陈立武谈拯救英特尔：文化、客户与 CPU 复兴

听众朋友们，欢迎回到 No Priors。今天我和另一位主持人与陈立武（Lip-Bu Tan）一起，他是 Walden 的传奇投资人，曾任 Cadence 的 CEO，现在是 Intel 的 CEO。我们讨论了他改造 Intel 的计划、美国政府成为主要股东、如何成为出色的半导体投资者，以及我们能否在美国制造芯片。欢迎你，Lip-Bu。很高兴见到你。我们从最直接的问题开始：担任这家极其重要的美国半导体公司的 CEO 是一项非常艰巨的工作。你为什么要接受这份工作？

Hi listeners, welcome back to No Priors. Today a lot and I are here with Lip-Bu Tan, the legendary investor from Walden, then CEO of Cadence, now CEO of Intel. We talk about his plan to transform Intel, having the US government as a major shareholder, how to be an amazing semiconductors investor and whether or not we can make chips in the United States. Welcome, Lip-Bu. Lip-Bu, it's great to see you. We'll start with the obvious question. This is a really hard job to go be CEO of this incredibly important American semis company. Why take the job at all?

这是个好问题。我 66 岁了，人们觉得你应该退休，而不是接下这个行业里最艰难的工作。有几个原因：一是这家标志性公司对半导体生态系统至关重要，对美国也至关重要。所以我决定，在 Cadence 之后再干一次。

It's a good question. I'm 66 and people thought that, well, you should retire rather than taking on this hardest job in the industry. And so a couple reasons: one is this iconic company is so important for the semiconductor ecosystem and also so important for the United States. And so I decided, you know, do one more after Cadence.

过去一年发生了很多事。最让你惊讶的是什么？

A lot has happened in this past year. What has been the most surprising to you?

最让我惊讶的是，在我之前的工作甚至培训中都没遇到过——某天清晨，特朗普总统要求我因利益冲突辞职，没有任何例外。我首先说服自己，我不需要这份工作，我做这件事纯粹是为了拯救 Intel。所以把个人问题放在一边。然后我想，我能做些什么来帮助 Intel？好消息是，我周四早上有个会议，周一又开了会，他听我解释了自己的情况。我出生在马来西亚，在新加坡长大，去了 MIT，住在美国，从未在国外生活过。我分享了这些，他听得很认真，然后给了我机会，我很高兴。

Well, the most surprising that I didn't learn from my previous job or even training is one day early morning President Trump asking me to resign due to conflict of interest and there's no exceptions. And so I convinced myself, first of all, you know, I don't need this job. I do it purely to save Intel. And so take that personal issue out of the way. Then I figure out what can I do to be helpful to Intel. And so good news is I had a meeting Thursday morning and then Monday I had the meeting and then he listened to me like I had a chance to explain myself. You know, born in Malaysia, grew up in Singapore, went to MIT and I live in US and never lived outside country. And so something that I shared and that somehow he listened very well and then he gave me the chance and so I'm delighted.

现在你有机会做这份工作了。你说这份工作是拯救 Intel，这家公司非常重要。对你来说，这具体意味着什么？Intel 获胜或蓬勃发展是什么样的？

And now you have the chance to do the work. When you said the job is to save Intel, it's a really important company. What does that look like to you? What does Intel winning or thriving look like?

我刚过了 14 个月。这 14 个月里发生了很多事。有几件事：一是改变文化，明确要推动更多问责制，决策速度必须更快。我非常习惯创业文化，以光速行动，没有层层官僚的会议。所以我改变了一些东西：问责制，倾听客户，让客户满意。要谦逊，愿意倾听，解决他们面临的问题，努力让客户满意。另一部分：从第一天起，我决定所有工程部门都向我汇报。我是工程师出身，我想知道哪里出了问题，需要纠正什么，倾听客户，让客户满意，确保我们有正确的产品，简化产品线，真正拥有未来 5-10 年的路线图和愿景。

Yeah, I just passed 14 months. A lot of things happened in these 14 months. So a couple of things. One is to change the culture and then clearly want to drive more accountability and also in terms of decision-making have to be faster. You know, I'm so used to startup culture and you move fast at the speed of light and don't have that bureaucracy layer of layer of meeting. And so something that I change: accountability, listen to the customer, and the customer delighted. You know, someone like also humble, willing to listen and then address some of the problems that they face and then try to delight the customer. And also the other part: from day one I decided all the engineering reports to me. I'm an engineer by training, I want to know what went wrong and what are the things that I need to correct, listen to the customer and delight the customer, and then make sure that we have the right product, simplify our product line, and really have the roadmap and the vision for the next 5-10 years.

你对 Intel 10 年后应该处于什么位置有什么愿景？

What is your vision of where Intel should be in 10 years?

我认为有几件事。首先，我在 Cadence 和 Intel 时一直相信：先爬行，保持谦逊，倾听客户；然后开始走；最后开始跑和冲刺。这是我一步一步来的文化。第一步是加强我的资产负债表。资产负债表非常沉重，甚至有些糟糕。所以我很高兴美国政府成为大股东。我向特朗普总统解释过，台积电起步时也有台湾政府作为股东。看看日本、新加坡，这是美国政府提供支持的基础设施。其次，我很高兴我的老朋友黄仁勋也投资了 50 亿美元支持我。我很高兴至少我做了一些好事。他的 50 亿美元现在变成了 250 亿美元甚至更多。另一部分是软银的孙正义。我曾是软银董事会成员，他伸出援手帮助我。所以我们加强了资产负债表，然后专注于产品，我真正简化了产品，倾听客户，推动下一代领先产品。在某种程度上，这很幸运。现在 AI 和推理 CPU 的需求非常高。所以，与训练中 CPU 与 GPU 的 1:8 比例相比，现在我看到 1:4，甚至可能 1:1。我很高兴 CPU 变得重要。我和一些 AI 模型开发者聊过，他们说：“在强化学习方面，在编排所有智能体的速度方面，CPU 实际上更好。”所以在某种程度上，我很高兴。现在我的 CPU 需求非常高。所以我认为总体而言，在数据中心服务器端打造产品，另一部分是我们的代工业务。最初，这是一个资本密集型业务，不容易，你需要几样东西：你需要所有正确的 IP 来支持客户。例如，如果是移动相关，你需要低功耗 IP 集。没有这些，你就无法服务他们。这是一个服务业务，一个信任业务。如果人们给你订单生产晶圆，如果良率不好，他们就会因收入损失而完蛋。所以，我认为非常重要的是真正关注良率、缺陷密度和周期时间，确保你能高质量、可靠地满足和服务客户。这些是我真正关注的事情。最终，你必须真正转向全栈。所以不仅仅是硅片，你还需要软件，有些客户要求我：“给我整个机架。”所以你需要构建一个系统。

