#62 - 开放性如何成为AMD的AI战略 - Andrew Dieckmann

好的。

Alright.

安德鲁，感谢你参加《深度视角对话》节目。

Andrew, thank you for being on the Deep View Conversations.

对于不熟悉你的人，请简单介绍一下你在AMD的工作内容。

For those who aren't familiar, tell us a little bit about, what you do here at AMD.

我是AMD数据中心GPU部门的总经理。

So I'm the general manager of the data center GPU, group at AMD.

这包括我们的Instinct产品，包括我们今天投入生产的Helios机架。

So that involves our instinct products, including the Helios rack that we launched into production today.

我的工作是管理整体业务、客户关系、产品管理团队，以及负责整体的盈亏和未来的业务发展。

So my job is to manage the overall business, customer relationships, product management teams, manage the overall P and L and business moving forward.

我们会聊很多关于Helios的内容，但在此之前，你能跟我聊聊，现在每天让你最兴奋地去上班的事情是什么？

We're gonna talk a lot about Helios, for But before will you, talk to me a little bit about, you know, what's the stuff that gets you most excited to come into work every day right now?

哦，天哪。

Oh, gosh.

我职业生涯是从工程师开始的，所以我对技术本身很兴奋，但变化的速度也让人兴奋，同时也很有挑战性。

I mean, I started out my career as an engineer, so, I mean, I get excited about the technology, but the pace of change is very exciting, also challenging.

所以，我觉得让我兴奋的是，每天、每周都能学到很多东西。

So, I think just what gets me excited is every day, every week, you learn so much.

总有新的挑战需要解决。

There's always new challenges to solve.

我们解决的问题范围很广，不只是芯片本身。

The breadth of what we're solving for, you know, it's not just silicon.

芯片是我们工作的基础，但还包括数据中心规划，从基础工具到芯片和系统的制造，以及中间的所有环节，还有正确的市场策略和客户关系。我觉得每周都有很多多样性和有趣的问题要解决。

I mean, silicon is the foundation of what we do, but doing data center planning and everything from picks and shovels to building the chips and the systems and everything in between, and the right go to market motions and customer relationships, I think it's a lot of diversity and interesting problems to solve every week.

对，很好。

Yeah, very good.

你的职业生涯中，你接触过基础设施的很多不同领域。

You've worked in your career across a lot of different areas of infrastructure.

这些经验对你现在的工作，尤其是Helios项目，有什么帮助？

How has that prepared you to work in the work that you're doing now, especially with Helios?

但更广泛地说，AI的变化也改变了AMD产品的需求形态，或者说需求的形状，以及我们正在打造的产品，看起来都大不一样了。

But but even just more broadly, the way that AI, you know, has changed has meant sort of the demand or maybe the shape of the demand, and of the products that AMD is building, you know, look a lot different.

对吧？

Right?

是的。

Yeah.

我觉得，我大概四年前加入AMD担任这个职位。

I think, so I joined AMD in this role about four and a half years ago.

幸运的是，我在半导体和数据中心基础设施产品方面有多年的经验，可以借鉴。

And so it's been Fortunately, I had a lot of background to draw on from working in semiconductor for data centers, infrastructure products for many years.

所以我对数据中心及其生态系统非常熟悉。

And so, I was very familiar with data centers and the ecosystem around data centers.

我觉得这些经验确实让我为现在面临的挑战做好了准备。

And I think that luckily has prepared me well for the challenges that we're embracing now.

而且，就像我之前说的，AI扩大了我们在产品上市时需要考虑的范围，因为不仅仅是发布一款芯片，让你去设计系统，或者给你一个基本的参考设计。

And I think AI has, like I said earlier, expanded the scope of what we think about when we bring a product to market because it's not just, Hey, we're releasing a chip, have fun designing it into a system, or, Here's a basic reference design.

现在已经发展到我们今天公布或披露的Helios系统了。

It's moved to what we announced or disclosed more today, the Helios system.

这是一个完整的机架级架构。

It's a full rack-scale architecture.

虽然AMD不直接向客户销售这个平台，但我们正在推动一个生态系统来构建这种机架级架构。

And although AMD is not selling that platform directly to the customers, we're enabling an ecosystem to build that that rack scale architecture.

这要求AMD在设计工作方面非常明确，包括供应链的运营方面，确保所有零部件以适当的数量存在，让合作伙伴能够获得，并确保机架高度可靠、易于维护，以及所有这些方面。

It requires AMD to be very prescriptive in in terms of the amount of design work that we do, the operational aspect of plumbing all the supply chain to make sure that all of these piece parts exist in the appropriate quantities for our partners to be able to obtain, make sure that the rack is, you know, highly reliable, serviceable, and all of those things.

