是价值 5 万亿美元的东西。我估计在未来 5 年内。这完全取决于进展速度，但我相信，有了正确的赋能和正确的模型，我们实际上可以实现。现在，以欧洲为例，欧洲的工资总额约为 9 万亿美元。所以，我们谈论的是未来 5 年 AI 支出大约略高于 1 万亿美元。目前欧洲因数字服务而流向美国的资金是 2500 亿美元。这很多，因为所有这些钱实际上都 reinvested 在美国的研发中，而不是在欧洲。所以，在某种程度上，过度依赖一个地区的技术会产生复合效应。如果没有替代方案，这种复合效应还会加剧。这就是为什么你会看到像印度这样的国家也开始考虑他们的全栈 AI 战略。这也是为什么我们与新加坡合作开发全栈 AI 战略，即使我们消失，他们仍然可以生产技术。这也是为什么欧洲开始像看待天然气一样将 AI 视为战略资产。我认为，即使是政策制定者也开始意识到需要采取行动。但真正推动这一切的是公司。我们在美国、欧洲、亚洲的所有客户中看到，我们提出的以可定制的开源模型为中心的方案引起了他们的共鸣。这带来了需求。我们相信，这个窗口实际上非常短，因为芯片、内存和电力都是有限的。我们相信，我们看到的需求使我们能够在所有与关键任务 AI 部署相关的事情中占据非常有意义的地位。所以，我认为这就是我们的希望。但我感到遗憾的是，这也是我被要求去见法国立法者的原因，我想分享的是，这不仅仅是技术问题。实际上是一个宏观经济问题。如果你想保持竞争力和参与竞争，你就不能承受一万亿美元的贸易逆差。所以，我认为人们正在意识到，我们谈论的是应该让我们每个人都担忧的事情。

It's not only a need for Europe. I think it's a need for every state that wants strategic autonomy. The main reason is economical in that this is a technology that is going to be a line in the budget that is maybe 10% of the wages. So, if you look at the wages in the world, it's around 50 trillion. So, we're talking about something that is worth 5 trillion tokens. I would say in the next 5 years. It all depends on how fast it goes, but I believe that with the right enablement and with the right models, we can actually get there. Now, if you take Europe, Europe is around 9 trillion in wages. So, we're talking about roughly a little north of 1 trillion in spending in AI in the next 5 years. The amount of money that goes back to the US because of digital services today in Europe is 250 billion. That's a lot because all of this money is actually reinvested in R&D in the US and not in Europe. So, in a way, you have some compounding effect of depending too much on technology from one region to another. And that compounding effect is going to increase if there is no alternative. That's the reason why you see states like India also start to think about their full-stack AI strategy. That's the reason why we work with the Singaporean state to develop a full-stack AI strategy in which if we are to disappear, they can still produce the technology. And that's the reason why Europe is starting to look at AI as a strategic asset the same way it has looked at gas. And I would say there is a realization even on the policy maker side that something needs to be done. But really the companies are the ones that are making it happen. What we see with all of our customers in the US, in Europe, in Asia is that the kind of proposition that we bring, which is centered around open-source models that can be customized, is resonating with them. And that brings demand. And we believe that that window, which is fairly short actually because there's only a limited amount of chips, a limited amount of memory, and a limited amount of electricity. We believe that the demand we see allows us to take a very meaningful position in everything that is related to mission-critical AI deployment. So that I think is the hope that we have. But what I'm regretting, and that's the reason why I was asked to go see lawmakers in France, and I wanted to share that this is not only a technological problem. It's actually a macroeconomic problem. You can't afford to have a commercial deficit of a trillion if you actually want to stay competitive and in the race. And so that's something I think that people are realizing that we're talking about something that should be concerning for any one of us.

如果我理解正确的话，你这里谈论的是数字服务，因为现实情况是，物理基础设施，芯片在台湾制造，对吧？英伟达和 AMD 等美国公司设计了一些最先进的 AI 工作负载芯片。内存高度集中在韩国，这似乎不会改变，对吧？

And I think if I'm hearing it correctly, what you're talking about here is the digital services because the reality of the situation is, physical infrastructure, the chips are being made in Taiwan, right? Nvidia is designing and AMD, the US companies designing some of the most advanced chips for AI workloads. Memory is so heavily concentrated in South Korea and that doesn't seem like it's going to change, right?

短期内不会改变。但这并不是什么大问题。我的意思是，我们生活在一个全球化的经济中，这也是过去 50 年我们增长如此之快的原因，这很好。问题在于，我们如何维持现有的平衡，让世界每个地区都能真正繁荣？如今，商品在美国、欧洲、中国、韩国、台湾之间交换。在构建用于部署 AI 的系统方面，我认为从好的方面看，这是相当平衡的，因为半导体供应链是完全交织的。有 ASML，它是关键一环，是一家欧洲公司。有台积电、三星、海力士。美国也有很多晶圆厂，当然还有英伟达。而且，越来越多的芯片设计公司试图颠覆这个领域。所以从好的方面看，欧洲当然可以建造更多。但要做到这一点，它需要一个市场。而要拥有市场，它实际上需要云提供商。所以我们去我们认为有优势的地方，那就是数字服务，部署 AI 无服务器系统，从而构建 AI 应用，部署高价值技能劳动力，将这些无服务器服务转化为交付价值的 AI 应用。我们为此创造价值。然后我们 reinvest in 研发。我们实际购买芯片，最终，我们认为欧洲的技术生态系统，特别是，可以成长到这样一个程度，使得像三星、SK 海力士或台积电这样的公司在欧洲设立晶圆厂成为一个好主意。但要做到这一点，这些公司实际上需要有一个市场。而这个市场就是正在建设的基础设施。所以，让我们从我们强大的地方开始。这就是我们分享的。然后，让我们创造一些东西，让世界上每个国家都能获得足够的杠杆，并以不造成不公平依赖的方式参与 AI 革命。