Yeah, I think a couple of things. One, I always believe in when I was at Cadence and also at Intel is first of all you crawl and then be humble, listen to customer, and then secondly you start to walk, and then finally you start to run and sprint. So, that's kind of my culture of step-by-step doing it. And then first step for me is to strengthen my balance sheet. And the balance sheet is really heavy, horrible in some way. So, I'm delighted the US government became a big shareholder. Just I explained to President Trump, TSMC when they started they had the Taiwan government as a shareholder. If you look at Japan, you look at Singapore, this is the infrastructure US government gets to provide the support. Secondly, very happy that Jensen Huang, my old time friend, he also put 5 billion in investing and support me. And I'm glad that at least I did some good work. His 5 billion became 25 billion now or more. And then the other part is SoftBank Masa. I used to be on the SoftBank board and then he lent a hand to help me. So, we strengthened the balance sheet and then focus on the products and I really simplify the product, listen to the customer, and then drive the next generation leadership products. And then in some ways it's very lucky. Right now the agenda AI and inference CPU become highly in demand. And so, versus 1 to 8 in the training CPU to GPU, now I can see 1 to 4, maybe 1 to 1. And I'm delighted CPU becomes important. I talked to some of the AI model developers, they said, "Oh, well, in terms of reinforcement learning, in terms of the speed of orchestrating all the agents, it turns out the CPU is actually better." And so, in some way, I'm happy. Right now, the demand is very high for my CPU. So, I think overall, build on the product on the data center server side, then the other part is our foundry business. And initially, this is a capital-intensive business, and it's not easy, and you really need to have a couple of things. You need to have all the right IP so that you can support the customer. Like, for example, if it is a mobile-related, you've got to have low power IP set that you need to have. If without that, you cannot serve them. It's a service business. It's a trust business. If people want to give you orders to have wafers come, if the yield is not good, they will be toast in terms of revenue miss. So, with that, I think it's very important to really focus on the yield, the defect density, and the cycle time, and then make sure that you really are able to meet and serve the customer in high quality and reliable. And so, those are the things that I really focus on. And eventually, you have to really move into a full stack. So, not just a silicon, you need to have software, and some of the customers asked me, "Give me the whole rack." So, there's a system that you have to build.

所以，我认为这些就是我一步步默默构建的事情，并招募我能找到的最优秀人才。顺便说一句，所有招聘都是我自己做的，没有猎头公司帮忙。所以，有时候拥有一个知道该联系谁的人脉通讯录是件好事。

And so, I think those are the things that I'm quietly building step by step and recruiting some of the best talent I can find. By the way, all the recruitment I do myself. No search firm helping. And so, I think sometimes it's good to have a Rolodex where you know who to reach out to call.

是啊，我的意思是，你在这一行干了这么久，而且，你知道，你之前掌管 Cadence 有，我想，12 年，所以……

Yeah, I mean, you've been in the business for so long, and you know, you ran Cadence for, I think, 12 years before this, and so...

准确说是 13 年。

13 years to be precise.

13 年。抱歉。对对对。那么你之后……

13 years. Sorry. Yeah, yeah, yeah. So, you did then...

再加上执行董事长的时间，总共 15 年。

More years as executive chairman, so 15 years.

年。我当时说只干 3 个月。所以现在我很小心。你说“我只干 3 个月”的那一刻，结果变成了 15 年。

Years. I signed up for 3 months. So right now I'm being very careful. The moment you said, "I'll just do it for 3 months." It turned out to be 15 years.

是啊，看来你在这里也有很长的路要走。那么，另一个被广泛讨论的重大举措是 Terafab 以及与 Elon Musk 的合作。你能再多讲讲这是如何形成的、你的参与以及你们是如何协作的吗？

Yeah, well, it seems like you have a lot of longevity ahead of you here, as well. And so, the other big initiative that has been sort of talked about is Terafab and working with Elon Musk on that. Can you tell us a bit more about how that came together and your involvement and how you're all collaborating?

是的，很好。我的意思是，Elon Musk，我想我们都同意，是本世纪最优秀的企业家之一，如果不是最优秀的话。他和我有相同的看法：半导体基础设施实际上没有跟上 AI 的增长。在需要产能、生产率和干法效率方面，这些都是他和我都认为缺失的东西。其次，我很高兴能与他合作。他非常——我称之为非传统。他基本上质疑每一步：为什么要用这种传统方式？在某些方面，这非常令人耳目一新。我喜欢这样——我喜欢有不同意见的人。我们一起努力，找到最佳路径，我们都会一起学到很多。而且我认为他显然有一个愿景：他的机器人和他的汽车需要大量的硅。

Yeah, good. I mean, Elon Musk, I think we all agree, is one of the best, if not the best, entrepreneur in this century. He and I share the same view that semiconductor infrastructure actually is not catching up with AI growth. And in terms of needing capacity, productivity, and dry efficiency, those are things he and I both see as missing. And secondly, I'm just delighted to work with him. He's very — I call it unconventional. He basically questions every step: why this traditional way of doing things? In some ways, it's very refreshing. I like that — I like people with different opinions. Let's work together, find the best route, and we'll both learn a lot together. And I think clearly he had a vision that his robots and his cars need a lot of silicon.

是啊，你能为不熟悉的人解释一下 Terafab 是什么吗？

Yeah, can you actually explain what Terafab is for people unfamiliar with it?

Terafab——他决定要建自己的晶圆厂。与此同时，我们很高兴与他合作，确保我们能一起努力，让他更快地投入生产，利用我们的一些技术和工艺。这是我们共同协作的事情。他有一个非常好的团队，我每周都和他们一起工作。与他合作真是令人耳目一新。

Terafab — he decided that he wants to build his own fab. And meanwhile, we are delighted to work with him and make sure we can work together to enable him to be faster and quicker to production, using some of our technology and some of our process. It's something we'll collaborate on together. He has a very good team that I work with weekly. It's just refreshing to work with him.

他还提到过一些事情，比如他希望你能在洁净室里抽烟，以及所有那些通常被认为……对对对。想想那 polka？是啊。

And he's talked about things like he wants you to be able to smoke inside the clean room and all these things that normally are considered... Yeah, yeah, yeah. Think the polka? Yeah.

我觉得我不会走那么远。我的意思是，也许洁净室的某些部分可以这样做。但我认为要保持开放心态，我们也在倾听并看看是否可行。

I think I don't go that far. I mean, maybe some part of the clean room you can do that. But I think something that is open-minded, and we are also listening and seeing whether we can do that.

是啊，看到你如何在美国逐步发展代工业务、与 Terafab 等项目合作，真是令人兴奋。如果你考虑全球 AI 和半导体供应链，比如说，你看到 AI 在宏观层面按国家推动的变化。当我观察某些国家时，比如那些声称因 AI 导致的裁员，我认为目前大多数都被夸大了。你知道，大多数裁员实际上只是 2020 年疫情期间过度招聘的修正。但我看到最先被削减的是外包公司，因为你会优先削减外部人员而非内部。所以你削减了外部客户支持，削减了外部 IT。我认为这对某些拥有大型 BPO 的国家影响更大，比如菲律宾、印度等。所以它们在短期内可能会受到 AI 的影响。然后如果你问公司如何以积极的方式参与 AI 的未来，你几乎必须逐个国家分析，对吧？能源便宜的地方会建数据中心。有能力训练模型的地方会训练模型，但这可能只有美国和一两个其他地方。你如何看待半导体行业全球供应链的转变？某些国家是否应该加大投资？比如以色列，考虑到 Mellanox、Nvidia 和 Intel 在那里的存在，是否应该做更多？他们是否应该在半导体领域做更多？菲律宾是否应该回归更多制造业基础？你从全球角度怎么看？

Yeah, I mean, it's very exciting to see how you're morphing the business here in the US in terms of incrementally building out the foundry business in terms of collaborating with things like Terafab. If you think about the global AI and semiconductor supply chain, so say that you were to look at the changes that AI is driving on a macro basis country by country. And if I look at certain countries when I look at the layoffs that are claimed from AI, for example, most of them I think are overstated right now. You know, most of the layoffs are actually just over-hiring during the 2020 COVID period. But the first things I see actually being cut are outsourced firms where you'd rather cut external headcount versus internal. So you're cutting external customer support. You're cutting external IT. And that has more of an impact I think for certain countries which have big BPOs, the Philippines, India, etc. And so they may be impacted in the short run by AI. And then if you ask how do companies participate in the future in a positive way in AI, you have to almost go country by country, right? Places with cheap energy will do data centers. Places with the ability to train models will train models, but that's probably only the US and one or two other places. How do you think about the shift in global supply chain for the semiconductor industry? Should certain countries invest more? Like should Israel be doing more given Mellanox, Nvidia, and Intel presence there? And should they try to do more in semiconductors? Should the Philippines move back to more of a manufacturing base? Like how do you think about that on a global basis?