所以这确实扩展了我们的思考范围，是的，我们是一家芯片公司，但我们远不止是一家芯片公司。

So it's really expanded the scope of what we think about, you know, yes, we're a chip company, but we're much more than a chip company.

交付这些解决方案需要大量的软件和系统工程。

There's an enormous amount of software and system engineering that goes into delivering these solutions.

你知道，计算是推动AI行业前进的燃料。

You know, compute is the, you know, the the fuel that drives the AI industry forward.

而如今关于计算的信息非常混杂。

And there's a lot of mixed messages around, compute these days.

你知道吗？

You know?

就是投入的资金太多了。

It's that there's so much that's being spent.

有些企业担心成本问题。

There's enterprises that are worried about their costs.

还有，超大规模云服务商和AI模型提供商担心资源不够用。

There's, you know, hyperscalers that are and and AI model providers that are are concerned they don't have enough.

比如，需求飙升，他们担心会供不应求。

Like, the the the demand is, you know, going through the roof, they're not gonna have enough.

而且总有一天，他们会遇到瓶颈。

And at some point, they're gonna, you know, run out of tracks.

对吧？

Right?

就像火车会跑到轨道尽头一样。

The train's gonna run out of tracks.

这些信号确实挺多的。

Those are a lot of different signals.

那你们在AMD是怎么把这些不同的信号梳理清楚，从而准确把握当前行业状况的呢？

Like, how do you all, at AMD, sort out all of those different signals to understand the industry right now?

说实话，这段时间变化真的非常大，而且未来还会持续有大量变化。

I mean, been a tremendously There's been a lot of change in a short period of time, and there will continue to be a lot of change.

事实证明，每次我们以为算力需求已经满足了，但每隔半年、一年，甚至一个月，我们就得重新评估市场机会的上限。因为客户给我们的反馈是，他们切实看到了这项技术的价值，它在业务成果上带来了实实在在的好处，所以他们愿意为此投入。

The thing that has proven to be true is every time we think, Oh, we might have solved for enough compute, there's like, incrementally, every six months or twelve months, sometimes on a monthly basis, had to raise our view of the market opportunity, because the signals that we're getting from our customers is they see real value in the technology, it's providing real benefit in terms of business outcomes, and therefore, they are investing in it.

我觉得，大家争论的焦点在于这笔投资的规模，以及它来得有多快。有人可能会问：是不是太快了？

And it is, I think, the scale of that investment and how quickly that investment has come upon us that if people debate, hey, is it too fast?

是不是太多了？

Is it too much?

但这恐怕是迄今为止地球上规模最大的基础设施项目之一。

Is it but it's one of the probably the largest infrastructure project on planet Earth that we've had to date.

对吧？

Right?

我觉得这些都是非常明智的决定，是由有远见的决策者做出的，他们投入巨额资金来建设这些基础设施，是基于对未来的承诺。

I think these are very well educated decisions that being made by well educated stakeholders that are spending enormous amounts of money to build out this infrastructure based upon the promise of

技术的承诺，以及它真正带来的经济价值。

technology and really the economic value it provides.

对。

Yeah.

好的。

Okay.

所以现在有个现象，连普通大众都能看到数据中心在建设。

So there's this story, you know, even the wider, public sees data centers being built.

对吧？

Right?

所以现在关于数据中心的讨论其实挺有争议的。

And and so there's this very challenging narrative around data centers right now.

是的。

Yeah.

当你们考虑这些的时候，你们怎么讲述这些故事，比如说，哦，这些会带来的好处是某某某，这样人们就不会只想着，为什么建这些大建筑，以及这对电力、水资源或其他方面会有什么影响。

When when you all think about that, how do you tell the stories about, like, oh, well, the the benefits that this is gonna bring are x so people don't just think about, well, why are these big buildings being built, and what is that gonna have impacts for power or water or other things.

你们会考虑这些吗？

You know, do you all think about that?

你们有没有想过怎么帮助讲述这些事物的好处？

Do you think about how can we help tell the story of the benefits of these things?

对。

Yeah.

我的意思是，我觉得我们在建智能工厂，对吧，是的。

I mean, I I think there's we're building intelligence factories, right, which Yeah.

这些工厂在为社会提供价值方面有很多潜力，对吧？

Have a lot of promise in terms of what they can provide in terms of value to society, right?

而且我觉得社会必须想办法，我们必须共同努力，这真的是全人类的事情，确保这些系统的价值以有益的方式积累。

And I think society has to figure out, we have to work as, really, it's a humankind thing, to make sure that the value of these systems accrue in a beneficial way.

就像任何强大的工具一样，它也可能被滥用。

And like any powerful tool, it can be misused.