It will not change short-term. And it's not that much of a problem. I mean, we live in a globalized economy that is the reason why we've been growing so fast in the last 50 years and it's great. The question is how do we maintain the equilibrium that we have in order for every part of the world to actually thrive? Today goods are being exchanged between the US, Europe, China, South Korea, Taiwan. When it comes to building the systems that are used to deploy AI, I would say on the good side, this is fairly balanced in that the semiconductor chain is completely intertwined. You have ASML, which is a critical piece of it, which is a European company. You have TSMC, you have Samsung, you have Hynix. You have a bunch of fabs in the US as well, and then you have Nvidia, of course. And more and more, I would say, chip designers that are trying to disrupt the space. So on the good side, of course Europe could actually build more. But for this, it needs a market. And for it to have a market, it actually needs to have cloud providers. So we go where we think we have an edge, which is the digital services, the deployment of AI serverless systems that allows to build AI applications, the deployment of a high-value skilled workforce that allows to turn those serverless services into AI applications that deliver value. And we build value for that. Then we reinvest in R&D. We actually buy chips, and eventually, we think that the tech ecosystem in Europe, in particular, can grow to a point where it becomes a good idea for a company like Samsung, or SK Hynix, or TSMC, to set up a fab in Europe. But for this to happen, you actually need for those companies to have a market. And the market is the infrastructure that is getting built. So, let's start where we are strong. That's what we shared. And then, let's create something that allows every country of the world to get enough leverage and to participate into the AI revolution in a way that is not creating unfair dependencies.

最后一个关于基础设施的问题，因为你提到了芯片颠覆者。Mistral 在自主设计 ASIC 方面做了多少工作？这在美国的许多云提供商中一直是个热门话题。我们看到谷歌、亚马逊、微软都在做。你们在这方面有进展吗？

And just a final one on infrastructure, just because you mentioned chip disruptors. How much work is going on at Mistral into designing your own ASICs? This has been a hot topic for a lot of the cloud providers in the US. We've seen it for Google, Amazon, Microsoft. Any work going on there?

我们目前还没有做。这当然很有趣。不，我们不排除这种可能性，因为如果你看看芯片设计领域，确实有，虽然不是唾手可得的果实，但也是果实。所以你可以真正地将部署 token 的成本降低到有意义的程度。

So, we don't do it yet. This is, of course, interesting. No, we're not ruling it out, because if you look at the chip design space, there are really, it's not low-hanging fruits, but they are fruits. So you can really lower the cost of deploying tokens to a meaningful extent.

因此，我们正在与几家真正在构建定制 ASIC 的芯片设计公司合作。他们比我们做得好得多。他们拿我们的模型，并试图让它工作。我们在无服务器基础设施上部署系统，这样我们就可以用多种芯片生成 token。我们认为这对客户很重要，对总拥有成本也很重要。如今，我们真正专注于制造模型和运营芯片。我们专注于将模型转化为对企业有用的东西，并具备正确的企业上下文等。这已经是堆栈的重要组成部分。这意味着我们有多个业务部门，但当我们审视我们的市场以及我们在开源模型之上构建的机会时，真正拥有基础设施、拥有产品非常重要。拥有芯片可能也会到来。我认为它应该在某个时候到来，但就目前而言，我们依赖英伟达，它是我们很好的合作伙伴。我们也在各处测试一些东西。

And so, we're working with a few chip designers that are really building custom ASICs. They do a much better job than we do. They take our models and they try to make it work. We deploy our systems on serverless infrastructure so that we can actually create tokens with a multitude of chips. We think it's going to matter for customers. It's going to matter for the total cost of ownership. Today, we're really focused on making the models and operating the chips. We're focused on turning the models into things that are useful for enterprises too, with the right enterprise context, etc. That's already a significant part of the stack. That means we have multiple business units, but really when we look at our markets and when we look at our opportunity of building on top of open source models, really owning the infrastructure, owning the products is very important. Owning the chips may come. I think it should come at some point, but for now we are relying on Nvidia, which is a great partner to us. And we're testing a few things here and there.

那么，就自行设计的芯片而言，目前更多只是一个阶段，看看它们如何与你的模型配合，而不是明显部署在你的数据中心。

So, in terms of self-designed chips, that's more right now just a phase, seeing how they work with your models rather than obviously being deployed in your data centers.