是的，好问题。所以我认为 AI 显然正在改变整个格局。而且我认为其影响将比互联网更大，也更深远。所以我认为 AI 最初能够帮助你更高效地做事。然后随着许多智能体帮你处理那些平凡的任务，它们现在可以更快地完成。所以在某些方面，你可以大幅提升效率。即使在半导体设计中，你在时序方面能提升多少效率——你能多快推出产品？其次是成本。所以我认为这些都会推动发展。然后我认为 AI 需求和增长有几个瓶颈：一是，当然，大家都知道电力约束。一些国家根本没有那么多电力，它们会受到影响。其次，很多人没有意识到氦气的影响对半导体也可能非常显著。第三，大家都知道现在内存短缺更严重，每个人都在争抢内存。即使你想建晶圆厂来增加产能，也需要几年时间。CPU、GPU 也一样——所有这些都会需求旺盛。而且我认为价格也会上涨，因为我们必须把成本转嫁给客户。所以我认为这些会影响行业增长。然后总体而言，我觉得受影响最大的公司是那些没有拥抱 AI 的公司。因为 AI 可以帮助你在企业的所有不同职能中提升效率。我们应该拥抱并找到更好的方式，将 AI 用于预测、设计以及工作负载的各个部分。我认为这潜力巨大。

Yeah, good question. So I think clearly AI is changing the whole landscape. And I think the impact will be bigger than the internet. It's more profound as well. So I think AI initially is able to help you do things more efficiently. And then with a lot of agents helping you do things that are mundane but now they can do it faster. So in some ways, you can drive a lot of efficiency. Even in semiconductor design, how much you can drive efficiency in terms of timing — how quickly can you come out? And secondly, the cost. So I think those will help drive that. And then I think a couple of bottlenecks for AI demand and growth: one is, of course, everybody knows power constraints. Some countries just don't have that power; they get impacted. Secondly, a lot of people didn't realize the helium impact can also be very significant for semiconductors. And thirdly, everybody knows right now memory is a bigger shortage, and everybody is scrambling for memory. And even though you want to build a fab to increase capacity, it will take a couple of years to do that. Same thing for CPU, GPU — all this will be highly demanded. And I think pricing will also go up because we have to pass the cost to the customer. So I think those will impact the industry growth. And then overall, I feel that the companies most impacted are those not embracing AI. Because AI can help you drive a lot of efficiency across all different functions of the enterprise. We should embrace and find ways to better use AI for your prediction, for your design, for all different parts of the workload. And I think that's tremendous.

很多人会说，反对 Terafab、反对 Intel Foundry 具有竞争力的简单论点，实际上是一个问题，你知道，建筑内部还有其他因素，对吧？你描述了 IP 和业务运营的速度。然后还有外部因素，你知道，很多已经被讨论过，但其中之一是劳动力成本和实际制造产能。

A number of people would say the simplistic argument against Terafab, against Intel Foundry being competitive, is really a question of, you know, there are other factors internal to the building, right? You described IP and velocity of just how you're doing business. Then there are external factors, and you know, a lot's been talked about a number of them, but one of them is the cost of labor and actually the manufacturing capacity.

在投资晶圆代工业务时，你显然相信有一种版本可以实现本土制造，埃隆也这么认为。你能谈谈这一点吗？以及劳动力限制有多真实？

You know, in investing in the foundry business, you obviously believe there's a version where you can manufacture domestically, and Elon does, too. Can you talk a little bit about that and, you know, how real that constraint is, the labor constraint?

没错。所以，当我决定是否要加倍押注晶圆代工还是退出时，我听到了市场中的各种声音，如你所知，有人说这太贵了，行不通。但我最终认为这对美国非常重要，对行业也非常重要。我们都经历过供应链的挑战，这对任何半导体大公司都至关重要。我们必须考虑供应链，必须拥有稳健且有韧性的供应链，不能只依赖不同地区的一两家厂商。所以，我认为越来越多的人会意识到在美国制造是关键。至于最先进的工艺，比如 14A 是 1.4 纳米，我们已经开始规划 1 纳米和 0.7 纳米。它越来越小，有点像我们的头发，非常细。这非常复杂，不容易做。每一步如果出错，就会前功尽弃。所以，某种程度上，你必须非常精确地进行制造。这越来越成为瓶颈。我们非常尊重台积电，我们是很好的合作伙伴。更重要的是，我们都需要更多产能来服务客户。所以，我认为我们决定长期坚持，这很关键，也是我能为行业创造更多价值的地方。

Right. So, I think, you know, when I decided whether I should double down on foundry or should I get out of the foundry. And I come to the of voices that are saying of voices in the marketplace, as you can tell. It's very expensive. It's not going to work. But I finally decided this is very important for United States and also very important for the industry. And I give you the idea that, you know, this we all live through this challenges of supply chain. And it's very important for any of the big company in semiconductor. And we really have to think about the supply chains. And you have to have a robust and resilient supply chain. You cannot just depend on one or two player in different geographical. And so, I think, you know, the more and more people going to realize making in United States is critical. And then the most advanced process, like, for example, we have the you know, 14A is like 1.4 nanometer. And we're already starting to plan for 1 nanometer and 0.7 nanometer. It's getting smaller and smaller. So, in a way, it's much like our hair. So, thin. So, it's a lot of complexity. It's not that easy to do. And every step, if you make a mistake, then you just go down, you know, go down the drain. So, in some way, you have to be really precise and in that manufacturing. So, in some way, this is a become more and more going to be the bottleneck. So, we felt that we, you know, we have a lot of respect for TSMC. We're a great partner. And then the more important, we both need to have more capacity to serve the customer. And then so I think we decided by the bullet longer term, I think it's critical and then that's where I can create more value for the industry.

长期以来，人们一直在谈论最终会达到一个分辨率极限，无法再进一步小型化。比如线宽变得太小，无法继续。你认为我们什么时候会真正达到这个极限？

People have been talking for a long time about eventually hitting a point of resolution where you can't really miniaturize things further. Like the line width gets too small to be able to keep going. When do you think we actually hit that limit?

好问题。我认为，目前我们有 18A，现在 14A 即将量产。我能看到 10 和 7。所以我认为这条路我们可以走通，但会越来越昂贵、越来越困难。这就是为什么我们需要合作伙伴，不能单打独斗。与材料供应商、设备供应商合作，确保我们能真正提升良率和性能。另一个瓶颈是封装，先进封装。我们都知道台积电的 CoWoS。现在我们有一个非常好的下一代封装叫 EMIB-T。我必须确保它能以满足客户要求的良率进行生产。现在正如你所说，开始看到后劲不足。所以我也在关注一些新材料，回到材料尺寸或化学元素周期表。比如氮化镓、碳化硅和磷化铟。我投资了这三种。看看这些新材料如何推动发展。在封装方面，我开始投资玻璃。玻璃是一种非常好的绝缘体。我甚至投资了一家风险投资公司 3D GS。然后我发现英特尔在模块方面有大约 1000 项专利。所以如何将材料与模块结合起来？我们刚刚宣布了一个与印度政府合作的大项目，在印度和美国新墨西哥州制造。所以我认为先进封装非常重要。我也开始关注人造钻石，那是另一种非常好的绝缘体。所以我投资了钻石晶圆厂。这是下一代需要关注的东西。新材料、新衬底材料以及新的设计方法论来推动它。作为工程师的好处是，你总是遇到障碍，然后找到方法跳过或绕过障碍，取得更好的结果。这是我作为投资者和从 EDA 工具到设计到制造构建半导体行业长期以来的经验。拥有这种经验很好，现在我可以找到方法为行业做出一点贡献。