它也可以被用在非常有用的地方，对吧？

It can be put to very beneficial use, right?

我们在医疗和教育领域看到了很多这样的好处，而且还有很多很多。

And we see a lot of the beneficial uses in medicine and education, and mean, the list goes on.

比如，以前只有很少一部分人能请得起私人教师来辅导孩子。

It's like, you used to, you know, a very small portion of society could afford a personalized tutor for their child or whatever.

现在，只要口袋里有钱，每个人都能拥有一个私人教师，真正加速孩子的学习体验。

Now you get a personalized everyone that has $20 in their pocket can get a personalized tutor to really accelerate the educational experience for children.

这只是一个例子。

That's just one example.

但是，是的，我认为我们必须以负责任的方式建设数据中心，要考虑到所在社区的感受。

But, yeah, I think it's we have to build data centers in a responsible way, in a way that's sensitive to the communities that they're in.

我们必须利用所有可用的技术，确保这些建设最终对所在社区是有益的、有增值的。

We have to make sure that we are using all of the technologies at our disposal to make sure that they're done in a way that is ultimately accretive and beneficial to the communities that they're in.

你们在谈论Helios和很多其他技术时，总是提到更开放生态系统的种种好处。

You all talk about when you talk about Helios, when you talk about a lot of your technologies, you talk about the open the the benefits of a more open ecosystem.

生态系统。

Ecosystem.

对。

Mhmm.

你觉得这是你们为对话带来的好处之一，也是对生态系统本身更广泛的好处吗？

Do you see that as one of the parts of the good that that you all are bringing to both the sort of the dialogue, but but broader sort of good to the the ecosystem itself?

比如，锁定效应更少了。

Like, there's there's less lock in.

选择更多了。

There is more choice.

你的客户和合作伙伴有更多机会去塑造产品的样子以及它的部署方式。

There's more opportunity for your customers and partners to sort of shape the what the product looks like and how it gets deployed.

这是你在思考时考虑的因素之一吗？你如何看待它作为你正在构建的重要方面之一？

Is that one of the things when you think about and how do you think about that as one of the sort of important aspects of what you're building?

我会说，拥有开放生态系统是AMD的核心原则。

I would say it's a core principle at AMD to have that open ecosystem.

我是说，我们宣布了广泛的产品线，从客户端到边缘再到数据中心的Helios。

I mean, we announced a wide breadth of products that we're doing, everything from client to edge to data center with Helios.

AMD拥有适用于多种不同应用的广泛产品组合，但我们并不想锁定客户。

AMD has a broad portfolio for many different applications, but we're not about locking customers in.

我们认为客户应该有选择权，而且没有一种通用的解决方案，丽莎经常说，没有一种计算方案能适用于所有情况，需要优化，而且市场如此广阔，应用种类繁多，没有哪家公司能提供所有这些。

We feel like there is customers should have choice, and there's not a one size, you know, Lisa constantly says, there's not one compute solution for everything, there's optimizations, and it's such a broad market, and with a variety of applications that no one company is going to provide all of those things.

现在，我们认为AMD可以提供其中的许多，并且我们很乐意在多个领域提供非常有竞争力的解决方案。

Now, we think AMD can provide many of them, and we're happy to provide very competitive solutions in a number of areas.

但无论是像Helios这样的特定平台，我们都有开放的机架架构，让客户可以选择不同的组件。

But whether it's specific platforms like Helios, we have an open rack architecture, which allows customers to have choice on different components.

可能更重要的是，我们拥有开源的软件栈。

Probably more importantly, we have an open source software stack.

这可以说是，我不想过度使用“改变游戏规则”这个词，但我的意思是，随着代理式AI的出现以及它加速软件开发的能力，我们的关键合作伙伴，就像你在台上听到的，无论是OpenAI、Anthropic还是Meta，他们以及其他非常有价值的合作伙伴都为这个开源软件栈做出了贡献，这确实加速了我们通过开源努力为广泛客户群提供更多价值的能力。

And that has been, I would say, I don't want to overuse the word game changing, but I mean, with the advent of agentic AI and the ability for that to supercharge the software development, the fact that our key partners, like you heard on stage, like whether it's OpenAI or Anthropic or Meta, they contribute, you know, them and other, you know, very valuable partners contribute to that open software stack, and it's really supercharged our ability to provide more value to our to our broad customer base through that open source endeavor.

是的。

Yeah.

你提到了生成式AI。

So you mentioned the genetic AI.

你知道，2026年的这个轨迹在一月份，第一季度，一月份、二月份发生了很大变化。对。

You know, this 2026 trajectory changed a lot in January in the first quarter, in January, February Yeah.

当OpenClaw出现，那个智能体时刻到来的时候。

When sort of the agentic moment happened with OpenClaw.