是的，这不是我们知道怎么做的事情，你知道。我们知道如何指定我们想要的东西。我们知道我们需要什么样的内存带宽。我们知道我们需要什么样的网络要求。我们知道我们需要什么样的内存芯片。但随后我们认为其他公司实际上比我们做得好得多。

Yeah, this is not something we know how to do, you know. We know how to spec the things that we want. We know the kind of memory bandwidth we need. We know the kind of network requirements we need. We know the kind of memory chips that we need. But then we think other companies actually do a much better job than we do.

也许有一天。

Maybe one day.

是的，也许有一天，但真正优化成本对整个行业都很重要。

Yeah, maybe one day, but really optimizing for cost is going to matter in the entire industry.

你得开始想些酷炫的名字，你知道，所有这些芯片都有酷炫的名字。所以，你得开始头脑风暴一些了。

You have to start thinking of cool names, you know, all these chips have cool names. So, you have to start brainstorming some of those.

我的意思是，我们已经找到了 Mistral，我认为这很酷。但寻找其他名字，我们留给别人去做。

I mean, we found Mistral already, which I think is pretty cool. But finding others is something we're leaving to others to do.

有一个关于智能体的概念。随着 AI 变得越来越复杂，它将开始自主地为我们做更多事情。你可以让它执行更长的任务等。但其中一个关键部分是编排。想象一下管弦乐队中的指挥。有很多不同的部分需要无缝地连接在一起。智能体连接到公司系统，因此他们的数据，以及人类参与其中。这一切都是关于编排，让这一切在组织内运作，最终目标是让这个智能体有效地成为一个数字助手。

There is this idea of agents. As AI gets more and more sophisticated, it's going to start doing more things autonomously on our behalf. You can ask it to do longer tasks, etc. But there's a key part of that, and that is orchestration. Think about a conductor in an orchestra. There's loads of different parts that need to be joined together seamlessly. Agents connecting to company systems, so their data, the humans in the loop as well. This is all about orchestration, getting this all working within an organization with the ultimate goal of this agent effectively being a digital helper.

那么让我们来谈谈智能体式体验。Arthur，你们推出的另一个新产品叫做 Vibe。据我所知，你们将两个现有产品合并成了这个所谓的智能体式企业产品。请向我们介绍一下这个面向市场的新产品是什么，以及它将如何对业务产生影响。

So let's talk about the agentic experience then. Arthur, one of the other new products you've launched is called Vibe. So here you're combining two existing products as I understand it into what you call this agentic enterprise product. Just run us through what this new product is for the market and how it's going to make an impact in the business.

所以这是一个基于我们开源模型的智能体式平台。我认为与其他平台相比，这已经是一个关键方面。什么是智能体式平台？首先，它是一个你可以像过去三年一样通过聊天交互使用模型的地方。但真正开始从像我们这样的 AI 系统中获得价值的时候，是你开始将一些任务委托给它的时候。所以如果你是一名软件工程师，这意味着连接到你的一个拉取请求并使其更好。这意味着获取一个 PRD 并将其转化为一个拉取请求。这意味着编排多个智能体，它们实际上代表你执行你的工作，模型与软件工程的整个上下文紧密连接。所以这意味着产品设计的产品表面，公司的文档，客户需求等。所以我们押注于这样一个事实：作为开发人员委托任务的工作将与作为非技术人员委托任务的工作非常相似。这就是为什么我们将两个产品合并在一起，其中一个是我们带有命令行界面的编码智能体平台，这是开发人员喜欢使用的。然后我们采用了 LeChat，我们用法语方式命名，但如果你用英语发音，你会得到各种结果。所以我们意识到我们可以把它变成一个相当统一的东西。所以它是一个智能体式平台。智能体式平台里有什么？你有模型。你有不断更新的业务上下文。你将它们连接到你的各种记录系统，然后启动智能体，它们构建你公司实际发生情况的中间表示。然后你有一个执行层。所以你可以告诉一个智能体实际一直运行任务，定期触发等。所以那个执行层很重要。我们做的有什么特别之处？我想说第一点是它建立在开源模型之上，这意味着你可以受益于我们带来的成本优化。第二点是你可以获得比其他提供商更多的控制权。所有状态、所有数据、对用户和组织的所有定制都可以托管在我们的客户租户上。我们可以将部署在客户租户上的内容连接到我们的 token 生成器，并将它们连接到他们 GPU 上的部署。有时我们的客户很难在他们的 GPU 上部署。所以我们结合了客户租户上的有状态托管组件和 Mistral 托管的状态无关 GPU，以两全其美：控制你的数据，以及效率、成本效益和延迟改进。第三点实际上是关于定制化。通过部署代表你做事的个人智能体，你会获得一些好处。如果你正在构建一个具有正确界面和正确编排的采购系统，同时通知多个人，你会获得更多好处。所以我们构建 Vibe 的方式是你可以托管你的自定义应用程序。所以你的采购智能体、你的客户服务智能体，都托管在同一个地方，具有相同的可观测性，正确的治理，连接到相同的数据，并且你可以控制哪些员工可以访问哪些数据。我们构建它使得我们的前向部署工程师可以快速构建业务应用程序，并且这些应用程序易于我们合作的客户访问。

