Inside Anthropic's $100 Billion Al Compute Commitment | CFO Krishna Rao

Every time we have a new model, there's a set of capabilities that are different. People tend to think about model intelligence as IQ. We think of it kind of differently. Intelligence for us is multi-dimensional. It's not just a score. What is the real-world capability of this model? Each model generation gives you the chance to do more with it, to do it better, to do it more efficiently, because we think the returns to frontier intelligence are extremely high. And it's extremely high, especially in enterprise. That's a core thesis of our business.

每次我们推出新模型时,都会带来一系列不同的能力。人们往往把模型智能理解为智商。但我们的看法有所不同。对我们来说,智能是多维度的,不只是一个分数。这个模型在现实世界中的实际能力是什么?每一代模型都让你能用它做更多事情,做得更好,做得更高效,因为我们认为前沿智能的回报极高。尤其在企业领域,回报极高。这是我们业务的核心理念。

Krishna, I have been so excited for this conversation because you get to see from the inside one of the most interesting businesses in world history at maybe the most interesting time in world history, at least if you're a technologist or care about technology. One of the things that fascinates me most, just to dive right into something that I think we're both quite

Krishna,我一直非常期待这次对话,因为你能从内部视角看到世界历史上最有趣的企业之一,而且是在世界历史上可能最有趣的时刻——至少对技术专家或关心技术的人来说是这样。最让我着迷的事情之一,我想直接切入我们都非常

passionate about is this question of compute that you have to deal with all day every day. It's a key part of what you do. It's a key part of what these companies are doing and there's just this whole revolution happening.

热衷的话题,就是算力问题,你每天都要处理这个问题。这是你工作的关键部分,也是这些公司正在做的关键部分,而且正在发生一场革命。

I'd love you to just start by explaining what it's like to have to deal with that.

我很想听你解释一下处理这个问题是什么感觉。

Like I understand at one point you were having like a daily meeting about how to allocate the compute and to who and why. Just like bring us into that part of your life because I think it's like right at the cutting edge of what's going on.

据我了解,你们曾经有一段时间每天开会讨论如何分配算力,分配给谁,为什么分配。带我们了解一下你生活中的这部分吧,因为我觉得这正处于当前发展的最前沿。

Look, the compute that we procure is the lifeblood of our business. It is the most important thing in the company. It is the thing on which, it's like the canvas on which everything else gets built.

听着,我们采购的算力是我们业务的命脉。它是公司里最重要的东西。它就像画布,其他一切都在上面构建。

And so the decisions we make and how much compute to buy are some of the most consequential and hardest decisions to make in the entire company.

所以我们关于购买多少算力的决策,是整个公司最重要、最难做的决策之一。

You know, think of it this way. If you buy too

你可以这样想。如果你买太多

Much compute, you go out of business. If you buy too little compute, you can't serve your customers and you're not at the frontier is the same thing.

算力,公司就会倒闭。如果买太少算力,你就无法服务客户,也无法保持在前沿,结果是一样的。

So, you know, we talk a lot about this cone of uncertainty, but the idea of just these purchases that have these real world implications, right?

所以,我们经常谈论这个「不确定性锥」,但这些采购决策会产生实际影响,对吧?

You can't just go out and, you know, buy a gigawatt of compute and have it delivered next week.

你不能直接出去购买一千兆瓦的算力,然后下周就能交付。

You have to really think ahead to plan for this. And so, we really take a very disciplined approach to how we think about it.

你必须提前认真规划。所以,我们对如何思考这个问题采取了非常严谨的方法。

So, we look bottoms up. You know, we model what we think demand will be. Obviously, we sometimes get that wrong.

我们自下而上地分析。我们建模预测需求会是什么样。显然,我们有时会预测错误。

We think about the compute we need to stay at the frontier and we really look ahead and try to estimate that. And then as we go out and actually do these deals to procure compute, you know, flexibility is really important to us and so we build that flexibility into the deals themselves.

我们考虑保持在前沿所需的算力,并真正向前看,尝试估算。然后当我们实际去做这些采购交易时,灵活性对我们来说非常重要,所以我们把灵活性构建到交易本身中。

into how we use the compute as well, because the way in which we bridge from a position we are today to where we want to go when the business is growing exponentially is to use that compute as efficiently as possible.

也构建到我们如何使用算力中,因为当业务呈指数级增长时,我们从今天的位置过渡到想要达到的目标的方式,就是尽可能高效地使用算力。

I would say I spend 30 or 40% of my time on compute even today.

我可以说,即使到今天,我仍然把30%到40%的时间花在算力上。

What does flexibility mean in that example?

在那个例子中,灵活性是什么意思?

So it means a couple of different things. Number one, you know, we use three different chip platforms.

它有几层不同的含义。首先,我们使用三种不同的芯片平台。

So we are customers of Amazon's Trainium chip, Google's TPUs and Nvidia's GPUs.

所以我们是Amazon的Trainium芯片、Google的TPU和Nvidia的GPU的客户。

You know, we use these chips fungibly.

我们可互换地使用这些芯片。

So if you think about the compute we buy, we're using it for model development.

所以如果你想想我们购买的算力,我们用它来开发模型。

We're using it internally to speed up our own product and model development.

我们在内部使用它来加速我们自己的产品和模型开发。

And then we're also using it obviously to serve customers.

然后我们显然也用它来服务客户。

Across those three chip platforms, we're using compute for all of those internal and external uses.

在这三个芯片平台上,我们将算力用于所有这些内部和外部用途。

And that

而这

Flexibility, it actually took us a long time to be able to do that.

在灵活性方面,我们其实花了很长时间才做到这一点。

We've invested in that over multiple years to be what I believe the most efficient users of compute amongst any of the frontier labs.

我们在这方面投入了好几年时间,我相信现在我们是所有前沿实验室中算力使用效率最高的。

And that's not something that just happened overnight.

这不是一夜之间就能实现的。

You know, when we started using TPUs, I think it was maybe the third generation TPUs was the first one we used at scale, people thought, "Oh, well, you're crazy. Everyone's using GPUs. Why aren't you using GPUs?"

我们开始使用 TPU 的时候,我记得第一次大规模使用的是第三代 TPU,当时人们觉得「你们疯了吧,大家都在用 GPU,你们为什么不用 GPU?」

And we've invested very heavily to be able to use that compute incredibly flexibly.

我们在这方面投入了大量资源,就是为了能够极其灵活地使用这些算力。

And then we look across the different generations of those chip platforms and use each generation for the best workload internally.

然后我们会研究这些芯片平台的不同代次,在内部为每一代选择最适合的工作负载。

And so we really built this orchestration layer that gives us that flexibility to use all different types of compute and in doing so we also are able to get the most value out of it.

所以我们真正构建了这样一个编排层,让我们能够灵活使用各种不同类型的算力,这样做也让我们能从中获得最大价值。

Am I thinking about this in the right way that like something like CUDA

我这样理解对不对,比如像 CUDA 这样的东西

That has been a part of NVIDIA's story for a long time now that allows you to do a lot with the underlying actual hardware that you want to sort of eke your way into being as close to the bare metal as possible, and that's part of this flexibility and being able to control as many of the variables as you can.

它一直是 NVIDIA 故事的一部分,让你能够充分利用底层的实际硬件,你想尽可能接近裸机层面,这就是灵活性的一部分,能够控制尽可能多的变量。

Is that the journey that you've been on?

这就是你们一直在走的路吗?

That's part of the journey for sure, but it's also been actually pretty collaborative.

这确实是我们旅程的一部分,但实际上也是一个很协作的过程。

So we work really closely with the Annapurna Labs team at Amazon to help influence the roadmap of these chips, because we believe what we're doing is really stressing the limits of what these chips are capable of.

我们与 Amazon 的 Annapurna Labs 团队紧密合作,帮助影响这些芯片的路线图,因为我们相信我们正在做的事情真正在挑战这些芯片能力的极限。

And that means that a dollar of compute inside our organization goes further than I think it does anywhere else.

这意味着在我们组织内部,每一美元的算力能发挥的作用比其他任何地方都要大。

But importantly, we basically want to utilize each chip to its best purpose within the company.

但重要的是,我们基本上想让每个芯片在公司内部发挥它最擅长的用途。

So that does mean that we're building our

所以这确实意味着我们在构建自己的

own compilers. We're really building things from the chip level up in order to have that customization and that flexibility to use it internally the way we think will generate the most ROI.

编译器。我们真的是从芯片层面开始构建一切,以便拥有这种定制化能力和灵活性,按照我们认为能产生最大投资回报的方式在内部使用。

Can you explain this cone of uncertainty thing? Like I want to ask about all the component parts of this, but that feels like a really key starting point or overall frame for how to think about both sourcing and then the uses of compute. Can you just explain what that concept is?

你能解释一下这个「不确定性锥」的概念吗?我想问这个的所有组成部分,但这感觉是一个非常关键的起点或整体框架,用来思考算力的采购和使用。你能解释一下这个概念是什么吗?

Sure. When you're building and growing a business exponentially, you know, really small movements in monthly or weekly growth rates result in compounding very, very different outcomes. And so as we're thinking ahead, you know, even with our revenue growth, it's really hard to predict this business, right? And it's really hard. I think humans mostly think linearly and you think incrementally. And that's something, you know, I've been at the company for two years. That's something

当然。当你在指数级地构建和发展一个业务时,月度或周度增长率的很小变动,经过复利效应会导致非常非常不同的结果。所以当我们展望未来时,即使是我们的收入增长,也真的很难预测这个业务,对吧?这真的很难。我觉得人类大多是线性思考的,你会增量式地思考。这是我在公司两年来

That's a paradigm I've had to break for myself, right? To stop just thinking linearly and think on this exponential.

必须为自己打破的一个范式,对吧?停止线性思考,开始用指数思维思考。

When you're on this exponential again, the range of outcomes starts to be really, really wide.

当你处在这个指数曲线上时,可能的结果范围会变得非常非常宽。

We look at a range of scenarios and we look at different points in that cone of uncertainty over, you know, a one to two-year period, and then we kind of work backwards from that.

我们会看一系列场景,看这个不确定性锥在一到两年期间的不同点位,然后我们从那里反推。

And what we want to do is we want to be at a place where we can, you know, obviously still be at the frontier. That's the most important thing.

我们想要达到的状态是,显然仍然处于前沿,这是最重要的。

To be able to serve customers and then to be able to have enough internal compute to accelerate our employees.

能够服务客户,然后能够有足够的内部算力来加速我们员工的工作。

It's interesting. If we were to say to our employees, you can't use our models anymore, we could serve billions of dollars of revenue with that compute that we allocate to employees internally, but we want to take a long-term view and a long-term perspective on that cone of uncertainty.

有意思的是,如果我们对员工说,你们不能再使用我们的模型了,我们可以用分配给员工内部使用的那些算力来服务数十亿美元的收入,但我们想对这个不确定性锥采取长期视角和长期观点。

Because we want to range towards the top end of these outcomes, but we have to plan for that.

因为我们想朝着这些结果的上限发展,但我们必须为此做好规划。

And as we go, that's how we think about buying compute in a disciplined way.

随着业务发展,我们就是这样有纪律地购买算力的。

The most important thing is what happens if you are at one point in the cone of uncertainty but you've only bought compute for a different point.

最关键的问题是:如果实际情况处于不确定性锥体的某个点,但你购买的算力却是按另一个点规划的,会发生什么?

That's where this compute efficiency is something that has really helped us out.

这就是算力效率真正帮到我们的地方。

Can you bring us into the room for the conversations around the trade-offs between those?

能不能带我们了解一下你们内部是如何权衡这些取舍的?

I'm so interested by those three buckets of like training, research, internal use broadly speaking, and then serving customer demand.

我对这三大类特别感兴趣——训练、研究、内部使用,以及服务客户需求。

You naively might think like, okay, it's a third, a third, a third allocation or something. Like how much does that range around? What are the trade-offs like? What is that discussion like?

你可能会天真地以为是三三制分配,各占三分之一。但实际分配比例会有多大浮动?权衡考量是什么?这种讨论是怎么进行的?

On an ongoing basis, in addition to meeting about compute procurement, we meet a lot about compute allocation.

在日常运营中,除了讨论算力采购,我们还经常开会讨论算力分配。

I think what's

我觉得

Important is it starts with a place where our culture is one that's incredibly collaborative, and that informs how this conversation happens. So there's not like FTEs—it's done in a very collaborative, not a zero-sum way.

重要的是,这一切始于我们高度协作的文化,这种文化决定了讨论的方式。不像按全职员工数量分配那样——整个过程非常协作,不是零和博弈。

But there's a level of compute for model development that we will not go below, right? So even if it means it's harder to serve customers or we have to do kind of unnatural things when it comes to that, we want to continue to make that long-term investment in developing the best models.

但模型开发有一个算力底线,我们绝不会低于这个标准。即使这意味着服务客户会更困难,或者我们不得不采取一些不太自然的做法,我们也要继续对开发最佳模型进行长期投资。

Because we think the returns to frontier intelligence are extremely high, and it's extremely high especially in enterprise.

因为我们认为前沿智能的回报极高,尤其是在企业领域。

But so that kind of puts a floor on the compute that's allocated to model development.

所以这就为模型开发设定了算力分配的下限。

And then as we think about the internal use of compute, it really helps us to speed up that model development and to speed up finding those compute efficiency multipliers that really get us more from each dollar.

至于算力的内部使用,它确实帮助我们加快模型开发速度,并加快找到那些算力效率倍增器,让每一美元发挥更大作用。

Of compute. So when we're talking about it, each team is kind of representing what they would do with that compute.

所以在讨论时,各个团队会说明他们会如何使用这些算力。

And then we have a really open and frank discussion about how we think about ROI.

然后我们会非常开放坦诚地讨论如何看待投资回报率。

And because we can allocate that compute so dynamically, we can make changes. We can make adjustments in that on a relatively short time horizon.

而且因为我们可以非常灵活地分配算力,所以能够在相对较短的时间内做出调整和改变。

The efficiency thing is so interesting to me. I'm curious if you have a sense of like how much more efficient you are versus your own internal benchmarks from a year ago or something, or versus others that you have some sense of how efficient they are. How do you measure what efficiency means?

效率这个话题我很感兴趣。我好奇你们是否知道,相比一年前的内部基准,或者相比你们了解到的其他公司,你们的效率提高了多少?你们如何衡量效率?

So, there's a couple different ways I would think about it. From a model perspective, I think the analogy people have when these new models come out is like they're kind of like cars. You had a sedan before and then you might have the higher-end version of that sedan and you're kind of moving, moving, moving up.

有几种不同的角度来看这个问题。从模型角度来说,人们对新模型发布的类比就像汽车一样。你之前有一辆轿车,然后可能升级到那款轿车的高配版,就这样一步步往上走。

The chain, and I think that is true in terms of model intelligence. The way the place that analogy kind of breaks down a little bit is people think, okay, I'm going from the sedan to the sports car.

这个类比在模型智能方面确实成立。但这个类比有点不太准确的地方是,人们会想:好,我从轿车换到跑车了。

I'm going to get much less fuel efficiency, right? I'm not going to buy the sports car for the gas mileage.

油耗肯定会高很多,对吧?我买跑车可不是为了省油。

In our case, we actually see both improvements, huge improvements in capability, but also in model efficiency.

但在我们这里,实际上两方面都有提升——能力有巨大提升,模型效率也同样大幅提升。

And so, if you look at going from Opus, you know, 4 to 4.5, 4.6, and now 4.7, you know, each one of those leaps, they're not equal, but each one of those leaps to a new model has a multiplier in terms of how much more efficient it is at processing tokens effectively.

所以如果你看从 Opus 4 到 4.5、4.6,再到现在的 4.7,每一次跃升——虽然幅度不完全相同——但每次升级到新模型,在处理 token 的效率上都有倍数级的提升。

And that just doesn't serve customers. That also helps us internally as well, because if you think about if we're using the model for, if we're doing reinforcement learning on the model, it's basically inference within a sandbox with a reward function.

这不仅服务于客户,对我们内部也有帮助。因为如果我们用模型来做强化学习,本质上就是在一个带有奖励函数的沙盒环境中进行推理。

Right? And so if the model's better at more efficient inference, that RL is more efficient as well.

对吧?所以如果模型的推理效率更高,那强化学习的效率也会更高。

And so we're able to do this kind of win-win where the customer is getting more capability when we release a new model.

所以我们实现了双赢——当我们发布新模型时,客户获得了更强的能力。

And then we're able to serve that model sometimes again a multiple more efficient than the prior generation.