Good question. So I think I can see, you know, right now we have 18A and then now we do going to production of 14A. I can see 10 and 7. And so I think that path I think we can get there, but going to be more and more expensive and more difficult to do. And that's why we need partners. We cannot just do it ourselves alone. Partner with the sub-script vendor, partner with equipment vendors so that make sure that we can really drive those yield and performance. And then the other part also very become the bottleneck is packaging. The advanced packaging. And so we all know about CoWoS by TSMC. Now we have really good one called EMIB-T that is a really next generation. I have to make sure that you become able to do in a production yield that meet the customer requirement. And now see more starting to run out of steam like you described. So right now I also look at some new materials of become going back to the material size or the chemical table. So gallium nitride and silicon carbide and indium phosphide. So I invest in all three. And then looking at some of these new material, how can we really drive that? And in the term of packaging, I starting to invest into glass. Glass is a very good heat insulator. So we I even invest a venture cycle 3D GS. Then I realized that Intel we have like 1,000 patent on the module. So how do you know, sub-script and the module put it together and then we just announce a big program with Indian government to manufacturing in in India plus in US in New Mexico. So I think this are advanced packaging very important. I also starting to look at artificial diamond. And that's another very good you know insulator. So I also invest into you know diamond foundry. And that's something is a next generation to look at. So new material, new sub straight material and new in the uh design methodology to drive it. So one thing good about being an engineers you always hitting the wall then you find way to either jump over the wall or you work around the wall and then the to get to the better result and that's what I've been uh have been long time as an investor and a building semiconductor from the EDA tool to design to manufacturing. It's kind of nice to have that experience. Now I can help find a way to make a small contribution to the industry.

是的，这非常令人兴奋。我问这个问题的原因之一也是，如你所说，总有一些事情可以绕过去，但也有物理极限，一旦达到 7 埃或任何极限，你就会遇到问题，需要寻找新材料或其他变通方法。

Yeah, and that one is very exciting and one of the reasons I'm asking about it as well is to your point there's always some things that you can run around but there are also physical limits where once you hit seven angstroms or whatever the limitation is you start to run into Yeah, you need to find new materials or find other workarounds.

是的。

Yes.

然后一个有趣的问题是，我们讨论这个已经很久了。我记得 20 年前人们就在谈论我们最终会达到一个空间用尽的地步。你是否会遇到某种渐近线，实际上使不同代工厂之间的性能趋于一致，从而不再有差异？

And then the interesting question is and we've been talking about this for a long time. I remember 20 years ago people are talking about how we'd eventually hit this hit a point where we ran out of space on this um is do you run into some sort of asymptote that actually normalizes performance across different foundries so we're not?

这是个好问题。就摩尔定律而言，它是在翻倍，然后功耗和成本，你可以让性能翻倍，但无法让成本和面积也翻倍。所以这些是你必须让步的地方，除非你找到新的材料或新的设计方法。然后转向材料科学，我开始招聘更多材料科学方面的人才。这是我们领域的创新。我们如何做到？我仍然记得 18 年前，我还在投资半导体，实际上大多数风投公司，其中一些是非常优秀的一流风投，是我的好朋友。最初在合伙人会议上，整个房间的合伙人，在我谈论半导体之后，一半人找借口跑出了房间。然后另一半人说：“有价值，你有软件服务吗？”然后他们也走了，只剩下两个人同情地听我讲。所以，历史已经改变了。现在看看半导体，腾讯是一家市值 5.3 万亿美元的公司，博通和台积电是 2 万亿美元的公司，我的好朋友苏姿丰在 AMD 接近 8000 亿美元，而我接近 6000 亿美元。

That's a good question. In term of like Moore's Law is an uh double in the and then the power and the cost and then you can double the performance but you cannot double down on the cost and and area. So those are the thing you have to give give way unless you find some new way of material new way of design. And then become material science I starting to hire more people in the material science. So that is kind of innovation in our area. How can we do that? And I still remember 18 years ago, and I I still investing in semiconductor, and actually most of the VC firm, some of them are very nice tier one venture firm, my good friend of mine. And I initially the partners meeting, the whole partners in the room, then after I talking about semiconductor, half make excuse run out the room. Then eventually the other half they said, "Valuable, do you have any software service?" So, then they both left with only two sympathetically listen to me. So, it's kind of the history have changed. And now semiconductor, if you look at it, Tencent is a 5.3 trillion market cap company, and then Broadcom and TSMC is 2 trillion market cap company, and Lisa Su, my good friend at AMD, is almost 800 billion, and I'm close to 600 billion.

所以，从某种程度上说，半导体又变热了，而且变得至关重要，因为 15 或 20 年前我投资半导体时，没有风投愿意跟我一起投，只有一些大公司比如三星、ARM、软银等跟我一起投。现在我看到很多风投开始想投资半导体，所以我非常高兴。

So, in some ways semiconductor has become hot again, and it became essential because 15 or 20 years ago, when I invested in semiconductor, no VC wanted to join me except some big corporations like Samsung, ARM, SoftBank, and others investing with me. And now I'm starting to see a lot of VCs wanting to come invest in semiconductors, so I'm very happy.

鉴于人们对这个曾经被认为太难的投资领域产生了巨大兴趣，对吧？

Given the enormous interest in investing in this area that used to be considered too hard, right?

是的。

Yes.

你怎么看？我的意思是，你作为 Walden 的风投已经很长时间了，同时也是一位经营者。我列出一些普遍的担忧：资本非常密集，你应该告诉我我漏了什么。设计能否成功流片非常不可预测，错过流片，而且你需要非常了解工作负载。我认为还有一点是客户切换的风险非常高，对吧？我们一起参与过的公司里，即使赢得了设计，仍然存在扩大订单量的问题。还有周期性，对吧？你建立了硬制造能力，但需求每年都可能变化。你对这个行业的难点以及来自不同领域的长期需求增长有什么看法？人们认识到供应链多元化的重要性，同时 AI 方面需求爆发式增长。你仍然是一位投资者，同时又做出了有史以来最大的赌注，比如去当 CEO。你如何看待这些不同的风险，并建议其他人投资供应链的哪些环节？我知道这个问题很大，但鉴于你的经历，我觉得有很多 YOLO 行为，比如内存短缺就买内存股，以及不愿意承担 10 年周期的事情，比如材料科学。

What do you think? I mean, you've been a venture investor with Walden for a very long time as well as an operator. The general fears I'm just going to list a bunch of them. The general fears have been: it's very capital intensive. And you should tell me what I'm missing. It's very unpredictable in terms of shipping a design that works, missing tape out, and you need to understand the workload very well. I think there's another which is just like it's very high risk for the customer to switch, right? I think we've been involved in companies together where there's a design win and then there's still the question of scaling order volume. And then there's cyclicality. Right? You build hard manufacturing capacity and demand may change or not in any given year. What is your view on how a bunch of what makes it hard as an industry and then the secular demand growth from a bunch of different areas? So you have the recognition of how important a more diverse supply chain is. And then you have this explosive demand growth on the AI side. How do you, you're still an investor and then you're making the biggest bet ever, like go be CEO. How do you think about these different risks and advise others about where to invest in this supply chain? I realize that's a very large question, but just given your history with it, I think there's a lot of YOLO action of like there's a memory shortage, buy memory stocks, as well as just unwillingness to take on things that have a 10-year timeline, like material science.