OpenClaw只是冰山一角。

OpenClaw was, like, the very, very tip of the iceberg.

从那以后发生了太多事情。

Like, there's been so many things that have happened, you know, since then.

这如何让你们重新调整了你们的工作，对需求的看法，以及你们构建产品的方式？

How has that had caused you all to recalibrate even what you do, the way you think of demand and and sort of the way you build things?

因为你们提前多年规划，来制定你们的产品路线图。

Because you plan years ahead of time to build the road maps that you build.

是的。

Yes.

那么，过去六个月世界的变化，这到底是怎么运作的？

So how does that work, the change in the world in the past six months?

对，我们的规划周期，芯片的开发周期从头到尾差不多要两年，对吧？

Yeah, our planning cycle, I mean, the development cycle of a chip is close to two years, start to finish, right?

所以我觉得，在每个规划周期和开发周期中，我们确实学到了很多。

And so I think, as learn a great deal during every planning cycle through every development cycle, really.

我们尽量保持灵活性和模块化。

And we try to retain optionality and modularity wherever possible.

我们不断重新评估，有时候外界不知道的是，我们内部有A计划、B计划、C计划，我们一直在审视和调整，这个和那个的比例怎么定才合适？

We constantly reassess, and sometimes, what's not publicly reported is we have plan A, plan B, plan C inside that we're constantly looking at and adjusting, what's the right ratio of this to that?

在这个特定的系统领域，什么才是正确的解决方案？

What's the right solution in this particular system area?

或者说，我们硅产品的芯片架构已经证明非常有价值，我们能够逐步推出产品变体，或者在开发过程中稍微调整方向，而且我们和客户之间一直有反馈循环，因为他们也在学习。

Or our chiplet architecture for our silicon products has proven to be extremely valuable, and that we've been able to incrementally spin off either product variants, or shift a little bit what we're doing partway through the development cycle, and we constantly have a feedback loop with our customers, because they're learning too.

就像，我们都在学习，坐下来聊的时候，就会说，好，这周你学到了什么？

Like, we're all learning, like, we sit down, it's like, Okay, what did you learn this week?

我这周学到了什么？

What did I learn this week?

我们合作得非常紧密。

And we collaborate really closely.

有很多协同设计。

There's a lot of co design.

在开发周期中，我们会收到客户的很多功能需求。

We get feature requests from our customers well into the development cycle.

我们不会忽视这些需求。

We do not ignore those.

我们会评估如何应对，有时候这会让我们的工程师抓狂，因为我们会说，好的，我还需要这个功能。

We assess how do we It drives some of our engineers crazy sometimes, because we're coming with, Okay, I need this one more thing.

但我觉得，因为我们的芯片架构是模块化的，软件栈是开放的，这给了我们很大的灵活性，可以把这些新学到的东西融入到我们的产品交付中。

But I think due to our silicon architecture being modular, our software stack being open, it's given us a lot of flexibility in terms of incorporating that new learning into our product deliverables.

所以，世界

So the world

随着智能体的出现，世界有点朝着AMD的方向发展了。

with agents, the world has kind of moved in AMD's direction a bit.

对吧？

Right?

这么说公平吗？

Is that fair to say?

我们确实有这种感觉。

I we feel that way.

是的。

Yes.

因为在那之前，大家主要关注的是GPU和AI加速器。

Because before that, it was very GPU AI accelerator focused.

而有了智能体之后，重点更多转向了编排调度。

And with agents, it's much more orchestration focused.

CPU变得不可或缺，你知道，在那种需要主导地位的情况下。

The CPU is needed, you know, where you have leader much more sort of a leadership sort of position

对。

Yeah.

在服务器CPU方面。

In in server CPU.

那么这对你们的路线图以及你们推出这项技术的方式，包括今天宣布的那些内容，意味着什么呢？

And so what has that meant for the roadmap, as well as sort of the way you're rolling out this technology, the stuff that was announced even today?

是的，有几个不同的方面。

Yeah, in a few different areas.

首先，能为客户提供更多价值总是好的，对吧？

I mean, one, it's always nice to be able to provide more value to your customers, right?

所以现在，我们的主要客户需要大量的GPU算力。

And so, now, our major customers, they need oodles of GPU compute.

他们也需要大量的CPU算力。

They also need oodles of CPU compute.

正如你所说，我们在CPU领域有着非常稳固的领先地位，这是我们多年来一直在努力的结果。

To your point, we have a very established leadership position in CPUs, which we've been working on for many years.

所以，我觉得这让我们能拿出更多客户真正关心的东西来。

So, I think it's allowed us to bring more ingredients to the table that our customers really care about.

所以，这挺好的。

And so, that's been nice.