然后我们就能部署那个模型,有时候效率能比上一代提升好几倍。

And then when we're in between generations, we're dynamically deploying efficiency improvements kind of in between these like more step function model changes.

在两代模型之间的时期,我们会动态部署效率改进,就是在这些更大的阶跃式模型变化之间持续优化。

And so it is always getting more efficient over time.

所以效率一直在随时间提升。

And what fuels that is the research team.

而推动这一切的是研究团队。

So if you think about it, all these things are very connected—these various tasks and workloads that we have internally all kind of fit together in this way of, you know, doing R&D for model capabilities, for compute efficiency, for serving customers, and then having internal workloads that can be sped up by using the best models, sometimes models that we

如果你仔细想想,所有这些事情都是紧密相连的——我们内部的各种任务和工作负载都以这种方式结合在一起,包括模型能力的研发、计算效率的优化、为客户提供服务,还有内部工作流程可以通过使用最好的模型来加速,有时候是我们自己的模型。

Unlike most software companies that try to maximize your time on their app to juice engagement, Ramp does the exact opposite. It understands that no one wants to spend hours chasing receipts, reviewing expense reports, and checking for policy violations.

不像大多数软件公司想方设法让你在他们的应用上花更多时间来提升参与度,Ramp 恰恰相反。它明白没人想花几个小时追踪收据、审核报销单、检查违规情况。

So they built their tools to give that time back, using AI to automate 85% of expense reviews with 99% accuracy.

所以他们开发的工具就是为了把时间还给你,用 AI 自动化处理 85% 的费用审核,准确率达到 99%。

And since Ramp saves companies 5%, it's no wonder that Shopify runs on Ramp, Stripe runs on Ramp, and my business does too. To see what happens when you eliminate the busy work, check out ramp.com/invest.

而且 Ramp 能为公司节省 5% 的开支,难怪 Shopify 在用 Ramp,Stripe 在用 Ramp,我的公司也在用。想看看消除繁琐工作后会发生什么,可以访问 ramp.com/invest。

Felix by Rogo is a personal finance agent that turns a single prompt into finished client-ready work using your firm's own templates, context, and standards.

Rogo 推出的 Felix 是一个个人财务智能体,能把一句简单的指令转化为完成的、可交付客户的工作成果,使用的是你公司自己的模板、背景信息和标准。

Send Felix an email like, "Take these comments and turn them for me," or "Update my tracker with the context of these emails," or "Run the ability to pay math on this buyer," and Felix sends back finished PowerPoint decks.

给 Felix 发个邮件,比如「把这些评论整理成文档」,或者「用这些邮件的内容更新我的跟踪表」,或者「对这个买家做支付能力分析」,Felix 就会返回完成的 PowerPoint 演示文稿。

Excel models, and sourced research." Felix works the way your team already does, delivering work quickly and accurately around the clock.

还有 Excel 模型和有来源的研究报告。Felix 按照你团队已有的工作方式运作,全天候快速准确地交付工作成果。

Learn more at rogo.ai/felix.

了解更多请访问 rogo.ai/felix。

OpenAI, Cursor, Anthropic, Perplexity, and Vercel all have something in common.

OpenAI、Cursor、Anthropic、Perplexity 和 Vercel 都有一个共同点。

They all use WorkOS. And here's why.

他们都在使用 WorkOS。原因如下。

To achieve enterprise adoption at scale, you have to deliver on core capabilities like SSO, SCIM, RBAC, and audit logs.

要实现大规模的企业级采用,你必须提供核心能力,比如 SSO、SCIM、RBAC 和审计日志。

That's where WorkOS comes in.

这就是 WorkOS 的用武之地。

Instead of spending months building these mission-critical capabilities yourself, you can just use WorkOS APIs to gain all of them on day zero.

与其花几个月时间自己构建这些关键能力,不如直接使用 WorkOS 的 API,从第一天起就拥有所有这些功能。

That's why so many of the top AI teams you hear about already run on WorkOS.

这就是为什么你听说过的那么多顶尖 AI 团队已经在使用 WorkOS。

WorkOS is the fastest way to become enterprise-ready and stay focused on what matters most, your product.

WorkOS 是最快达到企业级就绪状态的方式,让你能专注于最重要的事情——你的产品。

Visit workos.com to get started.

访问 workos.com 开始使用。

You said something really important before, which is the returns to being at the frontier are really high.

你之前说了一个很重要的观点,就是处于前沿位置的回报非常高。

Can you just explain

能不能尽可能详细地解释一下。

That in as much detail as you can.

这个问题。

Sounds obvious when you say it, but there's certainly been some camps that like, oh, I'll just, you know, I can use the six-month-old model and it's a fraction of the cost and I'll just use that and that'll be catching up all the time.

听起来很显而易见,但确实有些人认为,哦,我就用六个月前的旧模型,成本只是一小部分,我就用那个,它会一直追赶上来的。

And that just hasn't been the case. Like everyone, the second Opus 4.7 comes out, like even me as a consumer, the thing you do is you switch it on or GPT 5.5 comes out, you switch on the new one right away. Like I want the best.

但事实并非如此。就像所有人,Opus 4.7 一发布,甚至我作为普通用户,你做的第一件事就是立刻切换过去,或者 GPT 5.5 一出来,你马上就换到新版本。因为我想要最好的。

So talk about the returns to being on the frontier and why it's so high.

那么谈谈处于前沿的回报,以及为什么这个回报如此之高。

I think it's a couple things. It's every time we have a new model, there's a set of capabilities that are different. People tend to think about model intelligence as IQ. It's a single number. Okay, this model was at 110 and then it goes to 125. We think of it kind of differently. Intelligence for us is multi-dimensional. It's not just a score. In fact, we find that yes,

我认为有几个原因。每次我们推出新模型时,都会有一系列不同的能力。人们倾向于把模型智能看作智商,认为它是一个单一的数字。好吧,这个模型原来是110,然后变成了125。我们的看法有些不同。对我们来说,智能是多维度的,不只是一个分数。事实上,我们发现确实如此,

Everyone publishes their model benchmark cards and finds that a lot of those benchmarks are saturated.

每个人都会发布他们的模型基准测试卡,然后发现很多这些基准测试已经饱和了。

You know, we publish it too.

你知道,我们也发布这些。

But what our measurement is is what the customers tell us, like what is the real world capability of this model.

但我们的衡量标准是客户告诉我们的,比如这个模型在真实世界中的能力是什么。

And as we've released better and better models, what we've seen is it's not just, you know, the outright intelligence.

随着我们发布越来越好的模型,我们看到的不仅仅是纯粹的智能水平。

It's also the ability to do long horizon tasks, the ability to use tools or computer use, the ability to do things for agentic tasks that have specific value even faster, right?

还有执行长期任务的能力、使用工具或计算机的能力、更快地完成具有特定价值的智能体任务的能力,对吧?

Which means that in some sense, you know, if you have two employees and they're maybe both equally capable, someone takes a week to do, you know, an assignment, someone does it in a day.

这意味着在某种意义上,你知道,如果你有两个员工,他们可能能力相当,一个人花一周时间完成一项任务,另一个人一天就完成了。

Well, that second person, if they're continuing to do that, can be seven times better, right?

那么,如果第二个人持续这样做,他可以好七倍,对吧?

They might be equally capable at something, maybe just take longer times to do it.

他们在某件事上可能能力相当,只是完成所需的时间更长。

So all of those factor in to then how customers

所以所有这些因素都会影响客户的

Experience it. And what we found very consistently is by releasing new models, the TAM is unlocked in a unique way.

体验。我们非常一致地发现,通过发布新模型,总体可达市场(TAM)以独特的方式被解锁了。

Like more TAM gets unlocked, more use cases are possible.

更多的TAM被解锁,更多的用例成为可能。

And a good illustration of that is this last four months that we've had at the company, right?

一个很好的例证就是我们公司过去四个月的情况,对吧?

We started the year with about $9 billion of run rate revenue and we ended the quarter with, you know, north of $30 billion of run rate revenue.

我们年初的年化收入约为90亿美元,到季度末时,我们的年化收入超过了300亿美元。

I mean, that kind of a change is really enabled by these model intelligence leaps and then the products that we build around them.

我的意思是,这种变化真正得益于这些模型智能的飞跃,以及我们围绕它们构建的产品。

And so that's what I mean by the returns to frontier intelligence are really high.

所以这就是我所说的前沿智能的回报真的很高的意思。

I think that's unique to enterprise because in consumer sometimes you don't see that as readily, that consumers really are pushing the limits of what the models can do, whereas in enterprise like our customers are always—now you know it started with coding but it's really expanded beyond that very meaningfully—but each model

我认为这对企业来说是独特的,因为在消费者领域,你有时不会那么容易看到这一点,消费者真的在推动模型能力的极限,而在企业领域,我们的客户总是——现在你知道它始于编码,但已经非常有意义地扩展到了其他领域——但每一代模型

Generation gives you the chance to do more with it, to do it better, to do it more efficiently, and customers see that and then they invest really heavily in more tokens with the newer models, and we've just seen that cycle play out again and again, and that's a core thesis of our business that especially in enterprise, the returns to frontier intelligence are not slowing down.

都让你有机会用它做更多事情,做得更好,做得更高效,客户看到了这一点,然后他们在使用更新模型的更多token上进行大量投资,我们一次又一次地看到这个循环上演,这是我们业务的核心论点,特别是在企业领域,前沿智能的回报并没有放缓。

The things that push that frontier is like a sci-fi story or something from books I was reading when I was growing up.

推动这个前沿的事情就像科幻故事或我小时候读的书里的东西。

It seems as though in the major labs we've reached this point—someone on your team said it recently—of like recursive self-improvement, where the models themselves are building and doing a lot of the research to do, you know, the next generation of improvement.

似乎在主要的实验室中,我们已经达到了这样一个点——你们团队的某个人最近说过——就像递归自我改进,模型本身正在构建并进行大量研究工作,来实现下一代的改进。

And that there is some sort of—if I think about the frontier that you're pushing and OpenAI is pushing and compare that to the open source models—that maybe the gap will widen as a...

而且存在某种——如果我想想你们正在推动的前沿和OpenAI正在推动的前沿,并将其与开源模型进行比较——差距可能会作为一个...

Result of you getting there first to this like recursive thing. How do you think about that?

你们首先到达这种递归状态的结果而扩大。你怎么看待这个问题?

Like what, tell us how we should think about this idea of recursive self-improvement in the models themselves, because it seems like getting there first is incredibly important because then you just can continue to separate yourself versus those that haven't reached it yet.

告诉我们应该如何理解模型本身的递归自我改进这个概念,因为似乎首先到达那里非常重要,因为这样你就可以继续将自己与那些尚未达到的人区分开来。

I would say we do see progress accelerating. We see, you know, I can't speak for other companies, but for us the scaling laws are, you know, alive and well and we're seeing that, you know, even with releases more recently like Mythos, right now within the company, you know, 90 plus percent of our code is actually written by Claude Code, right? A lot of Claude Code is written by Claude Code, and so you think of this as like, why do we allocate compute internally? Why would we forego revenue for it? It's because the models themselves are helping us to build that next generation of models.

我会说我们确实看到进展在加速。我们看到,你知道,我不能代表其他公司发言,但对我们来说,规模定律是,你知道,活跃且运作良好的,我们看到,你知道,即使是最近像Mythos这样的发布,现在在公司内部,你知道,我们90%以上的代码实际上是由Claude Code编写的,对吧?很多Claude Code是由Claude Code编写的,所以你可以把这看作是,我们为什么要在内部分配算力?我们为什么要放弃收入来做这件事?这是因为模型本身正在帮助我们构建下一代模型。

And so in addition to this capability leap that you would have just from the scaling laws, talent is really important.

所以除了扩展定律带来的能力飞跃之外,人才也非常重要。

And that talent with the best models can really accelerate the development of the capabilities.

拥有最佳模型的人才能够真正加速能力的发展。

And we're really seeing that.

我们确实看到了这一点。

We don't really think about models as like closed or open.

我们并不真的把模型看作是闭源或开源的。

We think of them as frontier or not.

我们把它们看作是前沿的或非前沿的。

And the ones that are at the frontier, you know, clearly are capturing this economic value, driving meaningful ROI for customers.

那些处于前沿的模型,显然正在捕获经济价值,为客户带来有意义的投资回报。

And we are just investing behind that thesis.

我们就是在围绕这个论点进行投资。

And that means, you know, both compute but it also means talent to use that compute and use our own models to really accelerate the development.

这意味着,既要投资算力,也要投资人才来使用这些算力,并使用我们自己的模型来真正加速开发。

The other piece of it is it's not just the models, it's the products that get built on top of them.

另一个方面是,不仅仅是模型本身,还有基于模型构建的产品。

Right.

对。

So we had 30 different product and feature releases in January.

我们在一月份发布了30个不同的产品和功能。

The pace of that has accelerated as well and that's enabled in part by utilizing the models with

这个节奏也在加快,部分原因是利用模型配合

The talent that we have to really accelerate ways to access this underlying intelligence.

我们拥有的人才,真正加速了访问这种底层智能的方式。

That's kind of our theory of the case on the product side.

这就是我们在产品方面的理论。

How do you think about this weird world where you mentioned the talent and the leverage and they're not writing code themselves and Claude's writing its own code?

你怎么看待这个奇怪的世界?你提到了人才和杠杆作用,他们自己不写代码,而是Claude在写自己的代码。

It seems like the last step of that would be you don't even need the talent to tell the thing what to do, it just figures out what to do on its own and that's like the ultimate, you know, the thing then just runs and is only constrained by compute or something.

似乎最后一步就是你甚至不需要人才来告诉它做什么,它自己就能弄清楚该做什么,那就是终极状态,你知道,然后它就自己运行,只受算力限制之类的。

Is that, am I being too crazy about that or is that future possible do you think?

我这样想是不是太疯狂了,还是说这个未来是可能的?

I think that the core of our company is still a research lab.

我认为我们公司的核心仍然是一个研究实验室。

I think it's maybe not as well understood, maybe it's getting more understood from the outside, but we're doing experiments.

我觉得外界可能不太了解这一点,也许现在越来越了解了,但我们在做实验。

We are doing things that push the limits of what our models can do.

我们在做的事情是推动我们模型能力的极限。

And that research and that

而这个研究和这个

Engine is upstream of everything else that we've talked about, and so that is enabled by the models today.

引擎是我们讨论的其他一切的上游,所以它是由今天的模型赋能的。

It's not entirely done by the models. Over time, we think that the models will get better.

它不是完全由模型完成的。随着时间推移,我们认为模型会变得更好。

They'll be more helpful in that process. But having the best talent to set the direction, not just the priorities, but some of the new areas of discovery, it just actually makes that research talent even better, right?

它们在这个过程中会更有帮助。但拥有最好的人才来设定方向,不仅仅是优先级,还有一些新的发现领域,这实际上让研究人才变得更好,对吧?

And so I think of it as accentuating and accelerating the talent that we already have.

所以我认为这是在增强和加速我们已有的人才。

We talk a lot about how talent density beats talent mass, and I think that's true here.

我们经常讨论人才密度比人才总量更重要,我认为这一点在这里也适用。

Like we want the densest collection of AI research talent and inference engineering talent, and that enabled with the best models we think is a really winning combination.

我们想要的是最密集的 AI 研究人才和推理工程人才的集合,再配合我们认为最好的模型,这是一个非常有竞争力的组合。

How are scaling laws talked about internally? Like the sort of consensus has been you've got different components of them.

内部是如何讨论扩展定律的?目前的共识是扩展定律有不同的组成部分。

You've got like pre-training, you've got post-training, you've got reasoning, and that all of these are kind of moving at different paces and to hit a true wall, they would all need to fall down.

包括预训练、后训练、推理能力,这些都在以不同的速度发展,要真正遇到瓶颈,需要所有这些方面都停滞不前。

Like that's sort of like how the world is now conceptualizing scaling laws.

这就是目前业界对扩展定律的理解方式。

How are they talked about internally? How do you think about them?

你们内部是怎么讨论的?你是怎么看待这个问题的?

Yeah, I mean, we look at models at various points in their development.

我们会在模型开发的不同阶段对其进行评估。

We can see, you know, during a pre-training run, how does this model compare to a prior model that we did on these kind of loss curves?

在预训练过程中,我们可以看到这个模型在损失曲线上与之前的模型相比表现如何。

And that gives us a sense for model capability.