So it's an agentic platform that is based on our open-source models. I think that's already a pretty key aspect compared to others, I would say. What is an agentic platform? It's first of all, a place where you can use the models with chat interaction the same way you've been doing it for the last 3 years. But truly where you start to get value out of AI systems like the one we're doing is when you start delegating some tasks to it. So if you're a software engineer, it means connecting to one of your pull requests and making it better. It means taking a PRD and turning it into a pull request. It means orchestrating a multitude of agents that are actually doing your jobs on your behalf in a way where the models are well connected to the entire context of software engineering. So that means the product surface where the product gets designed, that means the documentation of the company, that means the customer requirements, etc. So we're betting on the fact that the job of delegating tasks as a developer is going to be very similar to the job of delegating tasks as someone who is actually not technical. So that's the reason why we brought together two products, one of them which is our coding agent platform with a command line interface which is the thing that developers like to use. And then we took LeChat which we named in a French way but if you pronounce it in English you get various results. And so we realized that we could actually turn it into something quite unified. So it's an agentic platform. What is there in an agentic platform? You have the models. You have the business context that is constantly updated over time. You connect them to your various systems of records and then you launch agents that are building intermediary representations of what's actually happening in your company. And then you have an execution layer. So you can say to an agent to actually run tasks all the time with triggers on a regular basis etc. So that execution layer matters. What is specific to what we do? I would say the first thing is it's built on open source models, which means that you benefit from the cost optimization that we brought. The second thing is that you get way more control than with other providers. All of the state, all of the data, all of the customization to the user and to the organization can actually be hosted on our customer tenant. We can connect what's deployed on our customer tenant to our token generator and we can connect them to deployment on their GPUs. It's sometimes hard for our customers to deploy on their GPUs. So we've combined a stateful hosted component on our customer tenant and a stateless GPU hosted by Mistral to have the best of both worlds: control over your data and then efficiency and cost efficiency and latency improvement. The third thing is really around the customization. You get some benefit by deploying personal agents that are doing things on your behalf. You get much more benefit if you're building a procurement system with the right interface and with the right orchestration that is pinging multiple people at the same time. And so we've built Vibe in a way where you can host your custom application. So your procurement agent, your customer service agent, all hosted in the same place with the same observability, with the right governance, so connected to the same data and you can control what kind of employees get access to what kind of data. And we've built it so that our forward deployed engineers can actually quickly build applications that are business applications and that are easily accessible by the customers we work with.

对于 Vibe，你们如何处理自主性以及智能体执行任务与人类参与程度之间的关系？

With Vibe, how are you approaching this idea of autonomy and execution of tasks from the agent versus how much is human in the loop?

自主性很重要，因为当你能够将长任务委托给智能体时，你会获得更大的杠杆作用。基础设施方面的一个关键点是如何将生成 token 的 GPU 连接到执行层。你需要一个监督者来确保智能体只做被允许的事情。为此，我们使用 Cohere 团队构建的沙盒，并很快在 Studio 中部署。这些沙盒提供无服务器接口，可以同时启动和关闭大量智能体。自主性需要正确的基础设施：GPU、CPU、无服务器部署和状态管理。当智能体被部署时，它会创建状态、学习、写入记忆，并使用文件系统来组织学习。Vibe 为你处理这些。第二点是人类的验证。你需要针对动态输入进行动态部署，但也需要确定性门控——在采购和计费等地方需要人类验证者。这需要持久的流程执行，能够中断自身以请求人类许可，并与多个人类交互。这就是我们使用工作流的原因，它结合了确定性和动态行为，让 CIO 和业务用户都满意。在 Vibe 中，底层是代码，因此开发者可以修改和维护它。如果智能体只由供应商的专有语言定义，你就无法长期维护它们。Vibe 允许从非技术用户部署到技术用户部署，以实现长期维护。

Autonomy matters because you get more leverage when you can delegate a long task to an agent. One critical aspect in terms of infrastructure is how you connect the GPUs producing tokens to the execution layer. You want a supervisor on top to ensure the agent only does what it's allowed. For this, we use sandboxes built by the Cohere team, deploying them soon in Studio. These sandboxes provide a serverless interface to spin up and down many agents simultaneously. Autonomy requires the right infrastructure: GPUs, CPUs, serverless deployment, and state management. When an agent is deployed, it creates state, learns, writes memories, and uses file systems to organize learning. Vibe handles that for you. The second thing is validation by humans. You need dynamic deployment for dynamic inputs, but also deterministic gates—human validators in places like procurement and billing. This requires durable process execution that can interrupt itself to ask for human permission, interacting with multiple humans. That's why we use workflows, combining deterministic and dynamic behaviors to satisfy both CIOs and business users. In Vibe, the ground source is code, so developers can modify and maintain it. If agents were only defined in a provider's proprietary language, you couldn't maintain them over time. Vibe allows deployment from non-tech to tech users for long-term maintenance.

随着越来越多的智能体进入企业，你认为组织架构或工作流程的最大变化会发生在哪里？

As more agents enter the enterprise, where do you think the biggest changes in organizational structures or workflows will be?