而且，即使在GPU系统里，我们高性能CPU的优势也会在整体系统性能上体现出来。

And then, also, even within the GPU systems, you know, the benefit of our high performance CPUs actually comes forward in some of the total system performance.

所以，这既关乎GPU那边的芯片和架构，也关乎整个系统的结构，因为我们的客户买的不是零件，而是整个系统，而系统性能，无论是每瓦性能还是每美元性能的竞争力，才是衡量好坏的标准。

So, it's both, the silicon and architecture on the GPU side, but also, it's the whole system structure, because, again, our customers are not buying components, they're buying a system, and the system performance, both perf per watt, perf per dollar competitiveness is really the metric of, is it good?

跟竞争对手比怎么样？

How does it compare to our competition?

这些都很重要。

All that stuff.

因为这一点，还带来了一个非常惊人的投资回报率的故事。

There's this incredible ROI story that's also happened because of this.

因为智能体做了太多事情，企业的成本飙升，他们正为此感到头疼。

Because of agents doing so much, you know, costs for enterprises have soared, you know, and they're feeling the pain of that.

这给你们带来了多大的压力？

How much pressure does that put on you all?

客户是不是会跟你们说，我需要这个更高效？

Do you get from customers of like, I need this to be more efficient.

我需要它做更多的事情。

I need it to do more.

我需要算力，但它必须更高效。

I need the compute, but it's gotta be more efficient.

我的每瓦成本、每瓦智能都要更好，诸如此类的。

My cost per watt intelligence per watt needs to be better, all of these kinds of things.

这些压力有多少会传到你们这边？

How much does that pressure come to you all?

我觉得这是每一代产品的基本设计目标之一。

I mean, I think it's it's one of the foundational design goals for every generation.

比如，我们需要提升性能、每瓦性能、每美元性能。

Like, we need to improve performance, performance per watt, performance per dollar.

我们需要给客户带来实实在在的进步，你知道的。

We need to, you know, pass a good amount of incremental improvement to our customers.

他们也需要把这些进步传递下去。

They need to pass that on.

不管是云客户，他们每一代都得能用更少的成本做更多的事，因为需求是指数级增长的。

Whether it's cloud customers, they need to be able to, every generation, do so much more for less because the demand is exponential.

这确实是个核心原则。

I mean, that's a core principle.

这会让我们更努力吗？

Does it make us try harder?

我不确定这会不会让我们更努力。

I don't know that it makes us try hard.

我们一直都是全力以赴的。

We were like, everything was You're always working out to the max.

对。

Exactly.

所有这些参数都调到了最高，我们对此进行了激烈的讨论。

All those knobs are dialed up, and we debate that vigorously.

我认为这就是AMD的核心运营方式。

And I think that's just core operating behavior at AMD.

但它的重要性，我觉得，只是因为支出在增长，而且问题越来越大，对吧？

But the importance of it, I think, just because the spend is growing, and it's a larger problem, right?

因为，我的意思是，世界每年部署的计算量比过去几十年都要多，对吧？

Because, I mean, the world is deploying more compute per year than in previous decades, right?

而且这个速度还在加快。

And that's accelerating.

所以，我觉得问题的规模，就像你说的，部署的规模确实令人望而生畏，所以每一个百分点的效率都非常宝贵，非常，对。

So, I think just the scale of the problem is, to your point, I mean, the scale of the problem is, the deployments is daunting, and so every percentage point of efficiency is extremely valuable, extremely Okay.

今年年初以来，我们听到了很多关于竞争产品的消息，比如Nvidia的Vera Rubin，这就是一个极端的例子，你知道，它将用来构建下一代模型，解决最大的问题。

We've heard a lot this year, the beginning, starting at beginning of the year, around competitive products, Nvidia's Vera Rubin, as like, this is the extreme example of, you know, what's gonna build the next models, what's gonna solve the biggest problems in that.

Helios，据我所知，是第一个达到同等水平的竞争产品。

Helios is is sort of the first thing that's the first competitive product, my understanding, that's on that, you know, same level.

你觉得Helios有什么独特之处？你跟客户聊的时候，或者你讲Helios的故事的时候，你是怎么讲的？

When you think about what's unique about Helios, when you talk to customers, when you sort of tell the story of of Helios, you know, how do you tell the story?

我觉得Helios的故事就是，它是一个非常有竞争力的机架级架构。

I think the story of Helios is it is extremely competitive rack scale architecture.

我们认为，在规格展示中，我们拿出了性能最高的GPU，能效比最好，绝对性能也是最好的。

We think we have produced in the spec show, it's the highest performing GPU, best performance per watt, best performance just straight out absolute performance.