这让我们对模型能力有了一个判断。

You can do the same thing as you think about RL.

在强化学习阶段也可以用同样的方法。

And then probably as importantly is when customers get their hands on it, like what are they seeing?

同样重要的是,当客户真正使用模型时,他们看到了什么。

Where are they identifying pain points?

他们发现了哪些痛点?

And those pain points then become like training targets for us, right? We don't train on customer data on the enterprise side.

这些痛点就成为我们的训练目标。不过在企业端,我们不会用客户数据来训练模型。

On the prosumer—

在个人消费者端——

Side, it's only if you opt in. But customers tell us things like, "Hey, I wish the model were better at this or I had this particular place where it got stuck and I could build this other product, but the capability needs to be further than that."

只有在用户选择加入的情况下才会使用。但客户会告诉我们:「我希望模型在这方面表现更好」或者「我在某个地方卡住了,我本可以开发另一个产品,但模型能力还需要进一步提升」。

What we usually tell them is, okay, build your product for that because we're going to on the R&D side improve that over time.

我们通常会告诉他们,可以先为那个场景开发产品,因为我们会在研发端持续改进模型能力。

And so there is this connected loop, but internally we're always looking at different models that are being trained, different snapshots that we have, and comparing them internally and to a lesser extent externally against our own measure and then ultimately how our customers view them as well.

所以这是一个闭环反馈机制。在内部,我们会持续观察正在训练的不同模型、不同的快照版本,在内部进行对比,也会在一定程度上与外部模型对比,既用我们自己的标准衡量,最终也看客户如何评价。

And it feels like there's just no slowdown in the scaling laws themselves. Is that a fair characterization?

感觉扩展定律本身并没有放缓的迹象。这样理解对吗?

For us that's a fair characterization. Yeah, we are extremely—I mean—

对我们来说确实如此。我们非常——我的意思是——

Obviously a bunch of the authors of the scaling laws papers are amongst our founders, but you know, notwithstanding that, we can be a bit of a skeptical bunch. Like, you know, we hold ourselves to a really high standard. Again, it's this kind of idea of a research lab that's very kind of scientific method, and people are constantly challenging previously held assumptions. But from what we see, the scaling laws are not slowing down.

显然,很多扩展定律论文的作者都是我们的创始人,但即便如此,我们也是一群相当skeptical的人。我们对自己的标准很高。这又回到了研究实验室的理念,非常注重科学方法,大家会不断挑战之前的假设。但从我们看到的情况来看,扩展定律并没有放缓。

So if that's true, you said before it's hard for humans to be exponential in their thinking and not linear. Like, if that continues to be true for however many more, you know, turns of the crank here, how do you do that thing of not thinking linear and thinking exponential yourself in your job and for the business? Like, the implications are really hard to reason through. Exponential growth rate is one thing, but exponential growth of capability—like, I don't even know how to get my head around it.

如果这是真的,你之前说过人类很难用指数思维而不是线性思维来思考。如果这种趋势继续下去,不管还要经历多少轮迭代,你如何在工作和业务中避免线性思维而采用指数思维?这真的很难推理。指数增长率是一回事,但能力的指数增长——我甚至不知道该如何理解。

So how do you get your head around it?

那你是如何理解的?

We think about the world as scenarios. It's very hard to have a point estimate in this business.

我们用场景化的方式来思考。在这个行业很难做出单点预测。

And then having a very low bar for updating your current prior or your current perspective, because it could be the case that something a month ago was true that's just not true today and that breaks your model.

然后要对更新当前认知或观点保持非常低的门槛,因为可能一个月前还成立的事情今天就不成立了,这会打破你的模型。

And so this old, like, well, we'll forecast, you know, once a quarter and we'll, you know, we'll revisit this in three months at the next board meeting. That doesn't work for our business.

所以那种「我们每季度做一次预测,三个月后的下次董事会再讨论」的老方法不适合我们的业务。

It's so dynamic that we have to always think about, oh, you know, our models couldn't do this before and they could do this now. What does that mean for the TAM? We've seen this in coding first, right, where, you know, starting with around Sonnet 3.5/3.6, we started to see this really remarkable jump in capability, which was then followed by adoption and usage and

市场变化非常快,我们必须时刻思考:我们的模型以前做不到的事情,现在能做到了,这对潜在市场规模意味着什么?我们首先在编程领域看到了这一点——从 Sonnet 3.5/3.6 开始,我们看到能力出现了非常显著的跃升,随之而来的是采用率、使用量的增长,以及

Revenue, and you know, it was a little hard to predict that, but now we can use coding as an analog for a lot of what's happening elsewhere in the economy and elsewhere in our business. And so we kind of look at pattern recognition in our own business to try to predict what's going to happen in the future.

收入的增长。虽然这有点难以预测,但现在我们可以把编程领域的情况作为类比,来理解经济其他领域和我们业务其他部分正在发生的事情。所以我们通过识别自己业务中的模式,来尝试预测未来会发生什么。

Literally 15 minutes before you got here, the news came out about your partnership with xAI in the Tennessee facility. Makes me curious about how you are canvassing the world for—like, that is an opportunity you decided to do. Like, I'm sure there's a universe of things that you've explored. What is the strategy for trying to get more in creative ways? Like, bring us a little bit more into that.

就在你到这里前 15 分钟,你们与 xAI 在田纳西州设施合作的消息刚刚发布。这让我很好奇你们是如何在全球范围内寻找机会的——比如,这是你们决定做的一个机会,我相信你们肯定探索过很多其他可能性。那么,用更有创意的方式获取更多算力的策略是什么?能不能多讲讲这方面?

We announced a partnership with xAI for their Colossus facility in Memphis. We're really excited about that. It's going to allow us to continue to expand, especially on the consumer and prosumer side. But that's just one example of us just, as you said, looking

我们宣布了与 xAI 在孟菲斯 Colossus 设施的合作。我们对此非常兴奋,这将让我们能够继续扩展,尤其是在消费者和专业消费者这一侧。但这只是一个例子,正如你所说,我们一直在寻找

For near-term compute, wherever we can get it, as the compute base grows, that near-term compute becomes a smaller and smaller fraction of what's available and what's out there.

近期可用的算力,无论在哪里能找到。随着算力基础的增长,这些近期算力在可用总量中所占的比例会越来越小。

But we look at it as can we deploy that compute that's available productively.

但我们关注的是:我们能否高效地利用这些可用的算力。

Sometimes the answer is yes and sometimes it's no, but if we can, then we look at the economic return on it based on what it's priced, what duration we have it for, where it's located, what type of compute it is, and how efficiently we can run it.

有时答案是可以,有时是不行。但如果可以的话,我们会根据价格、使用期限、所在位置、算力类型以及我们能以多高的效率运行它,来评估经济回报。

So we have a process to assess, and that same process, by the way, we use to assess longer-term deals as well.

所以我们有一套评估流程,顺便说一句,我们用同样的流程来评估长期合约。

So last month, we signed a 5 gigawatt deal with Google and with Broadcom for TPUs starting in 2027.

上个月,我们与 Google 和 Broadcom 签署了一份 5 吉瓦的协议,用于从 2027 年开始使用 TPU。

We also signed a deal with Amazon for Trainium for up to 5 gigawatts as well. It was an over hundred billion dollar commitment.

我们还与 Amazon 签署了一份最高 5 吉瓦的 Trainium 协议,这是一笔超过千亿美元的承诺。

And a lot of that compute is actually already

而且这些算力中的很大一部分实际上已经在

Landing and will land in the rest of this year into next year. And so if you think about it, it's a bit of this layer cake of compute that's starting at different times with different capabilities, and we're very dynamically comparing that compute.

交付中,并将在今年剩余时间到明年陆续到位。所以你可以把它想象成一个分层的算力结构,不同层在不同时间开始,具有不同的能力,我们会非常动态地比较这些算力。

It's price performance over time that's really, really important to us when it lands and what we think we can do with it internally in the business.

对我们来说,真正重要的是随时间变化的性价比——算力何时到位以及我们认为能在业务内部用它做什么。

And so there's so many different variables you have to optimize for around, you know, what compute it is, at what cost, and over what time horizon.

所以有很多不同的变量需要优化,包括是什么算力、成本多少、以及时间跨度有多长。

But we have a pretty dynamic way of looking at kind of near-term compute and then medium to long-term compute.

但我们有一套相当动态的方法来看待近期算力和中长期算力。

But the things we're assessing are largely the same. What is different is just the time horizon.

我们评估的内容基本相同,不同的只是时间跨度。

What about the trade-off? You said price per performance. The trade-off between like cost per token or something, throughput and speed.

那权衡取舍呢?你提到了性价比。比如每个 token 的成本、吞吐量和速度之间的权衡。

From the customer perspective, they care about both speed

从客户角度来看,他们两者都在意

Probably unlocks some capability and use cases that are really interesting that we don't know about yet as these things get faster.

速度可能会解锁一些我们还不知道的、非常有趣的能力和用例,随着这些东西变得越来越快。

Can you talk a little bit about that trade-off in compute as you're assessing it?

你能谈谈在评估算力时这种权衡吗?

As we look across three different chip platforms, we also have multiple generations of chips within it, right?

当我们审视三个不同的芯片平台时,每个平台内部还有多代芯片,对吧?

So it could be TPU, V5e and V6 and V7 and Trainium 2, Trainium 3, all of them are at different places on the price performance curve.

可能是 TPU V5e、V6、V7,以及 Trainium 2、Trainium 3,它们都处于性价比曲线上的不同位置。

And then we importantly look at how we will utilize it, right?

然后我们还会重点关注如何利用它,对吧?

Price performance is important because of efficiency.

性价比很重要,因为关系到效率。

Speed is also important for certain use cases as well.

对于某些特定用例来说,速度也同样重要。

So we look at the compute down to a very granular level in terms of what it can deliver for us and when, and that's something that we do. Again, our compute team leads that, but we closely collaborate across the business to say where do we...

所以我们会非常细致地审视算力,看它能在什么时候为我们提供什么样的能力。这是我们一直在做的事情。我们的算力团队主导这项工作,但我们会与整个公司密切协作,确定我们需要在哪些地方...

We need this compute and for what, right? Okay, we might need, you know, CPUs for RL, we might need this, you know, more leading edge compute, and we're going to deploy it for, you know, our best and fastest models or for training them. And so from our perspective, it's customer demand, but it's also really down quite granular in terms of what is each chip best for and then what will we have.

我们需要这些算力来做什么,对吧?比如,我们可能需要 CPU 来做强化学习,可能需要更先进的算力,然后我们会把它部署到最好最快的模型上,或者用来训练这些模型。所以从我们的角度来看,这既取决于客户需求,也要非常细致地考虑每种芯片最适合做什么,以及我们将拥有什么样的算力。

I'm always so curious by the metabolism of, in this case, Anthropic for new compute. Like, how fast could you take it? If I airdropped on you twice the compute that you have tomorrow, like, would you consume that? How fast would you consume that? If I airdropped 10 times the compute on top of you, how fast would you consume it? Can you calibrate us on that sort of thing? Like, is demand—actually, it feels like demand's unlimited between these three, you know, uses: training, internal, customer demand. Everyone's saying the same thing. Shortages everywhere, memory...

我一直很好奇 Anthropic 对新算力的「代谢速度」。比如,你们能多快消化它?如果我明天空投给你们两倍于现有的算力,你们会用掉吗?多快能用完?如果我空投 10 倍的算力给你们,你们多快能消化掉?能不能给我们一个概念?感觉需求是无限的——训练、内部使用、客户需求这三个方面,大家都在说同样的话:到处都缺算力,内存...

Stocks, you know, mooning. Is it that extreme that like if you 2x'd or 5x'd or 10x'd the amount available to you tomorrow, you would just like more or less instantly consume it?

股价飙升。真的有那么极端吗?就是说如果明天你们可用的算力翻 2 倍、5 倍或 10 倍,你们会或多或少立刻就用完?

This goes back to like how we use it and the fungibility of it. So the answer is, you know, we're constrained kind of across those use cases internally today.

这要回到我们如何使用算力以及算力的可替代性。答案是,目前我们在内部的这些使用场景中确实都受到算力限制。

And you know I would say that you know a year or two ago it would have been harder to consume especially like a heterogeneous kind of compute drop in your example really quickly because these chip platforms are different and they are different, some are harder to operate, some of them have you know idiosyncrasies in terms of how we use it.

我想说,一两年前,要快速消化算力会更困难,特别是像你举例中那种异构算力的突然增加,因为这些芯片平台是不同的,它们确实不同——有些更难操作,有些在使用方式上有特殊性。

I would say today that you know getting a bunch more compute I think it would be deployed you know very rapidly across those different use cases.

但我想说,在今天,如果突然获得大量额外算力,我认为它会非常快速地部署到那些不同的使用场景中。

We probably have the same kind of allocation or calibration that we do with compute today but it's

我们可能会采用与今天类似的算力分配或调配方式,但是

It's become a lot easier for us to spin up very quickly and deploy like almost any type of compute, and that's something again we think is a real advantage.

对我们来说,快速启动并部署几乎任何类型的算力已经变得容易多了,这也是我们认为的一个真正优势。

Back to the ways the customers are using Anthropic, one of the interesting tensions and trade-offs that I'm fascinated to hear how you think through is between sort of the platform approach where I build my business on top of Claude and it powers my thing versus you doing the thing that I wanted to build.

回到客户使用 Anthropic 的方式,有一个有趣的张力和权衡让我很想听听你们是怎么思考的:就是平台方式——我在 Claude 之上构建我的业务,它驱动我的产品——与你们直接做我想做的事情之间的权衡。

So this is like the classic, you know, Canva versus Figma or something like this.

这就像经典的 Canva 对比 Figma 的例子。

How do you think about the right balance of how deep into the application layer you should go versus just being a pure enabling layer of like, we're going to provide the reasoning engine and the intelligence and, you know, world go forth and build whatever you want, you know, pay us through the API or whatever? That seems like a fascinating internal discussion and tension to some extent.

你们如何看待应该深入到应用层多深,还是只做一个纯粹的赋能层——就是说,我们提供推理引擎和智能,然后大家去构建任何想要的东西,通过 API 付费之类的?这似乎是一个很有意思的内部讨论,在某种程度上也是一种张力。

Yeah, the way I would think about it is most of what we're building is platform, and we think that there's so many examples of where a platform can accrue a lot of value, but the customers who are building on that platform actually accrue even more value.

是的,我的想法是,我们构建的大部分是平台,我们认为有很多例子表明平台可以积累大量价值,但在该平台上构建的客户实际上会积累更多价值。

We think that's what we're setting up for today.

我们认为这就是我们今天正在建立的模式。

It's maybe akin to the early days of AWS, right?

这可能类似于 AWS 的早期阶段,对吧?

If you think about the cloud platform and all the tools and services that are now built in—because it's not just the raw model access, it is, you know, prompt caching and the ability to use virtual machines and, you know, cloud code being called within there or dispatch or the cloud agents SDK, managed agents.

如果你想想云平台以及现在内置的所有工具和服务——因为这不仅仅是原始的模型访问,还包括提示词缓存、使用虚拟机的能力、在其中调用 cloud code、dispatch、cloud agents SDK、托管代理等等。

All of these are effectively, I think of as vectors to access that model intelligence for other companies to build into their own products.

所有这些实际上,我认为都是访问模型智能的向量,让其他公司能够将其构建到自己的产品中。

That's most of what we're focused on and really most of where we think the business is going from where we are today.

这是我们关注的重点,也是我们认为从现在开始业务发展的主要方向。

That said,

话虽如此,

We will also build our own applications on that same platform where a couple things are true.

我们也会在同一个平台上构建我们自己的应用,前提是满足几个条件。

Number one, if we feel like we have a vision into where the models are going and we can kind of demonstrate that and create customer value in that, that might be something like Claude Code, right?

第一,如果我们觉得我们对模型的发展方向有洞察,并且能够展示这一点并创造客户价值,那可能就是像 Claude Code 这样的产品,对吧?

We were able to say actually a lot of what's out there in the market was developer-led. Claude Code is a platform that's Claude-led and we think the models couldn't quite do that today when it was launched a little over a year ago, but we think they'll get there and they have.

我们能够说,实际上市场上的很多产品都是开发者主导的。Claude Code 是一个由 Claude 主导的平台,我们认为在一年多前刚推出时,模型还做不到这一点,但我们认为它们会达到那个水平,而且它们确实做到了。

And so one is kind of building ahead to model capabilities.