两件事。首先，它颠覆了 SaaS 业务。一旦智能体能够操作很多事情，买家必须确保 SaaS 提供商让智能体能够访问他们的知识。将模型连接到公司的记录系统是关键。一旦完成，你就可以实现大量自动化。但一个大问题是，企业很快就会被自己的组织所瓶颈，而不是模型智能。你需要以可观察的方式编排智能体与人类，系统要能进化并自动修正。这意味着围绕编排流程的智能体重新调整组织。审视核心流程，思考如何让它们更快、更自动化，以及在哪里设置人类门控以保持质量和创新。这是自上而下的：对每个职能和核心业务，围绕一个以智能体为编排者和协调者的 AI 系统重新组织人员。其次，它改变了信息共享的方式。只要你有正确的上下文，就不再需要问同事发生了什么。我们称之为上下文引擎，将模型连接到记录系统和文档。这通过更快的周转节省了时间。但这需要非技术性的改变：业务领导者必须要求他们的下属记录他们所知道的内容。如果你不记录流程，智能体会迷失。作为员工，我们需要尽可能多地提供文本给将成为核心流程编排者的智能体。

Two things. First, it disrupts the SaaS business. Once agents can operate many things, buyers must ensure SaaS providers make their knowledge accessible to agents. Connecting models to a company's systems of record is key. Once done, you can automate a lot. But a big problem is that enterprises quickly become bottlenecked by their own organization, not model intelligence. You need to orchestrate agents with humans in an observable way, with systems that evolve and auto-correct. This means re-adapting the organization around agents that orchestrate processes. Look at core processes, think how to make them faster, automate more, and where human gates maintain quality and innovation. This is top-down: take every function and core business, and re-orchestrate people around an AI system where the agent is the orchestrator and coordinator. Second, it changes how information is shared. You no longer need to ask colleagues what's happening, provided you have the right context. We call it the context engine, connecting models to systems of record and documentation. This saves time with faster turnaround. But it requires non-technical change: business leaders must ask their reports to document what they know. If you don't document processes, agents will be lost. As employees, we need to give as much text as possible to the agents that will become orchestrators of core processes.

你即将听到一个你可能以前听过的术语：人工通用智能或 AGI。有许多不同的定义。广义上，它指的是与人类一样聪明或更聪明的 AI，不限于单一狭窄功能。Arthur 提出了一个有趣的观点：它不是一个神奇的时刻，而是进步的方向，而不是最终目的地。

You're about to hear a term you may have heard before: artificial general intelligence or AGI. There are many varying definitions. More broadly, it's taken to mean AI that is as smart or smarter than humans, not limited to one narrow function. Arthur mentions an interesting view: it's not some magical moment, but rather the direction of progress, not the final destination.

当你想到越来越多的智能体时，我前几天和 Claude Code 的创造者聊过，我问他：‘你认为今年 AI 最大的主题是什么？’他说：‘网络安全。’我理解这一点，因为 Anthropic 有 Me Thoth，这也是一个热门话题。但网络安全当然是企业最关心的问题，随着越来越多的智能体进入企业、接触敏感数据、可能更自主地行动，Mistral 是如何考虑网络安全的？你们是否有或正在考虑某种产品，可以补充你们的智能体产品，类似于 Me Thoth 风格的产品，帮助企业应对因智能体使用增多而可能出现的网络挑战？

When you think about more and more agents, I was speaking to the creator of Claude Code the other day and I asked him, 'What do you think is going to be the biggest theme this year in terms of AI?' He goes, 'Cybersecurity.' And I get that because Anthropic has Me Thoth and that's been a big topic of discussion as well. But cybersecurity, of course, is top of mind for enterprise and as more and more agents get into the enterprise, get on sensitive data, perhaps act more autonomously, how is Mistral thinking about cyber? Is there a product you have or are thinking about that can complement your agentic product, some sort of Me Thoth style product that can help enterprises deal with some of the cyber challenges that could arise from more and more agentic use?

是的，因为我们的客户越来越多地要求我们在网络安全方面提供多种服务。网络安全和 AI 有多个支柱。首先，你可以用 AI 检测漏洞、更快地更新代码库、防范渗透攻击。这不仅仅是模型的问题。模型当然重要，我们在每个类别都看到了巨大进步。我们的模型落后几个月，但正在快速追赶。这些模型能够检测漏洞、提出利用方法，还能用于启发式防御针对网络的渗透攻击。所以，你需要模型，但也需要利用正确工具和网络攻击工具的框架。你需要红队框架来通过渗透测试检验系统，需要蓝队框架来防御系统。这些是我们正在与客户合作的产品，他们迫切要求我们提供解决方案。我们基于开源模型构建这些产品，效果很好。我们认为，就像任何网络安全领域一样，最终你希望系统是开源的。加密如此，开源如此，AI 也将如此。因此，我们押注开源技术，让我们的模型对所有人开放，这样每个人都能理解模型的能力，甚至攻击者也能了解。我们认为这将带来更安全的系统。这是第一个支柱。其次，网络安全也关乎你部署智能体的方式。因为编写代码很容易。当你有一个 ID 时，模型擅长处理大量不同的异构数据源。所以，即使你作为单个用户使用开源工具，也能创建相当强大的部署和智能体，为你做事或充当个人助理。这对个人来说很容易，如果你愿意将数据交给闭源提供商。但对企业来说就困难得多，因为你天生存在一种张力：一方面要让员工和管理者能够构建非常适应业务流程的应用程序，另一方面要确保他们不会削弱安全态势。所以，你需要隔离、监控、护栏。你需要确保模型只访问你授予其访问权限的数据。即使你遵守 IT 提供的访问控制，你还需要确保没有“需要知道”的问题。当你将模型连接到所有记录和文档系统时，你一定会发现某些员工实际上有权访问他们不应该知道的内容。这种情况以前就存在，但因为你在减少摩擦，智能体可以搜索一切，你实际上需要担心这些问题。你需要建立动态访问控制系统。因此，企业需要设置各种原语，才能放心地将强委托交给 AI 系统。这需要系统和模型能力的结合。我们正在与客户一起构建这些，因为通常需要高度定制。你需要模型深入理解遗留系统，以及企业中通常相当混乱的整体 IT 架构。因此，定制化以及我们的安全工程师的部署实际上能让客户更快推进。未来几个月我们会发布更多相关公告。