把这些集成到机架里，我们相信能提供市场上最好的单机架AI性能。我们今天展示了很多规格，我们认为在负载性能上比Vera Rubin机架好10%到15%。

Those integrated into the rack, we believe, provides the best single rack AI performance in the market, and we showed a bunch of specifications today, and we think we're 10 to 15% better workload performance than a Vera Rubin rack.

当然，我们会继续评估他们的产品，等他们上市的时候，我相信他们也会评估我们的。

You know, we will continue to evaluate their products as they come to market, as I'm sure they will evaluate ours.

是的，我觉得这是个巨大的市场。

But yeah, I think it's a huge market.

有很多公司都有机会参与进来。

There is an opportunity for many companies to be engaged.

这不是零和游戏。

It's not a zero sum.

这不是一个零和游戏，我觉得我们正在尽最大努力，通过那些深度的联合工程合作，为客户提供强大的价值。

It's not a zero sum thing, and I think we're doing our best to provide just strong value to our customers through the those deep co engineering partnerships.

而且

And

我觉得这是一个健康的市场，有多家公司参与并激烈竞争是好事。

I think it's a healthy market to have multiple companies engaged and competing hard.

你们还没有透露Helios的价格，但有一些行业报告提到了它的成本，而且比Vera Rubin更贵。

You all haven't said how much Helios costs, but there are some industry reports about how much it costs and that it is more expensive than than Vera Rubin.

你们有没有一些指标，比如每性能、每瓦特的成本？

Do you have any metrics in terms of, the cost kind of, you know, per performance, per watt?

你们觉得在单位成本上，它是否具有竞争力，是否更好？

Do you do you feel like it's competitive, that it's better when you get to the down to the, like, what you're paying sort of per unit cost?

是的。

Yeah.

所以，Helios并不是所谓的比Vera Rubin机架更贵。

So, Helios is not, quote, more expensive than a Vera Rubin rack.

好的。

Okay.

我觉得我们

I think we

是以合作的方式与客户互动，我认为所有理性的客户都愿意为产品带来的价值买单。

engage our customers in a collaborative way, and I think all RATIONAL customers are willing to pay for the value that the product provides.

所以，如果他们的评估和我们类似——我们也认为确实如此——即这是全球性能最高的机架，他们愿意为这个机架支付合理的价格。同时，我们也清楚自己是这个市场中相对不那么成熟的竞争者。

So, if their assessment is similar to ours, which we think it is, that it's the highest performing rack in the world, they are willing to pay a reasonable amount of money for that rack, and we are also cognizant of the fact that we are the less established competitor in this particular market.

所以，我们在市场上表现得很积极，但也很理性，我们的客户同样理性，他们愿意为我们提供的价值付费。

So, we're being aggressive in the market, but also very rational, and our customers are very rational, they're willing to pay for the value that we're providing.

我们认为自己会处于有利位置来获取市场份额，并且以一种对客户和股东都有利的方式来实现这一点。

We think we're gonna be well positioned to gain share, and do so in a way that is beneficial to our customers as well as to our shareholders.

当我

When I

想到Helios时，你会考虑它首先部署在哪些地方。

think of Helios, you you think about the first places it's gonna be deployed.

你已经把它卖给你的第一批客户了。

You've already sold it to your first customers.

是的。

Yeah.

我敢肯定，你甚至已经把它卖给了第二批客户，在那些产品还没造出来之前，因为计算需求太大了。

I'm sure you've already sold it to, like, your second wave of customers, you know, too before those are even made because there's so much compute demand.

但说到第一批客户，当我想象谁会使用这类系统，比如Helios或Vera Rubin这个级别的，我觉得——如果我说错了你纠正我——是那些超大规模云服务商，可能还有那些新兴的大型云服务商，以及前沿实验室，他们才需要那种性能水平。

But those first customers, when I think of, like, who is gonna use these sort of systems, the level of the Helios or a Vera Rubin, I think of it, and you tell me if I'm wrong, that it's the it's the hyperscalers, maybe the neo the larger neo scalers, the, you know, cutting edge frontier labs, that they're the ones who need, you know, that level of performance.

不是每个人都需要Helios的性能，你们有不同级别的产品来满足不同层次的客户需求。

Not everybody needs the performance of a Helios, and you have multiple levels of products that can serve multiple levels of customers.

对。

Yeah.

那谁会是第一批部署它的客户呢？已经买了的那些。

Who who are the first ones that are gonna be deploying it that that have already bought it?

所以，

So,

我们有一些大客户，已经承诺了。我们跟五个非常大的客户谈过部署的事情。

some large committed customer I mean, we have five very large committed customers that we've talked about deployments with.

最近的一个是Anthropic，我们跟他们发布的新闻稿里提到，Helios的部署规模最高可达两吉瓦。

The latest one's Anthropic, so we've discussed that the press release with them was up to two gigawatts of Helios deployments.