好。你的问题范围很广。我来试着解释一下。首先，我认为风险投资创业是我的天性。我真的很享受。这不是在吹嘘。所以，有一些好的退出。你知道，我仍然有 159 个 IPO 和并购。这包括半导体。仅就半导体而言，我多年来的投资 238% 是在美国。所以我通常看半导体——

Good. You have quite a broad range of questions. Let me try to explain that. So, first of all, I think venture capital startup is in my blood. And I really enjoy it. And I think this is not trying to brag about it. And so, there's some good exits. You know, I still have 159 IPOs and M&A. And that includes semiconductors. Just break down the semiconductors, I invest over the years 238% is in US. So, what I usually look at in semiconductors—

澄清一下，这太不可思议了。

Just to be clear, that's incredible.

谢谢。谢谢。我只是享受构建的过程。但更重要的是，我会首先从投资角度看待这个问题。我总是看瓶颈在哪里？你要解决什么问题？例如，我投资了一家叫 Credo Semiconductor Group 的公司。互连是否成为瓶颈？所以我决定支持它。我还投资了 Celestial AI，在光学方面。因为速度在集群互连中变得越来越重要，所以我认为光学变得非常重要。看看 Jensen，他投资了几乎所有与光子学相关的公司。另一个我看的方面是，需要什么解决方案？例如，我们谈论设计、复杂性和成本。你能利用 AI 和机器学习来推动更好的设计和更好的解决方案吗？所以有几家新创公司正在进入 EDA 相关领域，以推动性能提升。我认为那是个金矿。另一个方面是新材料。我们谈论磷化铟。这就是我投资 Inphi 的原因，后来 Marvell 收购了它。然后你投资一些新材料，比如氮化镓和碳化硅，有些公司开始被收购，包括一家做电源管理的公司，ADI 刚收购了 Empower。所以又是 IBR。这是一个非常好的领域。电源管理现在成为瓶颈，从 40V 转换到 1V，在转换过程中你会损失很多功率。你如何推动功率改进？所以我认为功率和热管理成为瓶颈。所以我总是看我们要解决的问题是什么？它真实吗？客户迫切需要它吗？然后我开始投资。接下来，从第一天起就非常重要的一点是，你必须瞄准第一个客户。

Thank you. Thank you. It's just enjoy building it. But more importantly, I will look at this first of all on the investment side. I always look at where is the bottleneck? What are you trying to solve? For example, I invested in a company called Credo Semiconductor Group. Is this interconnect becoming the bottleneck? So I decided to back it. And also I backed Celestial AI on the optical side. And then because speed becomes more important in the interconnect in the cluster. So I think optical becomes very important. Look at Jensen. He invested in almost every company that is photonic related. And then the other part I look at is, okay, what are the solutions that need? For example, we talk about design and then the complexity and also the cost. Can you find some using AI and machine learning to drive better design and better solution? So a couple of new startups are actually going into the EDA related area to drive performance improvement. I think it's a gold mine to do that. And then the other part you look at is new materials. And we talk about indium phosphide. That's why I invested in Inphi and then Marvell bought it. And then you invest into some new materials like gallium nitride and silicon carbide, and some of the companies starting to be acquired, including one doing power management that ADI just bought called Empower. So again, it's IBR. That's a very, very good area. Power management becomes a bottleneck now in terms of converting from 40V down to 1V, and in that conversion you lose a lot of power. And how do you drive power improvement? So I think power and thermal become the bottleneck. So I always look at what is the problem we try to solve? Is it real? Is the customer crying for it? And then I start to invest. The next thing is, it's very important from day one you have to target the first customer.

嗯。

Mhm.

通常我喜欢客户是超大规模公司。他们有规模。如果他们喜欢你的产品，他们愿意在未来几年支付数百万美元。即使给一些认股权证也是值得的，因为你有一个大客户，你就可以规模化。所以我总是看一些公式，你怎么做？然后你怎么找到人才？有时找到人才非常重要。这就是为什么我对美国、硅谷、还有奥斯汀非常感兴趣。另一个地方是以色列。那里有很多人才。所以我投资了相当多，相当大比例的投资在以色列。因为他们有非常颠覆性、创新性的企业家。他们工作非常努力。即使在战争时期，他们仍然开电话会议，突然说：“好的，有警报。我得去地下室。”然后名字可能不好，也许我们只用语音。在某种程度上这有点有趣。他们那种坚韧的创业精神我真的很欣赏。所以总的来说，有很多机会。尤其是在 AI 领域。现在，除了巨大的 AI，物理 AI 是下一个大前沿。然后你必须真正看全栈。这就是为什么我仍然参与很多我们非常熟悉的前沿模型。我支持的一些投资是因为我真的很喜欢用于物理 AI 的开源前沿技术。我认为这是个好方法。

And usually I like the customer to be a hyperscaler. They have the scale. If they like what you have, they're willing to pay millions of dollars in the next few years. And even giving some warrants is worth it because you have one big customer, you can scale. So I always look at some formula, how do you do that? And then how do you get the talent? And then, yeah, sometimes it's very important to find the talent. That's why I'm really interested in the US, Silicon Valley, and then some Austin. And then the other part is Israel. A lot of talent. So I back quite a few, quite a significant amount of my investment in Israel. And then because they have very disruptive, innovative entrepreneurs. They work really hard. Even in this wartime, they still have conference calls and suddenly they say, "Okay, there's a warning. I have to go to underground." And then the name may not be good. Maybe we just use voice. In some way it's kind of fun. They kind of resilient entrepreneurship I really enjoy. So I think all in all, there's a lot of opportunity. And especially in AI. And right now, beside the gigantic AI, now you're looking at physical AI is the next big frontier. And then you have to really look at a full stack. That's why I'm still involved with a lot of this frontier model that we are very familiar with. And some of the investment I back because I really like open source frontier technology for physical AI. I think that's a good way.

你提到了用 AI 让芯片设计和测试的某些部分更快、更便宜、更具创造性的机会。

You mentioned the opportunity to make certain parts of the design and test of chips faster, cheaper, more creative with AI.

根据你在 Cadence 的经验，你认为哪个领域最有潜力？有什么已经见效的做法吗？

Given your Cadence experience, where do you think is most fertile? Is there anything you think is already working?

是的，我在 Cadence 工作了将近 15 年，最得意的成就之一就是在路上找到了我的继任者，培养了他，他后来成为了一位非常出色的 CEO。他真正拥抱了 AI，推动 AI 议程以提高效率。Synopsys 的 Sachin 也在尝试这么做，他们从 Nvidia 获得了投资，我记得是 20 亿美元，帮他做了很多事情。他收购了 Ansys 以进入整个系统设计领域。所以总的来说，他们都尽了最大努力，但初创公司也有机会做一些更具颠覆性的事情，最终它们可以上市或被其中一家或西门子收购。所以我认为机会对所有人都存在，取决于创业者的愿景。我的理念是：如果创业者想卖掉公司，那是更快的退出方式，你不用担心季度业绩。有些创业者从一开始就想上市。作为 VC，我们三个人都支持创业者的梦想，并帮助他们实现它。

Yeah, I think for almost 15 years with Cadence, one of my highlights was being able to find my successor on the road, train him, and he became a super great CEO. He's really embracing AI, driving the AI agenda to drive more efficiency. Synopsys's Sachin also tried to do that, and they have investment from Nvidia, 2 billion I think, helping him to do a lot. He acquired Ansys to move into the whole system design. So all in all, they all do the best they can, but there's also opportunity for startups to do something more disruptive, and eventually they can go public or be acquired by either of them or Siemens. So I think there's opportunity for all, depending on the entrepreneur's vision. My philosophy is: if an entrepreneur wants to sell the company, that's a quicker exit, you don't have to worry about quarter-to-quarter earnings. Some entrepreneurs from day one want to go IPO. As a VC, the three of us, we all support the entrepreneur's dream and help them fulfill it.