So, we do because our customers are increasingly asking us for multiple things on the cybersecurity side. Cybersecurity and AI has multiple pillars. First, you can use AI to detect vulnerabilities, do faster updates of your code bases, and guard against penetration attacks. It's not only about the models. Models matter, of course, and we've seen huge improvements in every category. Our models are a few months behind, but they're really catching up quickly. Those models are able to detect vulnerabilities, propose exploits, and also be used for heuristics to guard against cyber attacks that are penetration attacks against networks. So, you need the models, but you also need the harnesses that use the right tools, the right network attack tools. You need the red team harness to test your systems with pen testing. You need the blue team harness to defend your systems. Those are products we're working on with customers that are asking us to provide them urgently with a solution. We are building that on open source models. It works very well. We think that as with anything cybersecurity, at the end of the day, you want the systems to be open source. It's the case for encryption. It will be the case for open source. It will be the case for AI as well. So, we're betting on open source technology. We're betting on making our models available to everyone so that everyone can understand the capabilities they may have even for attackers. We think that's going to lead us to a safer system. That's the first pillar. Now, cybersecurity is also about the way you're deploying agents. Because it's very easy to write code. When you have an ID, models are great at dealing with a lot of different heterogeneous data sources. So, if you use even open source tools as a single user, you can create pretty powerful deployment and agents that are doing things on your behalf or your personal assistant, etc. That's easy if you're a single person and comfortable giving your data to a closed source provider. It becomes much harder if you are an enterprise because you have inherently a tension between making your employees and managers able to build applications that are very adapted to the business processes you want to automate, but you want to make sure they're not doing things that weaken your security posture. So, you want isolation, monitoring, guardrails. You want to make sure that models are accessing only the data you give them access to. Even if you respect the access control that IT provides, you also want to make sure you don't have need-to-know problems. When you connect your models to all your systems of records and documentation, you're bound to find places where certain employees actually have access to things they shouldn't know about. It was already the case, but because you're reducing friction and agents can just look for everything, you actually need to worry about those things. You need to have dynamic access control systems in place. So, you have a variety of primitives that need to be set for an enterprise to be comfortable doing strong delegation to AI systems. That requires a combination of systems and model capabilities. We are at work building them with our customers because often you need high level of customization. You need the models to deeply understand the legacy systems, the overall IT architecture that is often quite messy in enterprises. So, customization, deployment of our security engineers actually enables our customers to go faster. We'll have a few more announcements in the upcoming months about that.

Arthur，在我们结束之际，我想听听你对 AI 在市场上的一些宏观情况的看法。我感觉现在有一种 euphoria（狂热），不知道你是否也有同感。看看一些公开市场，一些内存股在过去一年上涨了 600%到 800%。人们在 X 上讨论把钱投到哪里才能赶上这波 AI 浪潮。我和一位芯片 CEO 聊过，他说甚至韩国的出租车司机都在问他内存短缺何时结束。这种狂热让你担心吗？因为如果出现任何崩盘，或者你认为存在泡沫破裂，我们过去经常看到涟漪效应波及基础设施投资等领域。你现在感觉如何？

Arthur, as we wrap up, I just want to get your take on some of the bigger picture things happening around AI in markets. There feels to me, I don't know if you feel this too, some kind of euphoria right now. I think when you look at perhaps some of the public markets we've seen, some of these memory stocks run up 6, 800% over the last year. People on X are talking about where to put their money to ride this AI wave. I was speaking to a chip CEO who said even a cab driver in Korea was asking him about when the memory crunch is going to end. Does this kind of euphoria concern you right now? Because if there is any kind of collapse or you believe there's a bubble that pops, we've often seen in the past ripple waves across things like investment in infrastructure building and other areas. What are you feeling right now?