他们显然会是一个大客户。

They will obviously be a large customer.

OpenAI是我们去年宣布的第一个大客户。

OpenAI was the first large customer we announced last year.

Meta，我们今年春天宣布了，是一个跨多代、多吉瓦的客户。

Meta, we announced this spring as a multi generational, multi gigawatt customer.

这三个客户作为基础模型构建者，会部署很多。

Those three customers will deploy a lot as foundational model builders.

微软，我们周一发布了公告，他们会在大规模部署Helios。

Microsoft, we had an announcement on Monday, where they're gonna be deploying Helios at scale.

甲骨文，我们通过Oracle Cloud有合作，他们正在部署大量容量，然后我们还跟一些Neo Cloud合作伙伴有合作，他们也在大规模部署Helios。

Oracle, we have a partnership through Oracle Cloud that they're deploying a bunch of capacity, and then we have partnerships with a number of Neo Cloud partners who are also deploying Helios at scale.

我觉得，回到你问的谁需要这种机架级架构，基本上任何在前沿领域工作的人都是这个系统的主要候选者，任何做大规模预训练的人。

And I think, to your question of who needs that rack-scale architecture, I mean, anybody who's working at the frontier, basically, is a primary candidate for the system, anyone who's doing any large scale pre training.

但是，那些为几亿甚至几十亿用户提供前沿模型、大型模型服务的公司，这也是提供这种规模推理的最有效方式。

But then, these guys that are serving frontier models, large models, to hundreds of millions or billions of users, it is the most efficient way of providing that scale of inferencing also.

所以，我觉得这些是主要客户，但我们还在推进主权倡议，在欧洲和亚洲，都有国家希望部署。

So, I think those are kind of the primary customers, but we have sovereign initiatives that we're working on, where both in Europe, as well as Asia, countries looking to deploy.

所以我觉得这不是一种专门的设备。

So I think it's not a specialized piece of equipment.

对于想要大规模训练或推理的人来说，这是一个很好的解决方案。

For people who are looking to do training or inference at large scale, it's a great solution.

而且，正如你所说，我们还有其他本地部署的解决方案，比如那些可能没有液冷数据中心的地方，我们有风冷解决方案，从PCIe卡到八路服务器，再到边缘和笔记本电脑上的推理。

And to your point, we have other solutions for on prem, where people maybe don't have liquid cooled data centers, or we have air cooled solutions, everything from PCIe cards to eight way servers, to inferencing on the edge and laptops.

所以，有一系列的解决方案，对吧。

So, there's a spectrum of solutions, Okay.

但是，你知道，Helios解决方案真正针对的是大型云部署，用于大规模训练和推理。

But, you know, the the the Helios solution is really targeted to the the large cloud deployments for at scale training and inferencing.

好的。

Okay.

你还宣布了与Cerebras的合作。

You also announced a partnership with Cerebras.

是的。

Yes.

给我讲讲这个合作吧。

Tell me a little bit about that.

这个合作主要是关于什么的？

What's the, what's that partnership, focused on?

对。

Yeah.

这是推理领域的一个新兴方向。

So that is an emerging segment of inference.

推理市场非常大。

So inferencing is such a large market.

现在推理市场正在出现细分。

There's segmentation now that's happening.

在低延迟服务方面，Cerebras 是做得最好的。

And the low latency serving is something that Cerebras is best in class at.

他们的晶圆级引擎加上 Helios，能让我们把这个服务推广给更多客户，这样他们单独运营的成本效益就大大提高了。

And the combination of their wafer scale engine with Helios allows us to scale that offering to a larger set of customers, so it dramatically improves the economics of their offering on a standalone basis.

他们能把推理任务拆分开，比如预填充和解码，这样就能用合适的工具做合适的事。

The fact that they can disaggregate the inferencing workload, you have a pre fill and a decode, you sort of use the right tool for the right job.

比如，我们的令牌吞吐量很高，而他们的低延迟响应非常强。

Like, we have incredible token throughput, they have very strong low latency responses.

把这两者结合起来，就能以更低的每令牌成本提供低延迟服务，而不是单纯扩展晶圆级引擎。

You kind of combine those things, and it allows the service to provide low latency attributes at a lower cost per token than just scaling out the wafer scale engines independently.

这就是这个合作的核心。

So that's what the partnership is all about.

好的。

Okay.

最后两个问题，我问每个来播客的人。

Last two questions I ask everybody, you know, that comes on the podcast.

第一，在人工智能时代，领导者经常用“杠杆”这个词。

One, in the age of AI, leaders use this term leverage a lot.