如果你看看你提到的不同领域，比如未来的产品开发或 AI 对半导体行业的影响，有像 Periodic 这样的公司做材料，你提到的人在 EDA 和设计方面工作，整个链条中还有制造。你认为英特尔或未来的半导体公司 10 年后会因 AI 而发生根本性变化吗？如果是，会怎样变化？

If you look at the different areas you mentioned in terms of future product development or impact of AI on the semiconductor industry, there are companies like Periodic doing materials, your point folks working on the EDA side and design, and throughout the chain there's manufacturing. Do you think that Intel or a future semiconductor company 10 years from now looks radically different from today given AI, and if so, how?

是的，我认为如此。首先，回到 Sara 关于资本密集、不可预测和周期性的问题，你必须把这些因素纳入决策和投资中。我通常喜欢很早进入，组建一个团队——这很有趣。我觉得你也应该这么做。其次，你试着找到合适的投资者作为合作伙伴。这不仅仅是公司的品牌；我通常看重个人。最重要的是找到一个能共度难关和顺境的伙伴。很多人只在顺境中合作愉快，但公司遇到麻烦时他们就走了。我喜欢能一起坚持的伙伴。很多成功的公司都多次濒临破产，但最终起飞了。所以找到愿意这样做的伙伴更重要。然后看看能帮助你的战略投资者，在制造、存储、连接或其他方面增加价值。还要有一些处于成长阶段和对冲基金的朋友。我很喜欢他们，因为他们有不同的视角；他们了解公开市场，可以指导创业者避开陷阱。所以这些都非常有帮助。初创公司的工程就像解决问题。每一步你都要找人帮你解决问题，然后触发下一个前沿。坦白说，我投资的公司中十有八九中途改变了商业计划，因为市场变了。所以我喜欢创业者是一个团队，而不仅仅是单个人。其次，思想开放，愿意倾听并接受我们的指导。最终他们制定自己的计划——不是只做我想做的事。他们找出最佳方案：你给他们足够的反馈，他们得出自己的结论。这正是你喜欢的，接受这一点是正确的决定。这就是做初创公司的乐趣——他们可以快得多。所以回到你的问题：10 年后，什么样的公司会胜出？这只是我的个人观点。那些能够清晰表达并专注于一个利基领域、找到合适合作伙伴、并且能够扩大规模的公司。在某种程度上，我回到了关于全栈的观点。你需要一个全栈解决方案。大公司可以转型为大型平台——比如 Jensen，我很钦佩他。他专注于 CUDA，他想成为一家平台公司，他做到了。像 Anthropic、OpenAI 这样的初创公司找到了更优雅的方式；他们改变了游戏规则，以光速前进，可以成为主导者。希望英特尔能发挥作用，因为我们有 XPU、先进封装和代工厂。如果把这些整合起来，可以为不同的工作负载构建专用芯片。这就是我的方向。

Yeah, I think so. First of all, back to Sara's question about capital intensive, unpredictable, and cyclical, you have to factor that into your decision-making and investment. I usually like to go in very early, put a team together—it's kind of fun. I think you should also do that. Secondly, you try to find the right investor that can co-partner with you. It's not just the brand of the firm; I usually go for the individual. The most important thing is to find a partner through difficult times and good times. A lot of people are enjoyable to work with in good times, but when the company gets into trouble, they just walk away. I like to have a partner that really works through it. A lot of successful companies have been almost bankrupt multiple times but eventually took off. So it's more important to find a partner willing to do that. Then look at strategic investors that can help you in manufacturing, memory, connectivity, or various ways to add value. Also have a couple of friends in the growth stage and in hedge funds. I really enjoy them because they have a different perspective; they know the public market and can guide the entrepreneur where not to go. So those can be very helpful. Engineering for a startup is like problem-solving. Each step of the way you have to find people to help you solve the problem, and then you trigger the next frontier. Frankly, nine out of ten companies I invest in change their business plan halfway because the market changed. So I like an entrepreneur as a team, not just one person. Secondly, open-minded, willing to listen and get coaching from us. Eventually they formulate their own plan—it's not just doing what I want. They figure out the best thing: you give them enough feedback, they draw their own conclusion. That's exactly what you like, and embracing that is the right decision. That's the fun of doing startups—they can move much faster. So back to your question: 10 years from now, what will be the winning company? This is just my personal view. The one that articulates and laser-focuses on one niche area, finds the right partner, and is able to scale the company. In some way, I'm back to my point about full stack. You need to have a full stack solution. Big companies can transform themselves to look at big platforms—like Jensen, I admire him. He focused on CUDA, he wanted to be a platform company, and he did it. Startups like Anthropic, OpenAI find a way to do it more elegantly; they change the game and move fast, at the speed of light, and can become dominant players. Hopefully Intel can play a role because we have XPU, advanced packaging, and foundry. If you put it all together, you can build purpose-built silicon for different workloads. That's the way I'm going.

这很有道理。我想问题的一部分是你的方向，另一部分是：这是否从根本上改变了你的工作方式？因为在软件世界，招聘对象发生了巨大变化——人们管理多个智能体。现在很多人招聘 30、40、50 岁的人，因为他们习惯于管理团队，这可以迁移到管理智能体上，包括理解复杂性、设置和质量保证等。我想知道在物理世界或晶圆厂的背景下，你如何看待团队结构、能力或 AI 如何叠加的变化？是自然的缓慢演变，还是有根本性转变的领域，比如“哦，从材料开始，我们现在应该只用这三个 AI 模型加上一些化学”？所以我很好奇你如何看待那里的未来世界。

That makes a lot of sense. I guess part of the question is where you're going, and the other part is: does it fundamentally change how you work? Because in the software world, there's a big shift in terms of who you hire—people managing multiple agents. Many people now hire those in their 30s, 40s, 50s because they're used to managing teams, and that transfers to managing agents in terms of understanding complexity, setup, QA, etc. I wonder in the context of the physical world or a fab, how you think about shifts in team structure, capabilities, or how AI layers on. Is it a natural slow evolution, or are there areas of radical shift like, "Oh, from materials now we should just use these three AI models plus some chemistry"? So I'm curious how you think about the future world there.

好问题。我想回到爬、走、跑这个阶段。在爬的阶段，你基本上尝试招募半导体行业的一些顶尖人才。现在我开始考虑需要引入哪些软件人才，以便构建一个全栈。

Good question. I think back to that crawl, walk, and run. In the crawl phase, you basically try to recruit some of the best talent in the semiconductor industry. And now I'm starting to look at what software talent I need to bring on board in order to build a full stack.

现在我开始看团队的平均年龄，在 40 多岁、接近 50 岁、50 岁。我需要引入一些新人才。让他们了解工作量和前沿模型开源，这很重要。所以，我知道我儿子现在成了我的老师。每次他邀请我去他家，我们和孙子们一起玩，我就开始向他请教 AI 和机器学习。他比我更懂行。所以我学到了很多，然后尝试理解投资，并引进一些人才。我们正在改变英特尔。过去它是一家非常老旧的、依赖电子表格的公司。现在我正在把它转型为 AI 原生和 AI 赋能的企业，利用我们的一些设计，并让整个组织都拥抱 AI。这样他们就减少了对电子表格和人工的依赖。你要把这两种人才加上我能用的最好的 AI 工具结合起来，不仅用于我的组织，不仅用于我的销售。现在我开始关注的不仅仅是营销，还有设计，也要拥抱 AI。

And now I'm starting to look at my team's average age, which is in the 40s, late 40s, 50s. I need to bring in some new talent. So they understand the workload and the frontier model open source. That is important. So, if I know that my son has become my teacher now. Every time he invites me to his house, we're playing with the grandkids. I start tapping him on all the AI machine learning. He's more plugged in than me. So, I learn a lot and then try to understand investing and bring some of the talent in. So, we are changing Intel. Used to be a very old legacy spreadsheet company. Now I'm transforming it to become AI-native and AI-enabled, using some of our design and also across all the organization embracing AI. And then they become less dependent on the spreadsheet and labor to do that. And you're going to combine the two talents plus the best AI tool that I can use, not only for my organization, not only for my sales. And now I'm starting to look at not just marketing but also the design to embrace that.