我们倾向于认为我们对这种狂热有抵抗力。我的意思是，有理由感到高兴，因为模型确实越来越强。我们看到在企业某些领域，应用开始起步。但还有很多工作要做。企业采用方面仍有很大的粘性，这意味着仍有大量价值有待创造。软件工程师可以使用 AI 系统，但工业工程师还不能。所以，如果你想真正用生成式 AI 打入 30 万亿美元的制造业市场，你需要解决许多尚未解决的新问题。你需要模型理解非常复杂的工具，需要模型理解物理。这是我们最近宣布投资并收购一家构建理解物理的模型公司的原因。

Well, we'd like to think that we're resilient to this kind of euphoria. I mean, there are reasons to be very happy in that the models are really getting stronger. We see in certain domains in enterprises that it's starting to pick up. There's still a lot of work. There's still a lot of viscosity in adoption in enterprises, which means that there's still a lot of value creation to be had. Software engineers, they can use AI systems. Industrial engineers, they cannot use AI systems. So, if you actually want to go and tap the 30 trillion market of manufacturing with generative AI, you actually need to solve a lot of new problems that we haven't solved yet. You need the models to understand tools that are very complex. You need the models to understand physics. That's recent announcements that we've made investing in and acquiring a company that is building models that understand physics.

所以，当我们谈论 AGI 超级智能时，一切都将变得简单。我认为这个想法过于简单了。你需要强大的语言智能，但也需要对物理空间有非常深刻的理解。我们还没有达到那个地步。所以，阻力很大。企业将在未来几年取得进展，但他们需要改变组织架构。有很多领域由于能力不足，AI 尚未产生影响。所以，我们需要构建这些能力。我们可以做到。配方就在那里。你需要更多的数据、更多的算力、专业知识。你可以做到，但还有很多工作要做。

So, when we talk about AGI superintelligence, it's all going to be simple. I think it's too simple an idea. You need strong intelligence in language, but you also need a very strong understanding of the physical space. We're not there yet. So, viscosity is high. Enterprises will make strides in the coming years, but they need to change their organization. There are a multitude of domains where AI is not having an impact yet for lack of capabilities. So, we need to build those capabilities. We can. The recipe is there. You need more data, more compute, expertise. And you can do it, but there's a lot of work to be done.

现在，关于市场上可能发生的事情以及在哪里投资的问题，我们认为，只要我们专注于为客户创造价值，只要我们做出正确的赌注，为他们提供所需的算力，让他们满意，让他们明白这不仅仅是一项技术，而是一场真正的工业革命，我们就没问题。我们认为我们正在做出正确的赌注。我们认为我们对市场上可能发生的任何事情都有韧性。当然，我们告诉客户他们应该关心供应链，因为由于购买热潮，它实际上变得更贵了。所以这也意味着我们需要加速建设我们的算力设施。我们需要确保他们获得好的成本，我们获得好的成本，这样他们就能获得好的价格。所以还有很多工作要做。再次强调，这需要全栈。需要在模型方面、基础设施方面、产品方面。还有很多工作要做，对我们的客户也是如此。但总体而言，未来是光明的，因为我们将进入一个能够更快增长、能够解决我们一直无法解决的问题（如全球变暖）、能够解决由于长寿而遇到的许多健康问题、能够建造更好的飞机（我们今天宣布了与空客的合作）、能够制造更好的光刻机、能够制造更好更安全的汽车（我们今天也宣布了与宝马的合作）的社会。未来是光明的，但我们需要共同努力，确保一切以公平的方式构建。为了构建公平，基于开源可能是正确的赌注。

Now, on the question of whether there might be things on the market and where to invest, we think that as long as we're focused on the creation of value for our customers, and as long as we take the right bets in terms of providing them with the compute they need, in terms of making them happy and making them understand that this is not just a technology, it's actually a true industrial revolution, we think we're good. We think we're doing the right bets. We think we're resilient to anything that may happen in the market. And of course we tell our customers that they should care about the supply chain because it's effectively getting more expensive because there's a frenzy of buying. And so that also means we need to accelerate on our compute facilities. And we need to make sure they get good cost and that we get good cost so that they get good prices. So still a lot of work to do. Again, it needs to be full stack. It needs to be on the model side, on the infrastructure side, on the product side. Still a lot of work, still a lot of work for our customers as well. But overall the future looks bright because we are going to enter a society that can grow faster, that can solve problems we have been unable to solve like global warming, that can solve a lot of health issues we're running into because of longevity, that can actually build better planes. We've announced the partnership with Airbus today, that can actually better lithography machines, that can build better cars, safer cars. We've announced the BMW partnerships today as well. And the future is bright, but we all need to work together and we need to make sure that everything is built in a way that is fair. And to build fairness, building on open source is probably the right bet.

即将到来的 OpenAI IPO，无论何时发生，对行业意味着什么吗？

Does the OpenAI IPO upcoming, whenever that happens, mean something for the industry?

它意味着什么吗？我认为，当然 IPO 总是被关注的。我们作为一家公司是私营公司，可以更长时间保持私营。但这当然是我们会感兴趣关注的事情。

Does it mean something? I think, of course IPOs are always looked at. We as a company are a private company and can stay a private company for longer. But of course that's something that we'll look at with interest.

当然。最后，在我们结束之前，Arthur，你简短地提到了 AGI，我们没有时间深入探讨，但人工通用智能这个想法在过去两三年里被一些 AI 实验室广泛讨论。感觉现在有点过时了。你提到了一些有趣的事情，比如要达到 AGI 我们需要理解物理世界。也有很多关于世界模型等事物的讨论，以便达到 AGI。鉴于我知道这个短语有多种定义，目前行业在实现其认为的 AGI 方面的辩论和进展状态如何？

For sure. And just finally, as we wrap up Arthur, you mentioned AGI very briefly and we haven't got time to go into too in-depth, but this artificial general intelligence idea was something that was spoken about a lot in the last say two three years by some of the AI labs out there as well. Feels like it's gone out of fashion a little bit at this point. You mentioned something interesting like to get there we need to understand the physical world. There's a lot of talk of things like world models emerging as well in order to get to AGI. What is the current state of debate and progress in terms of the industry achieving what it believes to be AGI given I know there's multiple definitions of the phrase.