比如，我怎样才能最大化利用我的时间？

Like, how do I get maximum leverage for my time?

那么，对于那些真的想最大化利用时间的领导者，在人工智能时代，你最好的建议是什么？

And so what's your best tip for for leaders who are really trying to maximize, you know, their time in the age of AI?

我觉得市场变化这么快，我们刚才也聊到了一点，这迫使你需要重新评估。

I think the the amount of change in the market, so quickly, we discussed that a little bit, it causes you to need to reassess.

不是说要做什么新的事情，而是你必须无情地优先排序，而且现在得更频繁地这样做。我发现，因为现在每个月都像过了一年一样，我得不断审视我的时间花在哪里，怎么把事情安排得井井有条，因为总有新工具不断出现，可以用来提高时间效率。

It's not new things, but you've got to ruthlessly prioritize, and you have to do that more frequently now, I find, because it's just like every month is like a year went by in terms of where am I spending my time, and how to have things organized appropriately, because there's new tools coming to coming to bear all the time that you can use to make your time more effective.

所以，我觉得关键就是不断评估这些。

So, I think it's really evaluating those.

毕竟，作为人类，我们自身也有处理能力的限制，对吧？

I mean, as human beings, we have a throughput limitation of our own, right?

你得留出时间，真正停下来评估一下，然后重新调整方向。我发现现在自己得比AI时代之前更频繁地这么做，纯粹是因为变化速度太快了。

You have to take the time to actually pause assess, and then recalibrate, and I find that for myself, I have to do that more frequently now than in the pre AI era, Just because of velocity.

就是这种速度。

Velocity.

对。

Yeah.

完全是这样。

Purely.

好的。

Okay.

那有没有什么AI工具是你现在正在用的，可能大家还不太了解，但你觉得值得关注的？因为它真的能改变游戏规则。

How about what's the AI tool, you know, that you're using right now that maybe not everybody knows about that you think people should be aware of because it could really change the game?

我们公司今天也聊到了这一点，我们正在使用合作伙伴提供的编程工具，这对提升工程师的效率起到了颠覆性的作用。

I mean, as a company, we talked about it today, we're using the coding tools from our partners have been game changing in terms of the effectiveness of our engineers.

正如Vamsi在主题演讲中提到的，这些工具解锁了以前根本不可能实现的功能。

And it has, as Vamsi talked about in the keynote, it's unlocked things that were just not possible before.

公司规模大小其实无所谓。

It doesn't matter how big your company is.

我们有三万名工程师。

I mean, we have 30,000 engineers.

我们不是小公司，但某些技能的人，你永远招不够，根本没法完成你想做的事。

We're not a small company, but you could never hire enough of certain skill sets to actually get done what you wanted to get done.

确实。

Sure.

所以现在，我觉得像内核优化这种组合，就是个很好的例子。

And so, now, I think the combination, like, kernel optimization is a perfect example.

在这种事情上，智能体的倍增效应对我们来说真是改变游戏规则。

The agentic force multiplier on something like that has been a game changer for us.

我们能做到以前做不到的事，为更多不同的模型做更多优化，而且真的让我们能把领先的硬件交到更多做领先工作的人手里。

We've been able to just do things that weren't possible before, have more optimization points for more different models, and it's really been an unlock in terms of ability to get our leadership hardware into the hands of more people doing leadership things.

所以我觉得，过去六个月里，这真的是一个突破，又向前迈了一大步。

So, I think that has been and that's been an unlock, really, in the last six months that took another step function forward.

所以我觉得这非常令人兴奋，而且对我们的客户和业务都极其有益。

So I think that has been very exciting to see, and it's been extremely beneficial to our customers and to our business.

你今天发布的那个新盒子，基本上就像桌上放着一支智能体团队，没错。

That new box that you announced today, essentially like a team of agents on a desk, Exactly.

你的团队已经在用这些了吗？

Is your team using those already?

我们已经把它们用在工作流程中了，而且我在尝试各种不同的工具和硬件，我们公司的情况也是一样的。

We're using those already as part of our workflow, and yeah, I experimenting with all types of different tools and pieces of hardware, and the story is the same in our company.

计算资源永远不够用。

There's never enough compute.

就像你给工程师更多计算资源，他们就会想要更多。

It's like you give the engineers more compute, they want more compute.

所以这是个好问题，但好问题也是问题，我们得解决它们。

So it's a good problem to have, but good problems are still problems, and we gotta solve them.

是的，非常好。

Yeah, very good.

安德鲁，谢谢你抽出时间。

Andrew, thank you for the time.

是的。

Yeah.

谢谢。

Thank you.

聊得很愉快。

Good stuff.

对。

Yeah.

很高兴和你聊天，杰森。

Great talking to you, Jason.