所以第一点是提前构建以适应模型能力的发展。

The second is thinking about ways to demonstrate value for the ecosystem that others might emulate.

第二点是思考如何为生态系统展示价值,让其他人可以效仿。

Right? If you think about Claude for financial services or Claude for life sciences or even something like Claude security, these are ways in which we've kind of composed the platform.

对吧?如果你想想 Claude for financial services、Claude for life sciences,甚至像 Claude security 这样的产品,这些都是我们组合平台能力的方式。

Again, we're building on the same platform as our customers and we think that creates like a level playing field.

我们和客户使用同一个平台来构建产品,这创造了一个公平的竞争环境。

We also think that there's so much value that's going to accrue in some of these areas that, you know, our customers can win and we can win as well, which is why you've seen as we've launched some of these products, we've done them in a collaborative kind of partnership-oriented way, whether that be on the security side or, you know, design or financial services—we've partnered with the ecosystem.

我们认为这些领域会产生巨大的价值,客户能赢,我们也能赢。所以你会看到我们推出这些产品时,都采用了合作伙伴导向的方式——无论是安全、设计还是金融服务领域,我们都与生态系统合作。

So I think of our strategy as mostly horizontal. We'll build vertical where, you know, we think we have some value to add or a perspective that's useful or a way to demonstrate to the market how we think about our platform adding value, and a lot of the value is going to accrue to the customers that are building on top of it.

我们的策略主要是横向的。只有在我们认为能增加价值、提供有用视角,或者能向市场展示我们平台如何创造价值时,才会做垂直产品。大部分价值会流向在平台上构建应用的客户。

Our goal is to build the best models and then build the products and tools.

我们的目标是构建最好的模型,然后开发相应的产品和工具。

And services that allow that intelligence to proliferate within customers.

以及让这种智能能力在客户内部广泛应用的服务。

How much do you care that it's just a reality that people are scared of you, that there's a sense that because you control the most essential piece of these new applications, the underlying intelligence reasoning engine, that may totally be true and maybe already is true, that more of the value is occurring on top of the Anthropic platform than is being captured by it.

你在意吗——人们确实害怕你们,因为你们控制着这些新应用最核心的部分,也就是底层的智能推理引擎。可能确实如此,甚至已经是现实:Anthropic 平台上产生的价值比平台本身捕获的更多。

But nonetheless, it's still scary to imagine, and I guess maybe you could say something similar about cloud and AWS or something like that.

但不管怎样,这种想象还是让人害怕。也许你可以说云计算和 AWS 也有类似情况。

But how much do you think and care about the fact that some of your would-be customers or existing customers are in fact scared of you as a competitor?

你有多在意这个事实——你的潜在客户或现有客户实际上把你当作竞争对手而感到害怕?

Part of what is hard in this business is it's changing so quickly, so the model capabilities

这个行业的难点之一是变化太快,模型的能力有时连我们自己都感到意外。

sometimes even surprise us, and so when we release models or products on top of

所以当我们发布模型或基于模型的产品时——

That like there is an element of what's happened, you know, in prior waves over the course of 5 years, 10 years, 20 years—it's happening in months now.

以前需要 5 年、10 年、20 年才发生的变化,现在几个月就发生了。

And when we release things, people are also surprised by it in some ways, in the same way that we were surprised by it.

我们发布新东西时,人们也会感到惊讶,就像我们自己当初惊讶一样。

But I think fundamentally what we are trying to do is be very partner-oriented towards the ecosystem.

但从根本上说,我们努力以合作伙伴为导向来对待整个生态系统。

And that means that, you know, we have early access programs. We work very closely with customers. We listen to them about what capabilities they want.

这意味着我们有早期访问计划,与客户紧密合作,倾听他们想要什么能力。

That doesn't mean that the things we release are sometimes not moments where you're like, "Wow, that's way more powerful than I thought it would be," or "I didn't realize the models would be able to do that this quickly."

但这不代表我们发布的东西不会让人惊叹「哇,比我想象的强大太多了」或「没想到模型这么快就能做到这个」。

I think that part of that is a reality of where we are in this cycle and in this kind of development of intelligence.

我认为这部分是我们所处的这个周期和智能发展阶段的现实。

But part of it is also we want to make those capabilities really accessible, and that—

但另一部分原因是我们想让这些能力真正易于获取——

should accrue a lot of value to customers as well, and customers that are forward-footed on that and adopt, and frankly also ones that are building and using the tools that we offer on our platform.

这会给客户带来很多价值,尤其是那些积极采用的客户,以及那些使用我们平台提供的工具来构建的客户。

We think we can actually accelerate them.

我们认为可以真正加速他们的发展。

I think some of it's a reality of kind of frontier model development, but our approach to it is probably a little different and more partner-oriented.

我认为这部分是前沿模型开发的现实,但我们的做法可能有点不同,更注重合作伙伴关系。

You said before going 9 to 30 in the first quarter.

你之前说第一季度从 9 个客户增长到 30 个。

The pace is so insane, which makes me wonder about pricing—like the dynamic of how to price tokens or use of the system is so fascinating to me, because I think a lot of people a year ago would say price is going to just constantly fall.

这个速度太疯狂了,这让我想到定价问题——如何给 token 或系统使用定价的动态变化让我很着迷,因为我觉得一年前很多人会说价格会持续下降。

But actually what's happening is it's going up in many cases, and this is true at different levels, whether it be the Mythos pricing that is quite high because it's so powerful, the cost of an H100, the rental price of

但实际上很多情况下价格在上涨,这在不同层面都是如此,无论是因为功能强大而定价很高的 Mythos,还是 H100 的成本,或者——

The cost of an H100 is, well, you know, looks like a smile curve. I'm very curious why, if everyone is compute constrained, why everyone doesn't just raise prices a lot to try to find like what the right equilibrium is. And so I'd love you to just like riff on pricing, like how you think about it, what the trade-offs are, why not raise prices a lot.

H100 的成本,嗯,你知道,看起来像一条微笑曲线。我很好奇,如果每个人都受算力限制,为什么不大幅提价来找到合适的平衡点。所以我想听你聊聊定价——你怎么考虑的,有什么权衡,为什么不大幅提价。

The company is only a little over 5 years old. This past March was the third anniversary of the first dollar of revenue into the business, and we only had a frontier model for real for the first time in March of 2024. So the time scale of these things is kind of, it's an important backdrop.

公司成立才五年多一点。今年三月是公司获得第一笔收入的三周年,而我们真正拥有前沿模型是在 2024 年 3 月才第一次实现。所以这些事情的时间尺度是很重要的背景。

Our pricing has been relatively stable across, you know, Haiku, Sonnet, and Opus, and now Mythos is obviously newer, but you know, we made very few pricing changes. The biggest pricing change we made was to bring down the price of the Opus family when we launched Opus 4.5. And if we thought about why did we do that,

我们的定价在 Haiku、Sonnet 和 Opus 这几个模型上一直相对稳定,现在 Mythos 显然是新推出的,但我们很少调整价格。我们做过的最大一次价格调整是在推出 Opus 4.5 时降低了 Opus 系列的价格。如果要说我们为什么这么做,

It's really because we found that Opus-class models were underutilized relative to their capability, right?

实际上是因为我们发现 Opus 级别的模型相对于它的能力来说使用率偏低,对吧?

And so people were trying to often fit an Opus problem into a Sonnet workload.

所以人们经常试图把本该用 Opus 解决的问题硬塞进 Sonnet 的工作负载里。

And because of the efficiency improvements that we were able to make, we were able to serve that very efficiently from our perspective, but actually bring down the price which made it more accessible to customers.

由于我们在效率上取得了改进,从我们的角度来看可以非常高效地提供服务,同时还能降低价格,这让客户更容易使用。

And so it goes back a little bit to we want our customers to generate a lot of value from it and they're generating a ton of ROI from our models today.

这又回到了一点,我们希望客户从中获得大量价值,而他们现在确实从我们的模型中获得了巨大的投资回报。

We want that to just continue because our goal is to proliferate this throughout the ecosystem.

我们希望这种情况持续下去,因为我们的目标是让这项技术在整个生态系统中普及开来。

We think we're in the very, very early innings on all of these use cases.

我们认为在所有这些应用场景上,我们还处于非常非常早期的阶段。

The best way to do that is to get this intelligence in the hands of as many businesses from startups to digitally native businesses to the largest companies in the world.

最好的方式就是把这种智能能力交到尽可能多的企业手中,从初创公司到数字原生企业,再到世界上最大的公司。

That means that you have to make it in a price point that's accessible and that allows them to get a lot of value from it.

这意味着你必须把价格定在一个可接受的点上,让他们能够从中获得大量价值。

The changing of the pricing for Opus actually, you know, you see this Jevons paradox, right?

Opus 的价格调整实际上体现了杰文斯悖论,对吧?

Like we lowered the price of it, but the consumption went up way, way more than what you would have expected.

我们降低了价格,但消费量的增长远远超出了你的预期。

And so because we kind of hit that sweet spot for customers, they were able to use it a lot more.

因为我们找到了客户的甜蜜点,他们能够更多地使用它。

We had the efficiency to be able to serve it to customers at scale.

我们有足够的效率能够大规模地为客户提供服务。

And then they were able to build that into their workload such that when we released Opus 4.6, it's a model improvement.

然后他们就能把它整合到工作负载中,这样当我们发布 Opus 4.6 时,这是一次模型改进。

They can slot it in. We didn't change the price.

他们可以直接替换进去。我们没有改变价格。

And so we think pricing stability is important.

所以我们认为价格稳定性很重要。

And we also think that pricing to get that value and to see that kind of Jevons paradox happen is really important.

我们也认为通过定价来获得那种价值,并看到杰文斯悖论发生,这真的很重要。

The other component of this is margins and how you think about margins as a business, again, because this...

另一个要素是利润率,以及你作为一家企业如何看待利润率,因为这个...

It's so unbelievably capital intensive to build these frontier labs. You've got the leverage we talked about, which is efficiency, price—both those things relate to margin.

建立这些前沿实验室需要极其庞大的资本投入。我们谈到的杠杆就是效率和价格——这两者都与利润率有关。

I apologize if there's a naive perspective, but given how much capital you need, why not just say we want to have a healthy margin and sort of set the price accordingly, and maybe that price can come down if efficiency is better or whatever? And so I'm curious how you think about margins as it relates to pricing in the business.

如果这个问题太天真请见谅,但既然需要这么多资本,为什么不直接说我们想要一个健康的利润率,然后相应地设定价格,如果效率提高了价格也许可以降下来?所以我很好奇你如何看待利润率与定价和业务的关系。

Yeah, I would say we think about what is the return on our compute spend, right, writ large. So that is all of the different workloads that we've talked about, whether it's serving customers, model development. If you think of all of those, they're kind of in support of revenue over different time scales, right? If I serve inference, it's in support of revenue today. If I do model development, it might help for a

是的,我会说我们考虑的是算力支出的整体回报率。这包括我们讨论过的所有不同工作负载,无论是为客户提供服务还是模型开发。如果你把所有这些都考虑进去,它们在不同的时间尺度上都是在支持收入,对吧?如果我提供推理服务,那是在支持今天的收入。如果我做模型开发,可能会帮助

Capability that unlocks TAM that drives revenue 6 months from now and everything in between.

解锁某种能力,从而扩大潜在市场规模,推动 6 个月后的收入,以及介于两者之间的一切。

If I do internal acceleration to launch a new product, all of these things are in support of that.

如果我做内部加速来推出新产品,所有这些都是在支持收入。

I will say our returns on that compute expense today are robust.

我要说的是,我们今天在算力支出上的回报是稳健的。

They're robust and we think of it as what is the return on that full envelope of compute.

它们很稳健,我们把它看作是整个算力投入的回报率。

And so we feel really good about where we are from that perspective.

所以从这个角度来看,我们对目前的状况感到非常满意。

And we're balancing delivering value to customers with also seeing a really strong return on that compute ourselves.

我们在为客户提供价值的同时,也在算力投入上获得了非常可观的回报,两者之间保持着平衡。

And if you think about when revenue grows, as we mentioned in Q1, it's not like we onboarded a bunch of new compute in that time period.

如果你想想收入增长的情况,就像我们在第一季度提到的,并不是说我们在那段时间里突然增加了大量新算力。

We talked about compute comes based on a ramp that might have been determined 12 months ago.

我们讲过,算力的部署是基于一个爬坡计划,这个计划可能在12个月前就已经确定了。

And so this idea of a variable cost that's like on the incremental to serve a customer is a little bit like it doesn't really fit our model.

所以那种认为每服务一个客户就会产生增量可变成本的想法,其实并不太适用于我们的模式。

business, right? It tries to maybe fit our business into like a software paradigm, but that's not the case.

对吧?这种想法试图把我们的业务套进软件行业的范式里,但实际情况并非如此。

And in actuality, compute is supporting all of these activities and we're really generating a robust return on that compute and that's our measuring stick.

实际上,算力支撑着所有这些业务活动,而我们在这些算力上获得了非常稳健的回报,这才是我们的衡量标准。

And so, you know, I think it's something where we think of the compute envelope that we have as the thing that is able to govern how much we're able to drive revenue both over the short term and the long term.

所以我认为,我们把现有的算力规模看作是决定我们能在短期和长期内推动多少收入增长的关键因素。

So if you're this great customer of the compute providers, what does that group need to do to be a great provider to you to help you drive that return?

既然你们是算力供应商的优质客户,那么这些供应商需要做些什么才能成为你们的优质合作伙伴,帮助你们实现那样的回报呢?

So, we're fortunate in that we have really great partners in Amazon, in Google, in Microsoft, but also with Broadcom and Nvidia as well.

我们很幸运,在Amazon、Google、Microsoft,以及Broadcom和Nvidia那里都有非常好的合作伙伴。

And so, our ecosystem, you know, we are the only model that's on all three clouds today. We're the only language

在我们的生态系统中,我们是目前唯一一个在三大云平台上都部署的模型,也是唯一一个

Lab that's using all three of these chip platforms, and really these collaborations are much deeper than just like procurement.

同时使用这三种芯片平台的语言实验室,而且这些合作的深度远不止采购那么简单。

I think that's something that's often lost.

我觉得这一点经常被忽视。

If you think about our relationship with Amazon, you know, our teams are deeply embedded with the Annapurna Labs team.

比如我们和Amazon的关系,我们的团队与Annapurna Labs团队深度合作。

You know, we are really good users of Trainium. We've spent a lot of time and energy and work closely with the team internally.

我们是Trainium的优秀用户,投入了大量时间和精力,与他们的内部团队紧密协作。

We plan capacity together, right?

我们一起规划算力容量,对吧?

If you think about the three clouds, they're great distribution engines for us too.

如果你想想这三大云平台,它们对我们来说也是很好的分发渠道。

We have a really robust first-party business as well.

我们自己的第一方业务也很强劲。

But these are multifaceted partnerships, whether it be on developing the chips themselves, landing that capacity, serving it, and then ultimately distributing it to the customers.

但这些都是多方面的合作关系,涵盖芯片本身的开发、算力容量的落地、服务的提供,以及最终向客户的分发。

I'm thinking about your function, like the finance team and the ways that you might—I'm picturing this like ROI on

我在想你的职能,比如财务团队以及你们可能采用的方式——我脑海中浮现的画面是关于算力投资回报率的

Compute things on different horizons with all these complex variables, which makes me wonder: how do you use these powerful tools yourself internally to run your group and the business? Like, what is the deployment of Claude Code and Claude in general on the finance team at Anthropic?

在不同时间维度上有各种复杂变量的计算,这让我好奇:你们内部是如何使用这些强大的工具来运营团队和业务的?比如,Claude Code和Claude在Anthropic财务团队的部署情况是怎样的?

Yeah, so this is really interesting because we were using Claude Code, you know, about a year ago, and I started asking people like, "Is everyone just vibe coding?" And we started to use Claude Code as almost like an assistant, a digital co-worker, not just for coding tasks. And that actually was early in what eventually became CoWork, right? That was kind of an extension of Claude Code to say what it's done for agentic software development, it should do for all of knowledge work.

这个问题很有意思,因为我们大约一年前就开始使用Claude Code了,当时我开始问大家「大家都在凭感觉写代码吗?」后来我们开始把Claude Code当作一个助手、一个数字同事来使用,不仅仅用于编码任务。这实际上是CoWork的早期形态,对吧?就是把Claude Code在智能体软件开发中的作用延伸到所有知识工作领域。

But then we started to productionize that, and I'm actually really proud. We spend a lot of time with our product team too. They kind of see how we use it and get input and feedback from that.