我认为很多投资者，至少对我来说，自从我创办公司以来的这几年，思考资本密集型公司的不同资本来源非常有教育意义。我以前做过很多软件。所以你需要有聪明的朋友，他们持有非常不同的立场和资产负债表，如果你说‘我需要 1.5 亿美元才能达到临界规模’，那就不太需要。你在这方面已经经历了很长时间，而且你还有与政府作为大股东合作的独特经验。你认为这种产业政策，它带来了像台积电这样的巨大成功？世界上最重要的公司。但在美国商业文化中，它长期以来一直有点被诟病。你认为现在应该如何改变，或者它在哪些方面是相关的？

I think a lot of investors, at least for me the last few years since I started a firm, it's been very educational thinking about the different capital sources for more capital-intensive companies. I did a lot of software before. And so your need to have smart friends with a very different stance and balance sheet was less if you're like, 'Ah, I need $150 million before this thing gets to some critical mass.' And you've lived that for a very long time and then you have the unique experience of working with the government as a large stakeholder. How do you think this sort of industrial policy, it's led to huge successes like TSMC? The most important company in the world. It's also been a bit frowned upon in American business culture for a long time. How do you think that should change now or where is it relevant?

这是个好问题。我认为，在明显的资本密集型业务和基础设施投资中，你需要获得资本。在某种程度上，我认为对于我们早期的风险投资来说，现在开始变得非常资本密集。一些风投公司愿意投入 10 亿美元到某家公司，这在风投行业是闻所未闻的。现在正在发生。所以，在某种程度上，你只需要，我喜欢这种钟形曲线。要么你非常早期进入，然后因为 A 轮估值已经超过 10 亿美元。所以你必须进入 pre-money、pre-seed 阶段，才能达到 200 亿、300 亿美元的估值，这现在非常罕见。所以你必须这样做，选对标的。另一部分是能够找到资本来规模化。这就是为什么一些共同基金也喜欢进入 pre-market、早期阶段，和我一起投资。我很高兴他们加入，因为他们对是否必须拥有公司 20% 的股份不那么敏感。没有那么多 20% 可以给。所以你必须找到合适的投资者进来。然后对于像 AI 工厂和晶圆厂这样的资本密集型项目，你真的需要利用政府资金或一些主权基金，以及一些非常大的资本。有一些大型基金专门支持基础设施，我们喜欢利用其中一些来确保他们能扩大我们的运营。所以，我认为总体而言，政府主权基金变得非常重要。同时，作为一家上市公司，我有意专注于那些更长期增长导向的投资者，这样他们可以帮助我发展业务，而不是短期询问资本配置，比如你打算在哪里回购股票。这些都是好问题，但与此同时我也必须建设业务。所以，我认为这种平衡很重要。

It's a good question. So, I think in a clearly capital-intensive business and infrastructure play, you need access to capital. And in some way, I think for our early-day venture capital investment, now it's starting to become very capital intensive. And some of the venture firms willing to put $1 billion into some company is very unheard of in the VC business. Now it's happening. And so in some way, you just have to be, you know, I like this kind of bell curve. Either you go in very early and then because it's starting to do the Series A at over $1 billion valuation. And so you have to go in pre-money, pre-seed to go into that kind of a $20, $30 billion valuation, which is very rare right now. So, you just have to do that, pick the right one. And the other part is able to find capital to scale. And that's why some of these mutual funds also like to move into the pre-market, early stage to join me in investing. I delight them because they are less sensitive to whether they have to own 20% of the company. There's not too many 20% to give. So, you have to find the right investor to come in. And then in terms of the capital-intensive like AI factory and foundry, you really need to tap either government funding or some sovereign fund and also some very big capital. There's some big fund that is organized to support the infrastructure, and we like to tap into some of them to make sure they can scale our operation. So, I think overall, government sovereign fund becomes very important. And also as a public company, I purposely want to focus on some investors that are more long-term growth oriented, so that they can help me grow the business rather than short-term asking about capital allocation, like where are you going to buy back your shares. Those are good questions, but meanwhile I also have to build the business. So, I think that balance is important.

你认为目前投资者对英特尔最大的误解是什么？

Do you think there is something that investors most misunderstand about Intel at this moment?

有不少事情。首先，我认为正如我回到这个爬行、行走和奔跑的阶段。过去四个月我在爬行，但人们开始认识到它的潜力。另一部分非常重要。我们需要真正推出最好的产品。在 PC 客户端我们仍然有很大的市场份额，但我们确实需要构建性能更强、表现更好的产品。所以，这就是为什么我悄悄地在组建 CPU 架构师、GPU 架构师和软件架构师团队，以便我们能实现跨越式发展。就像我看英特尔，我希望它拥有多种创业文化，这样我们就能快速行动，利用更好的技术实现跨越。另一部分是除了产品，还有一些新的能量在涌现，比如生成式 AI、物理 AI。有很多领域我们可以投资。市场巨大。这是产品方面。在晶圆代工方面，我们与台积电差距很大。就他们的性能而言，我们必须谦虚，着眼于构建我之前提到的积木：IP、良率、缺陷密度和周期时间，使其更高效、更可靠。这是一个信任生意。人们需要先信任你，才会把晶圆交给你。所以这些是需要更长时间的事情。但我认为到 2030、2032、2031、2032 年，我会开始浮出水面。人们可能不了解我在产品方面有多大的潜力，不仅仅是 PC 客户端。那是我们的主业。我们向上进入边缘，进入物理 AI 和智能体式 AI。因为现在不同了，过去你基本上是为人类提供服务器、提供 PC。现在开始有了另一个不同的维度。有数百万个智能体。它们需要访问算力。它们访问软件栈。所以，我认为那部分我们有机会真正发挥作用。游戏还没有结束。我们可以在智能体式 AI 和物理 AI 上发挥作用。这就是我的方向。而 AI 只是一个开始。你知道，有 Jensen 拥有的训练，还有边缘，以及在智能体式 AI 和物理 AI 方面，我认为这是巨大的机会。每个人都有机会。所以，我认为这是我要去争取的部分。所以，我希望投资者会知道，尽管在 14 个月内，我们为股东带来了 6 倍的回报。这只是一个开始。我们还有很大的空间。

Quite a few things. First of all, I think as I back to this crawl, run and walk. Last four months I crawled, but people are starting to recognize the potential of it. And the other part is very important. We need to really get the best product out. Either PC client we still have a big market share, but we really need to build more performant, better performance. So, that's why I'm quietly building up the CPU architect, GPU architect and software architect so that we can leapfrog. Just like I look at Intel, I want to be a multiple of startup culture so that we move fast and we can leapfrog using better technology. And then the other part is besides the product, there are some new energy coming in like in generative AI, the physical AI. There's a lot of area that we can invest. Market is huge. That's on the product side. And the foundry side, we are very distant from TSMC. And in terms of their performance, so we have to be humble, looking at building the building blocks that I mentioned earlier: the IP, the yield, the defect density, and the cycle time to make it more efficient and more reliable. It's a trust business. People want to trust you before they give you the wafer to count on you. So, those are the things that will take longer time. But I think by 2030, 2032, 2031, 2032, I think I will start to surface up. People may not understand how big potential I can be in terms of product, not on the PC client. That's our bread and butter. And we move up to the edge and move into the physical AI and agentic AI. Because not right now, in the past, you basically provide the server, provide the PC for human. Now it's starting to have another different dimension. It's millions of agents. They need to access the compute. They access into the software stack. So, I think that part we have a chance to really play. The game is not over yet. We can play on the agentic AI and also the physical AI. So, that's kind of where I'm going. And the AI is just a beginning. You know, you have the training that Jensen owns, and the edge, and also in terms of agentic AI with agents and also physical AI, I think it's the jumbo. Everybody has a chance. So, I think that's part that I want to go for it. And so, I think hopefully the investors will know even though in 14 months, we made six times return to the shareholder. This is a beginning. We still have a lot of room to go.