你知道，我认为 AGI 是一个 PPT 概念。所以我们在 2010 年到 2025 年间听到 AGI 大约 15 年的原因是，这个东西实际上还没有真正起作用。但现在它确实起作用了，我们可以把它变成对客户有价值的东西。所以突然之间，它不再是一个 PPT 术语，但突然之间，事情变得更加复杂，因为如果你在谈论一个还不存在的东西，你在谈论一个抽象的东西，你宁愿用一个单一的术语，而不是说它将会非常复杂，它必须理解许多不同的事物等等。现在我们处于这样一个阶段：实际上你只需要将模型连接到业务数据，并理解人们想用它做什么，从而真正创造价值。但这意味着你会遇到很多在 2010 年的 PPT 中没有预料到的事情。所以你需要它们理解物理，你需要它们理解人类行为，你需要它们连接到我们构建了 50 年的各种软件遗产。所以这一切都非常复杂，非常混乱。它涉及到组织和人，这比你能用技术构建的东西更加混乱。所以归根结底，我认为智能并不重要。我认为重要的是赋能。技术可以做很多事情，但需要大量的管道工作才能使其运转。所以突然之间，你从一个非常抽象的概念（我认为也有点弥赛亚式的）变成了一个非常具体的东西，你需要将不懂技术但懂业务的人与懂技术但不懂客户业务的人结合起来。所以他们需要分享知识，这就是我们构建真正有用的 AI 的方式。这就是我们加速技术进步的方式。所以如果你把 AGI 定义为加速技术进步，这更像是一个方向而不是一个终点，那么实际上我们正在构建 AGI。但这确实是一个方向，而且是一个非常混乱的方向。有很多东西需要构建，前沿是巨大的。

You know, I think AGI was a PowerPoint concept. And so the reason why we've heard about AGI for like 15 years between 2010 and 2025 is because the thing was not really working yet. But now it's actually working and we can turn it into something that has value for customers. And so suddenly it becomes no longer a PowerPoint term, but suddenly the thing becomes way more complex because if you're talking about something that does not exist yet, you're talking about something abstract, you would rather have a single term instead of saying it's all going to be very complex, it's going to have to understand many different things etc. Now we're at the stage where actually you just need to take the models and connect them to business data and to understand what the people want to do with it to actually build value. But it means you're running into a lot of things that were not anticipated in the PowerPoint of 2010. So you need them to understand physics, you need them to understand human behavior, you need them to be connected to all sorts of the 50-year legacy of software that we've been building. So it's all very complex, it's all very messy. It's holding to organization and people which is even messier than what you can build with technology. And so at the end of the day I don't think intelligence matters. I think what matters is empowerment. It's technology that can do many things, but there's a lot of plumbing to be done to make it work. So suddenly you move from a very abstract concept and what I think is a little bit Messianic as well into something that is very concrete and where you need the combination of people that do not understand technology but understand their business with people that understand their technology but do not understand the business of their customers. So they need to share knowledge and that's the way we can build AI that is actually useful. That's the way we can accelerate technological progress. So if you take AGI as the definition of accelerating technological progress, which is rather a direction than a point of arrival, then really we are building AGI. But it's really a direction and it's a very messy direction. There's a lot of things to be built and the frontier is enormous.

Arthur，我有一整张清单，我们下次见面时需要讨论。机器人技术、物理 AI，还有很多其他事情，但我们留到下次对话吧。

Arthur, I've got a whole list of things we need to talk about next time we catch up. Robotics, physical AI, and a ton of other things, but we'll save that for the next conversation.

必须抓住它。

Have to grab it.

非常感谢。非常感谢你加入我。我们正在全力推进 Tech Download 播客。未来几周我们还有一些大型采访，我是说大型。所以请确保你订阅、关注，做任何你需要做的事情以保持更新。它将在 YouTube、cnbc.com 以及任何你获取播客的地方上线。如果你想和我讨论这一集我们讨论的一些事情，或者对下一集应该讨论什么有建议，你可以直接联系我。只需在 LinkedIn、TikTok、Instagram 或 X 上搜索 Arjun Kharpal。感谢收听和观看，我们下次再见。

Thank you very much. Thank you so much for joining me. We are going full steam ahead with the Tech Download podcast. We have some big interviews, I mean big, coming up over the coming weeks as well. So make sure you subscribe, follow, do whatever you need to do to keep up to date. It'll be on YouTube, it'll be on cnbc.com, and anywhere you get your podcast. If you want to talk to me about some of the things that we discussed in this episode or have suggestions on what we should talk about next, you can get in touch with me directly. Just search Arjun Kharpal on LinkedIn, on TikTok, Instagram, or X. Thanks for listening and watching, and we'll catch you next time.