然后我们开始将其产品化,我真的很自豪。我们也花了很多时间和产品团队合作,他们会观察我们如何使用,并从中获取意见和反馈。

But like today, you know, all of our legal entities, we can produce the statutory financial statements using Claude. And yes, a human checks it, but all of those financial statements are produced with Claude.

但就像今天,我们所有法律实体的法定财务报表都可以用Claude生成。当然,会有人工审核,但所有这些财务报表都是由Claude制作的。

We also have a more real-time platform called Ant Stats. It used to take a lot of time to sift through all the data, get to conclusions, write a memo about it or publish a regular report on what's happening over the course of the day, what's driving it.

我们还有一个更实时的平台叫Ant Stats。过去需要花很多时间筛选所有数据、得出结论、写备忘录或发布关于一天中发生了什么、是什么在驱动变化的定期报告。

We now have a library of skills for Claude that are specific to finance. I think last I checked there were over 70 of them that everyone can access through this common repository.

我们现在为 Claude 建立了一个专门针对财务领域的技能库。我上次查看的时候,里面已经有超过 70 个技能,所有人都可以通过这个共享代码库访问使用。

And on top of that we built an MFR, a monthly financial review skill, and it can produce our monthly financial review. It's 90 to 95% ready and then all of our discussion becomes about what do we do, what are the implications, not what exactly happened because Claude is

在此基础上,我们还构建了一个 MFR(月度财务回顾)技能,它可以生成我们的月度财务报告。这个报告能达到 90% 到 95% 的完成度,然后我们所有的讨论就可以集中在「我们该做什么、这意味着什么」,而不是「到底发生了什么」,因为 Claude 已经帮我们处理好了。

Not just reporting the weather, it's also helping to think about drivers and like why did the number change in the way it did, and that gives you tremendous insight into the business, both in terms of this like MFR that we do, but also on a daily basis. And so what used to take hours to produce, you know, a weekly report for, you know, what's driving revenue or what's driving our compute utilization, now comes down to 30 minutes, and then we can spend our time on the actual strategic implications of the business.

Claude 不只是报告天气那样简单地呈现数据,它还会帮助思考驱动因素,比如为什么某个数字会以这种方式变化。这让你对业务有了深刻的洞察,不仅体现在我们做的这种 MFR 上,也体现在日常工作中。以前需要花几个小时才能完成的周报,比如分析收入驱动因素或计算资源利用率的报告,现在只需要 30 分钟,然后我们就可以把时间花在真正的战略层面思考上。

We can also get it in the hands of business, you know, leaders much more quickly, and so it's just meant that the insight engine is a lot faster within the company. We also have like, you know, I have a dashboard I look at token usage across...

我们也能更快地把这些洞察交到业务负责人手中,所以公司内部的洞察引擎运转得快多了。我还有一个仪表板,可以查看整个公司的 token 使用情况……

The leaderboard.

排行榜。

Yeah, we don't compensate people on it. No one's trying to token max for that, but it's really interesting because some of the most senior people...

对,我们不会根据这个给人发奖金。没人会为了上榜而刻意刷 token 使用量,但有趣的是,财务团队里一些最资深的人……

Within the finance team are actually the biggest users of tokens, so it is not just, you know, the 22-year-old who joined and has a coding background and was doing that on the weekends and brought it to work.

实际上是 token 使用量最大的用户。所以这不只是那种 22 岁刚入职、有编程背景、周末就在玩这些工具然后带到工作中的年轻人在用。

It's also people using the tools to change how they're working.

也有很多人在用这些工具改变自己的工作方式。

Like, I think our number one user is our head of tax, and he's, you know, really focused on tax policy engines and automating large parts of the workloads that are happening within the team.

比如我们使用量排第一的是我们的税务主管,他非常专注于税务政策引擎,并且在自动化团队内部的大量工作流程。

So I love seeing that, and I think I tell people, if we're not super users of this, if we're not pushing the limits of it, how can you expect customers to do that?

我很喜欢看到这种情况。我跟大家说,如果我们自己都不是这个工具的超级用户,如果我们都不去突破它的极限,你怎么能指望客户去这么做呢?

As your business scales up, everything gets more complex, especially your compliance and security needs. With so many tools offering band-aids and patches, it's unfortunately far too easy for something to slip through the cracks. Fortunately, Vanta is a powerful tool designed to simplify and automate your security work and deliver a single

随着你的业务规模扩大,一切都会变得更加复杂,尤其是合规和安全需求。市面上有太多工具只提供临时性的修补方案,很容易就会有东西从缝隙中漏掉。幸运的是,Vanta 是一个强大的工具,旨在简化和自动化你的安全工作,并为合规和风险提供单一的……

Source of truth for compliance and risk. There's a reason that Ramp, Cursor, and Snowflake all use Vanta. It frees them to focus on building amazing differentiated products knowing that compliance and security are under control.

真实来源。Ramp、Cursor 和 Snowflake 都在使用 Vanta 是有原因的。它让这些公司能够专注于打造出色的差异化产品,因为他们知道合规和安全已经得到了妥善管理。

Invest Like the Best listeners get a special offer of $1,000 off Vanta when you go to vanta.com/invest.

Invest Like the Best 的听众可以获得特别优惠,访问 vanta.com/invest 可以享受 1000 美元的折扣。

I know firsthand how complex the tech stack is for asset management firms. And seemingly every new tool and data source makes the problem even worse, adding more complexity, more headcount, and more risk.

我深知资产管理公司的技术栈有多复杂。而且似乎每一个新工具和新数据源都会让问题变得更糟,增加更多复杂性、更多人力需求和更多风险。

Ridgeline offers a better way forward. One unified platform that automates away that complexity across portfolio accounting, reconciliation, reporting, trading, compliance, and more, all at scale.

Ridgeline 提供了一条更好的前进道路。一个统一的平台,可以大规模自动化处理投资组合会计、对账、报告、交易、合规等各方面的复杂性。

Ridgeline is revolutionizing investment management, helping ambitious firms scale faster, operate smarter, and stay ahead of the curve.

Ridgeline 正在革新投资管理行业,帮助有雄心的公司更快扩张、更智能运营,并保持领先地位。

See what Ridgeline can unlock for your firm. Schedule a demo at

看看 Ridgeline 能为你的公司带来什么。在 ridgeline.ai 预约演示。

ridgeline.ai. Just as a human, does it freak you out at all that all of these—I've heard so many examples like this.

作为一个人类,这些事情会不会让你感到有点不安?我听到过太多这样的例子。

It starts to feel like we just start doing the stuff that AI tells us to do, like in the sales example or the calendar or whatever.

感觉我们开始只是做 AI 告诉我们去做的事情,比如销售的例子,或者日历管理之类的。

And maybe that's great. Maybe it's just such a better coordinator and, you know, wide thinker and optimizer than we ever could be that we should do what it tells us to do.

也许这很好。也许它就是比我们更好的协调者、更广阔的思考者和优化器,所以我们应该按它说的做。

But it feels like ever so slightly dystopian to me that that reality is coming quickly.

但对我来说,这种现实正在快速到来,感觉有那么一点点反乌托邦的味道。

I've had examples of it, too. And it feels kind of cool. Like it's helpful, but at the same time, if I really close my eyes, like, oh, I'm just like doing what it tells me versus me telling it what to do.

我自己也有过这样的经历。感觉挺酷的,很有帮助,但同时,如果我真的闭上眼睛想一想,会觉得「哦,我只是在做它告诉我做的事」,而不是我告诉它该做什么。

It's a really interesting human dynamic that I'm curious for your take on.

这是一个很有意思的人类动态变化,我很好奇你对此的看法。

I maybe have a slightly different view on it in that I think like it has made—you know, we've been able to hire great

我的看法可能有点不同。我觉得它让我们能够在公司招到优秀的人才……

People at the company, but it has made even those incredibly talented people so much more productive, and there's a little bit of this—I think of it again like Jevons paradox, but for labor—which is that we have people who become incredibly more productive. We actually, we've hired a lot more people because of that, because there's no shortage of work to do. And now with the assistance of Claude, you know, people are spending less time in that MFR trying to reconcile some number, but they're actually thinking, oh, how do we reinvest this in the business?

但它也让这些极具天赋的人变得更加高效。这有点像 Jevons 悖论在劳动力上的体现——人们的生产力大幅提升后,我们实际上雇佣了更多人,因为要做的工作永远不会短缺。现在有了 Claude 的帮助,人们在 MFR 上花更少时间去核对某个数字,而是真正在思考「我们如何把这些节省下来的时间再投资到业务中」。

How do we think about, you know, kind of dynamically allocating resources? Whereas before I'm working to tie out a number, or I'm, you know, in that accounting example, taking a long time to close the books.

我们如何动态地分配资源?而以前我需要花时间去核对数字,或者像会计的例子那样,花很长时间才能结账。

So I actually think of it, you know, maybe even more optimistically, that it is an accelerant to our productivity, and that actually means that we can get a

所以我其实更乐观地看待这件事,它是我们生产力的加速器,这意味着我们能完成更多工作。

A lot more done, and that even as we grow the team, those people are more productive as well as they come up the curve on how to use Cloud within our company, and I think that's starting to be true across many companies as well.

即使团队在扩大,新成员在学习如何在公司内使用 Cloud 的过程中,他们的生产力也会更高。我认为这在很多公司都开始成为现实。

I'd love to talk about investors and capital formation. Of course, you've had to raise tons and tons of capital. At the same time, it seems as though, like, if I just squint my eyes and think about the multiple on current revenue, it's like not that crazy in terms of like where you're raising money.

我很想聊聊投资人和资本形成的话题。你们当然筹集了大量资金。但同时,如果我眯着眼睛看当前收入的估值倍数,你们融资的估值其实也没那么夸张。

I'm so curious for you to teach us about what it's been like to interact with investors, like how you've seen their understanding of the company evolve and mature. Where do you think the investors as a group kind of like understand it now? Where are their misunderstandings about Anthropic? Tell us that side of your life.

我特别好奇你能不能跟我们讲讲和投资人打交道是什么体验,他们对公司的理解是如何演变和成熟的。你觉得投资人群体现在对 Anthropic 的理解到了什么程度?他们对 Anthropic 还有哪些误解?跟我们聊聊你生活的这一面。

So I joined the company about two years ago. We were closing our Series D.

我大约两年前加入公司,当时我们正在完成 D 轮融资。

At the time, you know, look, that was not a straightforward fundraising. The company really only had a frontier model in the middle of that fundraising. Towards the tail end of it, the FTX transaction was happening, which was liquidating a bunch of Anthropic shares, and so that was kind of the starting point. And at that point, the questions were around like, why do you need to have a frontier model? Like, what's the returns to this?

那次融资并不顺利。公司当时只有一个前沿模型,而且在融资快结束时,FTX 的交易正在进行,清算了一批 Anthropic 的股份,这就是起点。那时候投资人会问,为什么你们需要有前沿模型?这能带来什么回报?

They were also around, you know, our mission and how we approach things. People said, well, hey, aren't AI safety and building a really big business—aren't those things at odds? And there were also a lot of other misconceptions: your sales force is really small, don't you need to scale it like all these enterprise software companies? And so there was just a bit of a paradigm around trying to fit us, you know, into a particular mold that had existed before.

他们也会质疑我们的使命和做事方式。有人说,AI 安全和建立大型商业公司,这两件事不是矛盾的吗?还有很多其他误解:你们的销售团队这么小,难道不需要像其他企业软件公司那样扩大规模吗?所以当时有一种范式,就是试图把我们套进某个已有的模式里。

Over time, you know, it's—

随着时间推移,情况发生了变化。

evolved. We at the end of 2024 we raised the Series E. You know the business had scaled to close to a billion dollars of run rate revenue, but the day of our first close was the day the DeepSeek news came out.

2024 年底我们完成了 E 轮融资。业务已经扩展到接近 10 亿美元的年化收入,但我们首次交割的那天,正好是 DeepSeek 新闻爆出的那天。

Obviously we got the close done, but certainly a ton of volatility as people then said wait a minute, should I just totally rewrite how I think about AI in total? And so that was a Series E.

我们显然完成了交割,但市场出现了巨大波动,人们开始说,等等,我是不是应该完全重新思考对整个 AI 行业的看法?这就是 E 轮的情况。

Obviously, we brought on great investors across all of these, but people still had some of those questions, but they looked at our forecast and they thought,

我们在所有这些轮次中都引入了优秀的投资人,但人们仍然有一些疑问。不过他们看了我们的预测后会想:

okay, you know, I get it. You've grown to a billion dollars of run rate revenue so quickly, but there's no way you're going to be able to keep up. Yeah, that's just not possible, right? And there's laws of physics. You're in enterprise, which is great, but the

好吧,我明白了。你们这么快就增长到 10 亿美元的年化收入,但你们不可能保持这个速度。这不可能,对吧?这违反物理定律。你们做的是企业市场,这很好,但

Adoption is going to take so much longer. I mean, look at how long it took with cloud and how many people are still on-prem.

采用速度会慢得多。看看云计算花了多长时间,还有多少公司仍然在用本地部署。

The business continued to prove out the thesis that the return to frontier intelligence is really high that we are really focused on.

但业务持续证明了我们的论点:前沿智能的回报确实很高,而我们真正专注于此。

What's really happened is model growth enabled by products and our go-to-market team and our distribution.

实际发生的是,模型的增长是由产品、市场团队和分发渠道共同推动的。

And then I think what they also saw was that this thesis of like, hey, it's really important to build this transformative technology but to do it in the right way and do it responsibly, that that had this really interesting interlink with our business that most people didn't really understand or really believe.

我认为他们还看到了一点,就是「以正确和负责任的方式构建这项变革性技术非常重要」这个理念,与我们的业务有着非常有趣的关联,而大多数人并不真正理解或相信这一点。

Which was that we invest in research not just in model development but also in AI safety research, right? Like we've pioneered interpretability, which is, think of it as like an MRI for the model to see inside the neural network how it works.

我们不仅投资模型开发研究,还投资 AI 安全研究。我们在可解释性方面是先驱,可以把它想象成模型的核磁共振,能看到神经网络内部的运作方式。

We pioneered alignment science, which is

我们在对齐科学方面也是先驱。

You know, you want the model to do what you tell it to do, and how often does it do that and how often does it stray from that. And those things are important for our mission and that's why we did them, but they had these downstream effects where it turns out if you can look inside the model, you're better at building them.

你希望模型按照你的指令行事,它多久能做到这一点,多久会偏离。这些对我们的使命很重要,这也是我们做这些的原因,但它们产生了下游效应——如果你能看到模型内部,你就能更好地构建它们。

And then the last linkage, if you're selling to enterprises, like we now sell to nine of the Fortune 10, all of those enterprises are entrusting us with, you know, customer information, with their data.

最后一个关联是,如果你要向企业销售,比如我们现在服务财富 10 强中的 9 家,所有这些企业都把客户信息和数据托付给我们。

They're interacting with their employees, sometimes even interacting with their customers as well. That is, those are the most sensitive workloads.

我们与他们的员工互动,有时甚至与他们的客户互动。这些都是最敏感的工作负载。

The more and more of these businesses are running on cloud and our cloud platform. When you have this investment that we've made and will continue to make in safety, interpretability, alignment, like that actually inures to the benefit of the enterprise.

越来越多的业务运行在云和我们的云平台上。当你在安全、可解释性、对齐方面进行了我们已经做的和将继续做的投资,这实际上对企业有利。

Customers as well, because all of our customers, if they're going to entrust us with all that access and all that data and the ability to work in the most sensitive workflows within their company, they want a company that they can trust.

对客户也是如此,因为我们所有的客户,如果要把所有访问权限、所有数据以及在公司最敏感工作流程中工作的能力托付给我们,他们需要一家可以信任的公司。

And that's not why we invested in it, but it did have this kind of downstream effect that we've really seen prove out again and again to be a company that is both at the frontier, but one that is investing in safety and that you can trust.

这不是我们投资的初衷,但它确实产生了一种下游效应,我们一次又一次地看到这一点得到验证——成为一家既处于前沿、又投资于安全、值得信赖的公司。

We've raised, you know, $75 billion since I joined the company.

自从我加入公司以来,我们已经筹集了750亿美元。

We have another $50 billion that'll come in in the future from the Amazon and Google deals that we inked last month.