从这里开始还有风险投资的回报。

There's venture returns from here.

是的，你知道，我总是在寻找 10 倍回报。

Yeah, so you know, I always look for 10x.

你知道，我骨子里是个风投人，总想追求 10 倍回报。在 Cadence，我卸任 CEO 时，我们创造了接近 76 倍的回报。从临时 CEO 接手时股价 2.42 美元，到我以执行董事长身份退休时，股东回报大约 85 倍。但在 Intel 很难做到这点，因为基数太大了。所以我就说：“好吧，那我们就做 10 倍。” 如果 5 年、10 年内能实现 10 倍，我认为就是很好的回报。骨子里是个风投人，这就是我的目标。

You know, being a venture at heart, you want to look for 10x. At Cadence, when I stepped out as CEO, I think we made close to 76 times. Starting from interim CEO, $2.42. And then when I retired as executive chairman, about 85 times return to the shareholder. So it's hard to do that at Intel because the base is bigger. So I kind of said, "Okay, let's do it at 10x." And then in 5 years, 10 years, if we can do 10x, I think it's a good return. Being a venture capitalist at heart, that's kind of my goal.

所以在这个已经很大的基础上，你是在为一项非常宏大的使命加速。你描述的工作负载方向里蕴含着一种信念。好吧，我想有些人会说：“我们只会建造越来越大的数据中心，1 吉瓦只是起点……” 但集中化以及集中运行推理计算所带来的效率，是主导方式，而不是考虑边缘、考虑客户端。你认为计算存在一个你相信的均衡状态吗？还是说我们只能由工作负载来决定？你怎么看？

So there's a Godspeed on this very, very large mission from this huge base already. There's an embedded belief in what you described about where the workload is. All right, where I think some would say like, "We're just going to build bigger and bigger data centers and a gigawatt is the beginning and then..." But the centralization and the efficiency from running even the inference compute in a centralized way is the dominant way versus thinking about the edge, thinking about the client. Do you think that there's like an equilibrium state that you believe in of where the compute is, or is it just we will find out from the workload? How do you think about that?

嗯，我觉得这是个很好的问题。现在 AI 领域正在大规模建设，我认为这是正确的做法。我看不到任何会放缓的因素，因为工作负载增长非常快。然后我觉得问题在于如何……

Yeah, I think that's a very good question. Right now there's a massive build-up in terms of AI. I think it's the right thing to do. I don't see anything to slow it down because the workload is increasing a lot. And then I think the question mark is how...

而且我们还受供应限制。

And we are supply constrained.

受供应限制。是的。所以我认为任何放缓都是供应限制造成的。但另一方面，我一直在观察所有这些基础设施的建设。最终你必须看解决方案是什么，你要驱动什么应用？我更关注应用。如果你能找到一个巨大的应用，或者把几个应用加起来变得有意义，然后专注于它，那么并不是所有建设者都会赢。

Supply constrained. Yeah. So I think anything slow down is a supply constraint. But I think the other part is I always look at all this infrastructure build out. At the end you have to look at what is the solution, what is the application you want to drive? And I'm more focused on application. So if you can identify the application that is humongous or add up a few applications to become meaningful and you focus on that, it's not everybody built is going to be winning.

嗯。

Mhm.

所以有些会大获成功，有些会逐渐失败，或者停滞不前。就像互联网，你可以看到有些变得非常大，比如亚马逊、Netflix，而有些则停滞不前、消失或被收购。所以对我来说，方法是一样的。真正关注他们试图服务什么应用，那个应用有多大，是否可持续，或者是否过于拥挤。如果太拥挤了，可能只有一两个能存活，其他的可能会被整合。所以我认为这个行业会经历高速增长，然后开始整合，最终可能只有一两个成为真正的赢家。我们以前看过这种电影，所以对我来说并不意外，但重点是应用，比如 Netflix 就是应用，

And so some are going to win big time and some are going to lose over time or some go sideways. Just like the internet, you can see some of them turn out to be very big like Amazon, like Netflix, and then some of them kind of go sideways and disappeared or being acquired. So I think to me it's the same approach. Then really focus on what application they try to serve and how big that application is and whether it's sustainable or not or is very crowded. So if it's too crowded, maybe one or two may survive, the others maybe just consolidate. So I think this industry goes through that big growth and then starts to consolidate, maybe eventually one or two become the real winner. So I think we've watched the movie before, so it's not a surprise to me, but focus on application like Netflix is application,

嗯。

Mhm.

你知道，亚马逊是一个真正的应用，在我看来他们赢了。

you know, Amazon is a real application that to me they're winning.

但你假设其中一些应用更适合由客户端或边缘计算来服务，而不是仅由数据中心来服务。

But you're assuming that some of these applications will be better served by client or edge compute than only by the data center.

好的。

Okay.

正是。

Exactly.

是的。我的意思是，作为多家公司的投资者，这些公司做机器人、做国防，所以设备上的计算是一个非常重要的选择，关系到我们围绕它所做的假设。比如，如果最终家里有一个机器人，你对家庭内部和周围连接的假设决定了你能做什么。

Yeah. I mean I will say as I'm an investor in a number of companies that are doing robotics, they're doing defense, and so the compute on the device is a very important choice in terms of what we assume around it. Like let's say if a robot in the home eventually, what you assume is in the home and in connectivity around it determines what you're able to do.

嗯，我认为这在 SaaS 时代有点被遗忘了。

Um and I think that that's been kind of it was kind of forgotten for a little bit in the SaaS era.

嗯。是的。是的。所以我认为我的投资理念更多是：找到一个真正需要解决的问题。其次，谁是可以合作的参与者？然后第三，看应用。应用有多大？是否可持续？如果它真的很大，你相信它，就加倍、三倍下注。

Mhm. Yes. Yes. So I think I have more my investment thesis is find a problem that is really need to solve. And secondly, who are the player that you can partner with? And then thirdly, look at the application. How big is the application? Is that sustainable? And if it's really big, you believe in it, double, triple down.

是的。但你包括押注那些尚未广泛部署的应用。

Yeah. But you're including betting on applications that have not yet been broadly deployed.

好的。太棒了。非常感谢你今天加入我们。很愉快。

Okay. It's amazing. Well, thank you so much for joining us today. It was a pleasure.

非常感谢。

Thank you so much.

谢谢，Lip-Bu。

Thanks, Lip-Bu.

谢谢。

Thank you.

在 Twitter 上关注我们 @No Priors Pod。如果你想看我们的脸，请订阅我们的 YouTube 频道。在 Apple Podcasts、Spotify 或任何你收听的地方关注节目，这样你每周都能收到新剧集。注册邮件或在 no-priors.com 上查找每集的文字稿。

Find us on Twitter at No Priors Pod. Subscribe to our YouTube channel if you want to see our faces. Follow the show on Apple Podcasts, Spotify, or wherever you listen. That way you get a new episode every week. And sign up for emails or find transcripts for every episode at no-priors.com.