未来还会有500亿美元陆续到账,这些资金来自我们上个月与Amazon和Google签订的协议。

And so that's a tremendous amount of capital, but it's a capital-intensive business and we need this capital to support that growth.

这确实是一笔巨额资本,但我们这个行业本身就是资本密集型的,需要这些资金来支撑业务增长。

But you know, it all goes to the fact that, you know, the business is running very efficiently and so the reason we raise...

但实际上,这一切都说明我们的业务运营效率很高,所以我们筹集资金的原因是……

This capital is more because of that cone of uncertainty than it is to fund, you know, actual losses in the business today.

筹集这些资本更多是为了应对未来的不确定性,而不是为了弥补当前业务的实际亏损。

How is your own perception of this 10x perspective for a 10x growth of the business? Like the first time that happened, did you personally believe that it was possible? Did that seem absurd? And has it, like now that it's becoming consistent and so maybe it's becoming more commonplace to you or something, but what was your own view staring at this cone about the odds of hitting, you know, a 10x type of growth so many years in a row?

你自己如何看待这种业务增长10倍的预期?第一次遇到这种情况时,你个人相信这是可能的吗?这听起来会不会很荒谬?现在这种情况变得越来越常见,可能对你来说也变得更加习以为常了,但你自己面对这个不确定性区间时,对于连续多年实现10倍增长的可能性是怎么看的?

Well, when I joined the business it had about 250 million of run rate revenue, and the plan was to get to a billion. And I said, great, in what year? And that was like linear thinking, right? And you know, consistently, Dario has been a much better predictor of the revenue than I have. I think we're going to close the gap over time as we get better.

我刚加入公司时,年化收入大约是2.5亿美元,当时的计划是达到10亿。我就问,那是哪一年?这就是典型的线性思维,对吧?而事实证明,Dario对收入的预测一直比我准确得多。我想随着时间推移,我们会在预测方面做得更好,逐渐缩小这个差距。

At forecasting and understanding the business. But yeah, definitely the first time I saw it, you have all these arguments about the laws of physics and law of large numbers and this can't, you know, where is the revenue coming from and how can it be added this quickly and how can customers move this quickly and is this even possible in enterprise and all of those things start to get broken down over time as you see how the business works internally and you see how the adoption curves and the exponentials that are happening. Again, we have the exponential that's happening on revenue, but underlying that are these many other exponentials that support that.

在预测和理解业务方面会越来越好。但确实,第一次看到这种增长时,你会有各种质疑——物理定律、大数定律、这不可能实现、收入从哪里来、怎么可能增长这么快、客户怎么可能转换这么快、在企业市场这真的可能吗——所有这些疑问随着时间推移都会被逐一打破,当你看到业务内部的运作方式,看到采用曲线和正在发生的指数级增长时。我们的收入呈指数级增长,但支撑这一增长的是许多其他的指数级因素。

You start to see and believe in that more. Now, that doesn't mean we're not disciplined and thoughtful about the forecast and how we think about the range of scenarios. But it does mean that my thinking has at least shifted a lot more from linear and incremental towards, you know,

你会开始看到并相信这一点。当然,这并不意味着我们在预测和思考各种可能情景时不够严谨和周密。但这确实意味着我的思维方式已经从线性和渐进式转变了很多,转向了……

Leaning into this exponential and really, you know, believing in its potential and how this is just different than how other businesses have evolved.

拥抱这种指数级增长,真正相信它的潜力,认识到这与其他业务的发展方式完全不同。

As you've talked to investors at every stage, I'm sure every stage, every round that you've raised, there's something that is like the most common or hardest thing to explain to investors or that they're struggling the most to understand and get their heads around. What is that today?

在你与投资者交流的每个阶段——我相信每一轮融资都是如此——总会有一些最常见或最难向投资者解释的问题,或者他们最难理解和把握的东西。现在是什么?

I think it is this paradigm of how compute is used. Thinking of it as, you know, not just something that is like a variable cost over some time period, but really this resource that's so fully utilized, right? We run workloads on one day in the morning on a chip for inference and in the afternoon or evening we use it for model development. That is something that's, you know, that paradigm does not exist in a company like a software company or a factory, right? If you, you can't...

我认为是算力使用的这种范式。不能把它仅仅看作某个时间段内的可变成本,而应该看作一种被充分利用的资源,对吧?我们在同一天早上用一块芯片做推理,下午或晚上就用它来做模型开发。这种范式在软件公司或工厂里是不存在的,对吧?你不能……

Repurpose if you have a bunch of people doing R&D and that's your R&D expense, they can't go and become cogs, right? And vice versa in most traditional companies.

重新调配用途——如果你有一群人在做研发,那是你的研发支出,他们不能转而变成生产线上的工人,对吧?在大多数传统公司里反过来也不行。

Here, you really have that fungibility that's possible, and I think that's where the return on compute is so important. And I think people are beginning to understand that, but there's still a tendency towards treating it like, you know, oh I have to separate these two costs, when in actuality, you know, they're very self-reinforcing and that flexibility is actually what helps to drive revenue short-term and long-term.

但在我们这里,这种可替换性是真实存在的,我认为这就是算力回报率如此重要的原因。我觉得人们开始理解这一点了,但仍然倾向于把它当作两种独立的成本来对待,而实际上,它们是相互促进的,这种灵活性正是推动短期和长期收入增长的关键。

If I was to force you out of your role and into an investor seat at a great big investing firm and then I said your job is to go grill these companies and invest in the best ones, like what questions would you be asking of the labs or companies that are building models to really get at the heart of, you know, the points of uncertainty, of skepticism, of like things that might not—

如果我强行把你从现在的职位上拉下来,让你坐到一家大型投资公司的投资人席位上,然后说你的工作是去拷问这些公司,投资最好的那些,你会向那些构建模型的实验室或公司提出什么问题,来真正触及问题的核心——那些不确定性、质疑点,以及那些可能不会……

Make these the best businesses of all time. I'm curious maybe from that angle how you would approach it.

让这些公司成为有史以来最好的企业的因素。我很好奇你会从这个角度如何看待。

So I would say a couple of things. First, what is the ROI on compute kind of all up? How are you utilizing it? And what return are you seeing today? And how is that coming over time? Right?

我会问几个问题。首先,算力的整体投资回报率是多少?你们如何利用算力?今天看到了什么样的回报?随着时间推移回报如何变化?

These are the massive kind of unprecedented investments that companies like us are making. What is the return that you're generating on that and when does it come and what is the shape of it?

像我们这样的公司正在进行前所未有的大规模投资。你们从中获得了什么回报,何时获得回报,回报曲线是什么样的?

So I think that's one. A second one is, how do your customers see ROI in what you do? Are people just using this for testing? Are they actually deploying this at meaningful scale?

这是第一个问题。第二个是,你们的客户如何看待你们产品的投资回报率?人们只是在测试,还是真的在大规模部署?

You know, I could say for our business, like we're seeing that in spades. Our net dollar retention rate is over 500% on an annualized basis. And so, you know, with nine out of the Fortune 10...

对于我们的业务,我可以说我们在这方面表现突出。我们的净美元留存率年化超过500%。而且财富10强中有9家都是我们的客户……

These are real, real customers making significant buying decisions.

这些都是真实的客户,在做重大的采购决策。

No pilots anymore.

不再是试点项目了。

Exactly. Like on the way here I was in an Uber and I signed two double-digit million-dollar commits during the car ride, which was like 20 minutes.

没错。来这里的路上我坐在Uber里,20分钟的车程里我就签了两个千万美元级别的承诺订单。

So from that perspective, we're seeing it and we're now being judged by some of the biggest companies in the world, the most sophisticated buyers and startups that also have choice in the market and they're choosing us.

从这个角度来看,我们确实看到了成效,现在世界上一些最大的公司、最成熟的买家以及市场上有选择权的初创公司都在评判我们,而他们选择了我们。

But I think one question I get a lot, or I would ask from the investor seat, the skeptical investor seat, is like how are your customers getting return from this?

但我认为我经常被问到的一个问题,或者说如果我坐在持怀疑态度的投资人位置上会问的问题是:你们的客户如何从中获得回报?

Maybe a third one is, you know, how do you think about compute in the future and like where does it come from?

可能第三个问题是,你怎么看待未来的算力,以及这些算力从哪里来?

Because obviously some of the places that we buy compute from, you know, they sell the compute to others, they...

因为很明显,我们购买算力的一些供应商,他们也会把算力卖给其他人,他们自己也会...

might use the compute internally like what is the balance of that over time again for us like one reason why we have multiple different

内部使用这些算力,那么随着时间推移这种平衡会是怎样的?对我们来说,我们与多家不同供应商合作的一个原因就是

And so your philosophy there is just like be involved with great players and have flexibility?

所以你们的理念就是与优秀的合作伙伴保持合作,同时保持灵活性?

That's right, that's right. There's this crazy stat about AI just the generic concept being less popular than like Congress amongst like the general populace and it's kind of funny when you first hear, but when you really think about it, you're like, "This is kind of scary. Like, we need to solve this problem."

没错,就是这样。有个很疯狂的统计数据,AI 这个概念在普通大众中的受欢迎程度甚至不如国会,第一次听到时觉得挺好笑的,但仔细想想会觉得「这其实挺可怕的,我们需要解决这个问题」。

It doesn't seem like the general world that isn't in technology, doesn't live in the Bay Area or New York, does not yet feel or understand why this is good for them just as measured by their opinion of it. What do you think we need to do as an industry about that problem?

看起来不在科技行业、不住在湾区或纽约的普通人,还没有感受到或理解 AI 对他们有什么好处,至少从他们对 AI 的看法来衡量是这样。你认为作为一个行业,我们需要为这个问题做些什么?

Look, I think that if we think about the transformation that's happening, there's

我觉得,如果我们思考正在发生的这场变革,

There have been other transformative waves, right, before all the way back to the industrial revolution, the internet, cloud, etc.

之前也有过其他变革性的浪潮,对吧,一直追溯到工业革命、互联网、云计算等等。

I think what's one of the things that's different about AI is it's all happening so quickly.

我认为 AI 的不同之处之一在于,一切发生得太快了。

You can have, you know, years or decades of progress that are being compressed into months.

原本需要数年甚至数十年的进步,现在被压缩到了几个月内。

And going back to, you know, humans thinking in terms of exponentials versus linear, that can be jarring, I think.

回到之前说的,人类习惯线性思维而不是指数思维,这种变化速度可能会让人感到不适应。

We are very optimistic generally about the potential for this technology.

总体来说,我们对这项技术的潜力非常乐观。

I think that we as an industry can continue to do a better job of articulating, you know, Dario wrote this essay, Machines of Loving Grace.

我认为作为一个行业,我们可以继续更好地阐述这一点,Dario 写过一篇文章叫《Machines of Loving Grace》(充满爱的机器)。

It's all about the potential for this technology to transform the way that we live.

文章讲的就是这项技术如何有潜力改变我们的生活方式。

Whether that be in drug development and curing diseases that are more mainstream, but also rare diseases.

无论是在药物研发和治疗常见疾病方面,还是罕见病方面。

Number two, in healthcare and how healthcare is delivered to raise our

第二,在医疗保健以及医疗服务的提供方式上,提高我们的

Standard of living, you know, in the developing world and in places where resources are not as plentiful.

生活水平,在发展中国家以及资源不那么充足的地方。

I think that all of those things are part of the promise and the potential of AI, and so we could probably do a better job of painting that picture and we want to show more tangible results for that over time.

我认为所有这些都是 AI 的承诺和潜力的一部分,所以我们可能可以在描绘这幅图景方面做得更好,我们也希望随着时间推移展示出更多切实的成果。

I think that is coming and that's one of the things I'm most optimistic about.

我认为这些成果正在到来,这也是我最乐观的事情之一。

I think on the other side though, we do, and this is again cultural to us, like we do want to articulate the risks.

但另一方面,我认为我们确实需要,这也是我们文化的一部分,我们需要阐明风险。

Like I don't think we should just tell everyone everything's going to be great, because, you know, there are likely to be bumps on the road.

我不认为我们应该只告诉大家一切都会很好,因为在这条路上很可能会遇到一些坎坷。

And so I think people generally gravitate towards more honest and balanced assessments, right?

所以我认为人们通常更倾向于更诚实、更平衡的评估,对吧?

If I feel like somebody's just telling me all the good news and none of the bad news, then I'm like, okay, do I really trust this perspective?

如果我觉得有人只告诉我好消息而不告诉我坏消息,那我就会想,好吧,我真的能相信这个观点吗?

I think that's where there's a need for balance and to

我认为这就是需要平衡的地方,需要

Say like, look, these are some of the things that happen when change is compressed over a short amount of time.

说,看,当变化在短时间内被压缩时,就会发生这些事情。

How do we work across, you know, commercial and government to actually come up with some of the solutions to that?

我们如何跨越商业和政府部门,真正找到一些解决方案呢?

So I think it's about a clear articulation of the opportunities.

我认为关键在于清晰地阐明机遇所在。

It's about really thinking about what those solutions may be.

还要真正思考这些解决方案可能是什么样的。

And that's not any one company that can, you know, come up with it.

这不是任何一家公司能够独自想出来的。

We don't have the blueprint that's going to solve everything, but to at least have that dialogue about some of the risks and downsides and what we can do to address it.

我们没有能解决一切问题的蓝图,但至少要就一些风险和负面影响展开对话,讨论我们能做些什么来应对。

And then I think it's being transparent about that, about both of those things when we talk about it.

然后我认为在谈论这些问题时,要对这两方面都保持透明。

And so over the long term, the opportunity is going to be significantly higher and greater than, you know, some of the risks and the downsides that will happen.

从长远来看,机遇将远远大于那些会出现的风险和负面影响。

But that doesn't mean it's going to be perfectly smooth on the curve.

但这并不意味着发展曲线会完全平滑。

The release of Mythos was such an

Mythos 的发布是一个非常

Interesting moment. It was the first time many people, friends of mine that are careful watchers of this stuff said something like, "I'm like, this one kind of makes me scared."

有意思的时刻。这是第一次,我的很多密切关注这个领域的朋友说:「这个模型让我有点害怕。」

So it relates back to the safety question.

所以这又回到了安全问题。

It's also the first example of you coming out and saying, "We want to make sure this isn't used for bad, and it's maybe the first one that we are worried could be used for bad."

这也是你们第一次站出来说「我们要确保它不被用于坏事」,可能也是第一个你们担心可能被用于坏事的模型。

I'm curious what that discussion was like internally before the world heard about it, the decision-making process around it.

我很好奇在向外界公布之前,内部的讨论是什么样的,围绕它的决策过程是怎样的。

And just using that as an example to talk about the things that do scare you as we continue to advance and the scaling laws continue to hold.

就以此为例,谈谈随着我们不断进步、scaling laws 持续有效,有哪些事情确实让你们感到担忧。

Yeah. I think one of the things about Mythos is that people maybe misconstrue it as just a cyber model.

是的。我觉得关于 Mythos,人们可能误解了,以为它只是一个网络安全模型。

It is an incredibly capable model across many different dimensions.

它其实是一个在很多不同维度上都极其强大的模型。

What we found was that cyber in particular was a place where it spiked

我们发现网络安全是它能力特别突出的一个领域。

And so this was the first model that we decided to release in a different way, and the way in which we did that again is consistent with our mission, our principles, like we wanted to do it in that way.

所以这是我们第一个决定以不同方式发布的模型,而我们采用的方式再次符合我们的使命和原则,我们想要这样做。

And so we have this phased approach to it because we think that when a model is this capable, and again cyber is the thing that people focused on, but you know there are other things as well.

所以我们采取了分阶段的方法,因为我们认为当一个模型如此强大时——网络安全是人们关注的焦点,但其实还有其他方面。

We think again it can be used in a positive way, right, to patch code bases.

我们认为它可以被用于积极的方面,比如修补代码库。

You've seen these examples where, um, you know, we had an open source code base that, you know, a prior model found 22 security vulnerabilities in and Mythos then found 250.

你看到过这些例子,我们有一个开源代码库,之前的模型在其中发现了 22 个安全漏洞,而 Mythos 发现了 250 个。

Uh, so that is kind of scary, right, but that informed the way in which we released it.

这确实有点吓人,对吧,但这也影响了我们发布它的方式。

So we didn't say we're never going to release it. We said let's do it in a phased way. Let's do it to, you know, a group that will expand over time where we can, you know, focus on this one cyber capability and how it can actually be used.

我们没有说永远不发布它。我们说的是分阶段发布,先向一个会随时间扩大的群体发布,这样我们可以专注于这一项网络安全能力以及它实际上可以如何被使用。

Used positively, you know, in a defensive way as opposed to in an offensive way, and we think that's a template that could be used for the future. But because of this one particular area, we wanted to be cognizant of that in how we released it.

用于积极的、防御性的方式,而不是攻击性的方式,我们认为这可以作为未来的一个模板。但正是因为这一个特定领域,我们在发布时想要对此保持警觉。

You're so big now that you run into everything and everyone.

你们现在规模太大了,会遇到所有的事情和所有的人。

And one example of this is the government just a couple days ago said maybe there'd be this new system where you have to sort of pre-approve the release of a new model with the government before it was released to the public.

举个例子,就在几天前,政府表示可能会推出一个新系统,要求在向公众发布新模型之前,必须先获得政府的预先批准。

Obviously, you had the crazy experience with the Department of Defense, which I'm really curious what that was like as you went through it.

显然,你在国防部那边有过一段很疯狂的经历,我真的很好奇你当时是怎么度过的。

Like now everyone cares about this company and this technology and the couple other companies that are building it.

现在每个人都在关注这家公司、这项技术,以及其他几家正在开发它的公司。

How do you navigate that stuff? And some of it's just, I guess, beyond your control, but I'm sure you're trying to

你是如何应对这些事情的?我想有些事情可能超出了你的控制范围,但我相信你在尽力

Work with people as best you can. Maybe talk about those two examples of like the government now as a very relevant partner, player, you know, overseer, etc.

尽可能地与人们合作。也许可以谈谈这两个例子,比如政府现在作为一个非常相关的合作伙伴、参与者、监管者等等。

Yeah. So, I think first and foremost, like we prioritize having a strong relationship on this because we do think that, you know, regulation has a role to play in how these models are developed over time.

好的。首先,我认为我们优先考虑在这方面建立牢固的关系,因为我们确实认为监管在这些模型随时间发展的过程中应该发挥作用。

We are very like America first in our approach. We want the technology to support the US as well as, you know, democratic countries around the world.

我们的做法非常「美国优先」。我们希望这项技术能够支持美国以及世界各地的民主国家。

And that's one of the reasons why we've been working closely with the administration on something like Stargate.

这也是我们一直与政府密切合作推进 Stargate 这类项目的原因之一。

I do think that there's a balance, right? You want to be able to have innovation happen really quickly and have that not be slowed down, but you also want to have this kind of responsibility framework for how these things are deployed because we've long said

我确实认为需要平衡,对吧?你希望创新能够快速发生而不被拖慢,但同时你也希望有一个责任框架来规范这些技术的部署,因为我们一直在说

That, you know, this technology has implications and we should have an honest conversation about them, and that includes with the government. And so I think that's, you know, I think the Nitros process is a good example of that.

这项技术是有影响的,我们应该坦诚地讨论这些影响,包括与政府讨论。所以我认为 Nitros 流程就是一个很好的例子。

Can you teach us a bit more about the culture, like how you would describe the cultural tenets to your parents or something like this?

你能多讲讲公司文化吗,比如你会如何向你父母描述这些文化准则?

And what feels like it really drives most of the culture? I'm especially curious about the writing. You hear often that, you know, Dario publishes these long essays every so often externally. My understanding is he does that way more frequently and there's a lot of writing culture internally. I'm just trying to get a sense of what the culture is like to be in and what makes it the most distinctive from other companies, maybe that you've worked at, or from other companies that are trying to do the same thing. And what's your sense of

什么真正驱动了大部分文化?我特别好奇写作文化这一块。经常听说 Dario 会不时对外发布长篇文章。据我了解,他在内部发布的频率要高得多,而且内部有很浓厚的写作文化。我只是想了解一下身处其中是什么感觉,以及与你工作过的其他公司,或者与其他试图做同样事情的公司相比,最显著的区别是什么。你对

Of the differences and the distinctiveness, the culture is a real unique aspect of Anthropic and it is something that, you know, we do talk about externally, but it's different when you're kind of in there living it, and maybe I can tell you a little bit about some of my observations.

这些差异和独特性有什么看法?文化确实是 Anthropic 非常独特的一面,我们确实会对外谈论它,但当你真正身处其中时感受是不一样的,也许我可以跟你分享一些我的观察。

First of all, we have seven co-founders, right? That shouldn't work on paper, but it really does in practice.

首先,我们有七位联合创始人,对吧?这在纸面上看起来不应该行得通,但在实践中确实很有效。

And I think they've really set the example for the culture and the things that really matter to the company.

我认为他们真正为公司的文化和真正重要的事情树立了榜样。

We do a culture interview and it's not some pro forma, you know, thing we do just to kind of check a box.

我们会进行文化面试,这不是那种走形式、只是为了打勾的事情。

It is a real part of the evaluation process.

它是评估流程中真正重要的一部分。

So somebody could be flying colors on everything else and really, really the smartest person you've met in this role, we won't hire them if they don't pass the culture bar.

所以即使某人在其他所有方面都表现优异,真的是你在这个岗位上见过最聪明的人,如果他们没有通过文化关,我们也不会录用。

And the way I would describe it, um, I like that frame. How would you describe it to your parents is

我会这样描述,嗯,我喜欢这个框架——你会如何向你父母描述它

It's one incredibly collaborative, and this means that we don't really tolerate the fiefdoms or the sharp elbows or the like "I need to take credit for this."

首先是极度协作,这意味着我们真的不容忍各自为政、争抢地盘或者「我需要为此邀功」这种行为。

It's incredibly humble. It's like, you know, our competitors are incredibly capable and success is far from guaranteed.

极度谦逊。就像,你知道,我们的竞争对手非常有能力,成功远非板上钉钉。

And I think that's really part of how the company operates. If we reach a milestone and something good happens, there's not confetti on the floor, it's like "what's next?"

我认为这真的是公司运作方式的一部分。如果我们达成了一个里程碑,发生了好事,地上不会撒彩纸,而是「接下来做什么?」

And I think it's just that focus on the mission and the alignment that kind of is imbued throughout the culture of the company.

我认为正是这种对使命的专注和贯穿整个公司文化的一致性。

The other thing I would say is, you know, there's rigorous debate, right? There's an intellectual openness and intellectual honesty that happens where people question things.

另一点我想说的是,你知道,有严谨的辩论,对吧?有一种智识上的开放和诚实,人们会质疑事情。

People will really express a point of view, but then there's dialogue around it that's productive, and then we'll decide on a path forward.

人们会真正表达自己的观点,然后围绕它进行富有成效的对话,最后我们会决定前进的路径。

After that happens, there's real alignment. So in something like compute allocation we were talking about before, people might have different perspectives on how to allocate that compute, but they will engage in a thoughtful discussion about where the returns are the highest or the best. And when they do that, you know, and we come to a decision, then there's alignment on it. There's not second-guessing, there's not this kind of politics or fiefdom. The other piece of it is it's remarkably transparent, the culture, right? So

这之后,就会形成真正的共识。比如我们之前提到的算力分配问题,大家可能对如何分配算力有不同看法,但他们会进行深思熟虑的讨论,探讨哪里能获得最高或最好的回报。当他们这样做并最终做出决定时,就会达成共识。不会有事后质疑,也不会有那种政治斗争或山头主义。另一个方面是,这种文化非常透明。

Dario gets up in front of the company every two weeks, usually writes a short document, and he talks about, you know, usually three or four topics, and then takes open questions from the company. And these are not softballs, they're not like planted questions, they're just real questions that are on people's minds, and he answers them the best that he can. And it's not a decision-making forum, but it is a way for the company to get a window into how

Dario 每两周会在全公司面前做分享,通常会写一份简短的文档,讲三四个话题,然后接受公司的开放提问。这些问题不是那种软性问题,也不是事先安排好的,而是员工真正关心的问题,他会尽力回答。这不是一个决策论坛,但它为公司提供了一个窗口,让大家了解

Leadership is thinking how he's thinking, and there's debate and dialogue in that, and I think that is something that people really value. Like, it is a transparent culture.

领导层在思考什么、他是如何思考的,其中也会有辩论和对话,我认为这是大家真正看重的。这确实是一种透明的文化。

It is one where, you know, all seven of the co-founders are still at the company.

在这种文化下,七位联合创始人全部还在公司。

The vast majority of the first, you know, 20 to 30 employees are still at the company.

最早的二三十名员工中,绝大多数也还在公司。

And I think the culture underpins the reason why we've been able to attract and retain some of the best talent in the industry, right?

我认为这种文化是我们能够吸引和留住业内顶尖人才的根本原因。

Because we don't always pay people the most. We have, you know, very competitive compensation packages.

因为我们并不总是给员工最高的薪酬。我们的薪酬待遇很有竞争力。

But when you know Meta and others were out, you know, with these huge packages for some of the technical talent across the large language labs, I think we lost two people and other labs lost dozens.

但当 Meta 和其他公司为大语言模型实验室的技术人才开出巨额薪酬包时,我们只流失了两个人,而其他实验室流失了几十人。

What parts of the business and the culture—I mean, specifically for researchers—why do you think that stat is true?

业务和文化的哪些方面——我是说,特别是对研究人员来说——你认为为什么会有这样的数据?

I think it really is

我认为这真的是

Underpinned by the culture, and that's not just something we feel. It's like empirically when you talk to people, it's, you know, I want to have the most impact possible.

由文化支撑的,这不只是我们的感觉。从实际情况来看,当你和员工交流时,他们会说,我想产生尽可能大的影响。

I want to work in a place where, um, again, this idea of talent density mattering more than talent mass.

我想在一个地方工作,在那里,人才密度比人才总量更重要。

And I want to work in a place that is actually collaborative versus I have to like fight for this one thing and I feel like it wasn't discussed and debated in the right way, or there wasn't transparency around how a decision was made.

我想在一个真正协作的地方工作,而不是我必须为某件事争斗,感觉它没有以正确的方式被讨论和辩论,或者决策过程缺乏透明度。

I think that actually really matters because most of our team just wants to do really, really good work and they're attracted to the company for the mission.

我认为这真的很重要,因为我们团队的大多数人只是想做非常出色的工作,他们是被公司的使命吸引而来的。

The idea of having an impact on a company like ours that is trying to develop this transformative technology but to do it in a responsible way, I think that that really matters to the people, not just on the research team but across the company, and that we think is a real advantage for us and it's not

能够在像我们这样的公司产生影响——一家试图开发这种变革性技术但以负责任的方式进行的公司——我认为这对员工来说真的很重要,不仅是研究团队,整个公司都是如此,我们认为这是我们的真正优势,而且这不是

Something that we take lightly. We have this concept of a race to the top. We want, you know, we don't always have all the right answers. We don't always do everything perfectly, but we want others to look at some of the things we do and maybe emulate some pieces of that and actually have the technology be developed in a better way across the industry.

我们轻视的东西。我们有一个「向顶端竞赛」的概念。我们并不总是有所有正确答案,也不总是把每件事都做得完美,但我们希望其他人能看到我们做的一些事情,也许模仿其中的某些部分,让整个行业以更好的方式开发这项技术。

I think people are also really attracted to that as well. Again, not that we have all the answers, but that we can be a part of contributing and leading to how this can go well for humanity.

我认为人们也非常被这一点吸引。再次强调,不是说我们有所有答案,而是我们可以参与并引领如何让这项技术造福人类。

As you're having conversations with people internally, what does the frontier feel like to you? I don't just mean the model frontier. I mean like the next couple of rolls of the dice here in building AI in general.

在你与内部人员的对话中,前沿对你来说是什么感觉?我不只是指模型前沿,我是指在构建 AI 方面接下来的几次尝试。

Everyone is kind of wise to like these things are powerful. Everyone's using them. It's becoming, it's diffusing. People are becoming accepting of

每个人都意识到这些东西很强大。每个人都在使用它们。它正在扩散。人们正在接受

What feels to you like the frontier from the inside?

从内部来看,什么让你感觉像是前沿?

I think it's this idea, and again it's because we're focused on enterprise and because we're really trying to change the productivity of knowledge work that's done in the economy.

我认为是这个想法,这也是因为我们专注于企业,因为我们真的在试图改变经济中知识工作的生产力。

I think it is towards this vision or this goal of like a virtual collaborator.

我认为是朝着虚拟协作者这个愿景或目标前进。

And so think of this as something that has context within your organization that can use all of the tools that are specific to you, whether they be homegrown tools or tools that you purchase, that has memory and the ability to effectively learn from mistakes you've made, but also mistakes that it's made over time.

可以把它想象成这样的东西:它了解你组织内的上下文,可以使用所有你特有的工具,无论是自研工具还是购买的工具,它有记忆力,能够有效地从你犯过的错误中学习,也能从它自己随时间犯的错误中学习。

And then the ability to work over a very long time horizon on not just a task but an actual idea.

然后它能够在很长的时间跨度内工作,不仅是完成一个任务,而是推进一个真正的想法。

And so what that means for us is the model capability has to continue to grow to support that.

所以对我们来说,这意味着模型能力必须持续增长来支持这一点。

And then the products we build on top of it can unlock this.

然后我们在此基础上构建的产品就能释放这种能力。

Virtual collaborator that we think can really accelerate knowledge work.

我们认为这种虚拟协作者能真正加速知识工作。

But you have to get it in the right form factor.

但你必须找到合适的产品形态。

Right. This is where intelligence is not just a single dimension. It's multiple things, but the virtual collaborator kind of combines many of those things, right?

对。智能不是单一维度的,而是多方面的,虚拟协作者就是把这些方面结合起来,对吧?

Which is something that's not just generically smart, but is smart for your use cases.

它不只是泛泛的聪明,而是针对你的具体使用场景的智能。

And I think again, what we're seeing in coding is something that we expect to see elsewhere.

我们在编程领域看到的情况,我认为在其他领域也会出现。

For us, Claude for code has led the way on that as well as much of the business that we have great customers in that are pushing the coding frontier as well.

对我们来说,Claude for code 在这方面引领了方向,我们也有很多优秀的客户在推动编程前沿的发展。

But then you also see something like Cowork come along and start to unlock that co-working faster than Claude for code was if you index them to the same point in time.

但你也看到像 Cowork 这样的产品出现,如果把它们放在同一时间点比较,Cowork 释放协作能力的速度比 Claude for code 还要快。

That's kind of remarkable because developers are really fast adopters of

这很了不起,因为开发者对这项技术的接受速度本来就很快。

this technology. But I think it's because the model capabilities and the products are pushing towards this notion of a virtual collaborator where even our product development today is not done by like one product manager with two engineers shipping something over 3 months.

但我认为这是因为模型能力和产品都在朝着虚拟协作者的方向发展,现在我们的产品开发已经不是一个产品经理带两个工程师花三个月交付一个东西了。

It's shipping daily and there's a fleet of agents that are working across the company on a specific task.

而是每天都在交付,有一整队 agents 在公司内协作完成特定任务。

Everyone kind of becomes a manager and I think the implications of that and the productivity gain that can come from that when it's in the right form factor is we're very very early in that but the potential for it is incredible, crazy to imagine.

每个人都变成了管理者,我觉得当产品形态合适时,这种模式的影响和能带来的生产力提升——我们现在还处于非常早期的阶段,但它的潜力是难以想象的,太疯狂了。

I'm curious how you've had to personally evolve to be able to stay doing this.

我很好奇你个人是如何进化来持续做这件事的。

Like you hear a lot about these stories about how the executives have to scale with the company or else they'll get new executives.

你经常听到这样的故事,说高管必须随着公司成长,否则就会被换掉。

You know the business that you were at prior to this was a great business but it was a tiny—Cedar was a

你之前做的业务是个很好的业务,但规模很小——Cedar 是个

Tiny, tiny fraction of the scale, so like everyone is in this new unprecedented thing.

非常非常小的规模,所以现在每个人都在经历这种前所未有的事情。

You talked about the example of like getting out of linear into, you know, into more exponential type thinking.

你提到了从线性思维转向指数型思维的例子。

That's one example of what I mean, but how have you managed it personally?

这就是我说的一个例子,但你个人是怎么应对的?

Like, what have you had to do? What's been the most painful?

你必须做什么?什么最痛苦?

Like, how do you manage your own ability to scale with this thing that's scaling faster than what we've seen before?

你如何管理自己的能力,让自己能跟上这个增长速度超过以往任何时候的东西?

Yeah, it's really hard, but I think the important thing is to think in first principles, right?

这确实很难,但我认为重要的是用第一性原理思考,对吧?

So this is like everyone has priors when they come to something new.

每个人面对新事物时都有自己的先验认知。

Thinking in first principles and having like intellectual openness.

用第一性原理思考,保持思想上的开放。

You know, I spent a lot of time with Tom Brown, our chief compute officer.

我花了很多时间和我们的首席计算官 Tom Brown 交流。

He was actually one of the first people to interview me at the company, and I remember we went on a walk like, it was a bit before I

他其实是公司最早面试我的人之一,我记得我们一起散步,那是在我

It started and we walked around the Mission in San Francisco for two and a half hours, and he started to tell me about his vision for the future of the company, and this is in 2024, early 2024.

那次见面开始后,我们在旧金山 Mission 区走了两个半小时,他开始跟我讲他对公司未来的愿景,那是在 2024 年,2024 年初的时候。

And I'll be honest, it sounded crazy.

说实话,听起来简直疯狂。

You walked me all the way home, and I remember I came in, I told my wife, I was like, this is going to be wild, like if even 10% of that is true, this is going to bend all paradigms of what not just things I've seen but what most people have seen.

你一路把我送到家,我记得我进门后跟我妻子说,这将会是疯狂的经历,如果他说的哪怕只有 10% 是真的,这都将打破所有范式——不仅是我见过的,而是大多数人见过的。

And it turns out that a lot of what Tom said during that walk has come to fruition, but I remember that as like an early formative thing, coming home and being like, holy shit, like this is going to be totally different and new, and you know, a really incredible experience but also a really challenging one.

结果证明 Tom 在那次散步中说的很多事情都实现了,但我记得那是一个早期的关键时刻,回到家时我想,天哪,这将完全不同,是全新的东西,会是一段非常不可思议的经历,但同时也会非常具有挑战性。

And that's what it's been.

事实也确实如此。

The other piece of this is just hiring great people. You know, I try to hire people and I tell people during the interview process, I'm like, I'm not...

另一个重要方面就是招聘优秀的人才。我在面试过程中会告诉候选人,我不是……

Really hiring you as like a direct report of mine. I'm hiring you as a partner and I want you to treat it as a partnership which means that there might be things that you and I disagree on.

真的把你当作我的直接下属来招聘。我是把你当作合作伙伴来招聘的,我希望你也把这当作一种合作关系,这意味着你我之间可能会有意见不一致的时候。

I want to hear that and I want to like whiteboard it. I want to understand like you know we've hired people from some of the best companies in the world. They come to this from a different perspective right. They might come to it from a hyperscaler or a large software company or from financial services. In another lifetime you know I worked at Blackstone in the private equity group. Like that training is really valuable and thinking about things at a granular level and not losing that. Like I'm not somebody who is comfortable at 50,000 feet. That's just like not me. But you can't be at 500 feet at everything in this business. There's too much surface area and so having people who can be partners in that is really really critical.

我想听到那些不同意见,我想在白板上讨论。我想理解——我们从世界上一些最好的公司招来了人才,他们带来不同的视角,对吧。他们可能来自超大规模云服务商、大型软件公司或金融服务行业。在另一段人生中,我曾在 Blackstone 的私募股权部门工作过。那种训练非常宝贵,能让你在细节层面思考问题而不失焦。我不是那种习惯在五万英尺高空俯瞰的人,那不是我的风格。但在这个业务中,你也不可能在所有事情上都保持在 500 英尺的高度。涉及面太广了,所以有能够成为合作伙伴的人真的非常关键。

I think the last piece is to think about you

我认为最后一点是要思考

Know how the business evolves over time and where there might be moments or analogues to things that have happened in the past.

业务如何随时间演变,以及在哪些时刻可能会出现与过去类似的情况。

I helped lead the financing that Airbnb did in the middle of the pandemic.

我曾帮助主导了 Airbnb 在疫情期间的融资。

Very different situation, right? The business lost 70% of its revenue in seven weeks.

那是完全不同的情况,对吧?公司在七周内失去了 70% 的收入。

I know Brian was just—did a show with you.

我知道 Brian 刚刚——在你的节目上做过访谈。

That was a harrowing time, but it was also a time kind of without precedent, right?

那是一段艰难的时期,但也是一段几乎没有先例可循的时期,对吧?

Where you had to think about things with a clear perspective when it was rapidly changing and there was not a good template or pattern to match.

在那种情况下,你必须在快速变化中保持清晰的视角来思考问题,而且没有好的模板或模式可以参照。

And then on a personal level, look, it is hard to balance everything—family and friends—and certainly this job takes a big bite out of all that.

然后在个人层面,说实话,平衡所有事情——家庭和朋友——是很难的,这份工作确实占据了很大一部分时间。

But what I do try to do maybe once a week is in a quiet moment just think, wow, this is really cool. It's an incredible opportunity to—

但我确实会尝试每周一次,在安静的时刻想一想,哇,这真的很酷。这是一个难得的机会——

Work, you know, with this group of people on this problem at this company at this moment in time. And I try to do that, you know, again, maybe it's in a car ride, maybe it's late at night or something like that.

能够和这群人一起,在这个时刻,在这家公司,解决这个问题。我会尝试这样做,可能是在开车的时候,可能是深夜或类似的时候。

Just having that recognition and that appreciation is really important.

保持这种认知和感激真的很重要。

What did Tom tell you on the walk that sounded most crazy?

Tom 在那次散步中告诉你的哪些事情听起来最疯狂?

I mean, we talked a lot about the scale of the compute infrastructure, what models could do in a short amount of time.

我们聊了很多关于计算基础设施的规模,以及模型在短时间内能做到什么。

I think, you know, he described a world that I would have said is kind of sci-fi. But a lot of what we're experiencing here and now have really roots in that conversation.

我觉得,他描述的世界我当时会认为是科幻小说。但我们现在正在经历的很多事情,其实都源于那次对话。

And so there's even more things that he talked about that are probably beyond where we are today.

所以他还谈到了更多可能超出我们今天所处阶段的事情。

But I think the commonality of it was that, you know, everything is going to happen much quicker than we think and that both the implications but also the

但我认为共同点在于,一切都会比我们想象的发生得更快,而且这带来的影响以及

Capabilities of that can change, and then he also had like a really incredible optimism about the future that I think, you know, we talk about internally kind of holding light and shade.

能力都会发生变化,然后他对未来也有一种非常不可思议的乐观态度,我觉得这就是我们内部所说的「同时持有光明与阴影」。

That's one of the things we say, and I think like I came from that conversation with just a bunch of questions but also just a sense of positivity about what could happen in the future.

这是我们常说的一点。我觉得那次对话结束后,我带着一堆问题离开,但同时也对未来可能发生的事情充满了积极的期待。

It seems like we spent most of our time talking about because it's been the reality that we exist at the high end of that cone.

看起来我们大部分时间都在讨论这个,因为现实情况是我们一直处于那个锥形区间的高端。

What can you imagine that would cause that to change to the low end of that cone? Like how, if we were to do like some sort of premortem on a year from now, we're like, "Wow, actually we didn't need nearly as much compute as we thought or something like that." What can you imagine that would shift us meaningfully in that cone?

你能想象什么情况会让我们转向锥形区间的低端吗?比如说,如果我们现在对一年后做个事前复盘,发现「哇,实际上我们需要的算力远没有想象中那么多」之类的。你能想到什么会让我们在这个区间内发生显著变化?

I think the first thing would be the diffusion rate within our customers, the use cases are playing

我认为第一个因素是我们客户内部的扩散速度,也就是实际应用场景的推进情况。

Catch up to the model capability, and I think, you know, look, these are—we are talking about humans in large organizations with a set of tools and practices and things that they've been doing for a really long time. Change is hard, right? And so to the extent that that diffusion, you know, hits a wall or slows down or something like that, that could affect the kind of rate of change in terms of revenue growth.

能否跟上模型能力的发展。你要知道,我们面对的是大型组织中的人,他们有一套已经使用了很长时间的工具、实践和做事方式。改变是很难的,对吧?所以如果这种扩散遇到瓶颈、放缓或者类似的情况,就可能影响营收增长的速度。

Certainly the scaling laws slowing down or not holding—we don't see that. We can't say that with 100% certainty. I think that would be silly. We certainly believe in the trajectory, but the model capabilities leveling off would be another thing.

当然还有scaling laws放缓或者失效的可能——我们目前没有看到这种迹象。我们不能百分之百确定,那样说太愚蠢了。我们确实相信这个发展轨迹,但如果模型能力增长趋于平缓,那也会是一个影响因素。

And then, you know, maybe third is just how we think about being at the frontier. You know, today we're at the frontier. I think we're defining the frontier of agentic AI. We need to stay there, right? And it's a competitive

第三点可能是我们如何看待保持在前沿的问题。现在我们处于前沿,我认为我们正在定义agentic AI的前沿。我们需要保持在那里,对吧?这是一个竞争激烈的市场。

Market, and we're going to continue to invest in the technology and the compute and the go-to-market that's required to be there, but that's not guaranteed either.

我们会继续在技术、算力和市场拓展方面投入必要的资源来保持领先,但这也不是板上钉钉的事。

What are you most excited about? Like, you get to—you have a privileged seat. You sort of get to literally see the future because it's happening inside the business before those outside the business see it.

你最兴奋的是什么?你的位置很特殊,可以说你能真正看到未来,因为这些变化正在公司内部发生,而外界还看不到。

With that perspective and in that seat, what are you most excited about in the future?

站在这个角度,处在这个位置上,你对未来最期待的是什么?

I really think that the biotechnology and healthcare outcomes that can come from this technology are the things that I'm most optimistic about.

我真的认为,这项技术能在生物技术和医疗健康成果方面带来的改变,是我最乐观看待的事情。

We may live in a world where you're diagnosed with a disease that is not curable, but in your lifetime that cure can be found much more rapidly and you actually might not die of that disease.

我们可能会生活在这样一个世界:你被诊断出患有某种无法治愈的疾病,但在你的有生之年,治愈方法可以被更快地找到,你实际上可能不会死于那种疾病。

And I think of this as like, you know, a lot of what we're doing today is helping to speed up the drug development process, right? A lot of the paperwork and—

我觉得这就像,我们现在做的很多工作是帮助加速药物开发过程,对吧?大量的文书工作和——

Clinical studies, reports, and things like that that are needed to be done—AI and our solutions in particular are helping to rapidly accelerate that.

临床研究、报告之类需要完成的工作——AI,特别是我们的解决方案,正在帮助快速推进这些工作。

I'm really most optimistic and excited about when it goes further back into drug development and drug discovery, because, you know, our humans are incredibly capable at research, but if you think about these molecules and proteins, like, they're so complex and such small changes have such big implications for the outcomes—like, AI is perfect for that.

我真正最乐观和兴奋的是当它进一步深入到药物开发和药物发现阶段,因为人类在研究方面的能力确实很强,但如果你想想那些分子和蛋白质,它们是如此复杂,微小的变化就会对结果产生巨大影响——AI在这方面简直完美。

If you think about what can happen when the lab's throughput goes up 10x or 100x and we can run that many more experiments, probably get better results faster, and that can be something that helps, you know, people around the world, right? And it doesn't have to be limited to a small set of diseases or disorders—it can really go much further down the chain. And so I think that has the potential to, you know, greatly alter the...

想想当实验室的产出提高10倍或100倍时会发生什么,我们可以进行更多实验,可能更快得到更好的结果,这能帮助到全世界的人,对吧?而且不必局限于少数几种疾病或病症——它可以延伸到更广泛的领域。所以我认为这有潜力极大地改变……

Way that we live and the way that we interact, and that's really exciting to me.

我们的生活方式和互动方式,这真的让我很兴奋。

I sure hope you're right. It sure seems like we're on that trajectory and it's quite a future to imagine.

我真心希望你是对的。看起来我们确实在朝那个方向发展,想象一下那样的未来真是令人激动。

This is so much fun. I feel like we covered so many interesting aspects of the business that, you know, you don't—I don't think you've done this before. So, you know, don't get this amazing perspective.

这次聊天太有意思了。我觉得我们涵盖了业务的很多有趣方面,你之前应该没做过这种访谈。所以能得到这种精彩的视角真是难得。

When I do these, I ask the same traditional closing question. What is the kindest thing that anyone's ever done for you?

每次做这种访谈,我都会问同一个传统的结束问题。别人为你做过的最善良的事情是什么?

I have a brother who's five and a half years older than me, and we lived in California when he went to college.

我有个比我大五岁半的哥哥,他上大学的时候我们住在加州。

He got into everywhere he applied to, and he was going to go to medical school after that.

他申请的每所学校都录取了他,而且他之后打算去读医学院。

And so, I didn't know any of this at the time.

当时我完全不知道这些事。

So he ended up going to college in-state and he did exceptionally well.

所以他最后去了州内的大学读书,而且表现非常出色。

It's kind of years later that I kind of had to pull this out of him.

多年以后我才从他那里把这件事问出来。

But, you know, in deciding where to go to college, you know, we were solidly middle class as a family. And, you know, this was like 25-30 years ago.

但你知道,在决定去哪所大学读书时,我们家当时是标准的中产阶级。而且这是大约25到30年前的事了。

You know, the financial aid packages weren't, you know, as robust as they are today. And a big factor in his decision, I found out, you know, many, many, many years later, was, you know, wanting to give me the opportunity to go wherever I wanted.

那时候的助学金项目远没有现在这么完善。我是很多很多年后才知道,他当时做决定的一个重要考虑因素,是想让我将来能有机会去任何我想去的学校。

Even though, you know, that was 6 years out and who knows how it would turn out. I didn't know that.

虽然那是6年以后的事,谁也不知道会怎么样。但我当时完全不知道这些。

And it was something that, you know, 12-year-old me or 13-year-old me would have never really understood. But now, you know, many years later, I think that's something that was incredibly kind and is something that, you know, I still kind of hold with me today.

这种事情,12岁或13岁的我是永远不会真正理解的。但现在多年过去了,我觉得这是一件非常善良的事,也是我至今仍然铭记在心的。

Wow, I've done this like 600 times or something. I've never heard an answer like that type. That's awesome and amazing.

哇,我做这个节目大概做了600次了吧。从来没听过这样的回答。太棒了,太感人了。

Christian, thanks so much for doing this.

Christian,非常感谢你来做这期节目。

With me. Yeah, thanks for having me, Patrick. Really enjoyed it.

是的,谢谢你邀请我,Patrick。我很享受这次对话。

You know how small advantages compound over time? That's true in investing and just as true in how you run your company.

你知道微小的优势是如何随时间累积的吗?这在投资中是真理,在你经营公司的方式上同样如此。

Your spending system is your capital allocation strategy. Ramp makes it smarter by default. Better data, better decisions, better economics over time. See how at ramp.com/invest.

你的支出系统就是你的资本配置策略。Ramp让它默认就更智能——更好的数据、更好的决策、随时间推移更好的经济效益。访问ramp.com/invest了解详情。

As your business grows, Vanta scales with you, automating compliance and giving you a single source of truth for security and risk. Learn more at vanta.com/invest.

随着你的业务增长,Vanta与你一起扩展,自动化合规流程,为你提供安全和风险管理的单一可信来源。访问vanta.com/invest了解更多。

Every investment firm is unique and generic AI doesn't understand your process. Rogo does. It's an AI platform built specifically for Wall Street, connected to your data, understanding your process, and producing real outputs. Check them out at rogo.ai/invest.

每家投资公司都是独特的,通用AI并不理解你的流程。但Rogo理解。它是专为华尔街打造的AI平台,连接你的数据,理解你的流程,并产出真实的成果。访问rogo.ai/invest查看。

The best AI and software companies from OpenAI to Cursor to Perplexity use Work

从OpenAI到Cursor再到Perplexity,最优秀的AI和软件公司都在使用Work

OS to become enterprise ready overnight, not in months. Visit works.com to skip the unglamorous infrastructure work and focus on your product.

OS在一夜之间做好企业级准备,而不是花费数月时间。访问works.com跳过那些不起眼的基础设施工作,专注于你的产品。

Ridgeline is redefining asset management technology as a true partner, not just a software vendor.

Ridgeline正在重新定义资产管理技术,作为真正的合作伙伴,而不仅仅是软件供应商。

They've helped firms 5x and scale, enabling faster growth, smarter operations, and a competitive edge.

他们帮助公司实现5倍增长和规模化,实现更快的增长、更智能的运营和竞争优势。

Visit ridgelineapps.com to see what they can unlock for your firm.

访问ridgelineapps.com,看看他们能为你的公司释放什么潜力。