Picks & Shovels for an Insatiable Demand

"Well, what does it say about the expectations that investors have placed on companies like yours — that you can release results reaffirming a forecast of a hundred billion dollars in certain types of revenue, surpass estimates in a lot of different ways, and still there be this sense of disappointment?"

「那么,投资者对你们这类公司的期待,到底说明了什么?——你们发财报、重申某类业务上千亿美元的营收指引、在好几个维度都超出预期,结果市场居然还是有一种'不及预期'的失落感。」

"What does it say about where we are in the hype cycle around AI?"

「这说明在这轮围绕 AI 的炒作周期里,我们到底走到哪一步了?」

"I guess we are kind of, like AI, in a very surreal environment. Frankly, I don't think about it."

「我想我们就跟 AI 本身一样,处在一个非常超现实的环境里。老实说,我不去想这个。」

"No, it's hard not to."

「可这很难不去想吧。」

"Right. But no — just focus on fundamentals, great value, and stop thinking about your stock price."

「是啊。但不行——就盯住基本面、盯住实打实的价值,别去想你的股价。」

"The trouble is, it's very hard to do that. But we try to do it."

「难就难在,这真的很难做到。但我们尽量这么做。」

"Well, as we were just talking with Ali — every month or so, every couple of months, there's a new LLM, and it sort of changes the conversation around who's ahead, who's behind."

「就像我们刚才和 Ali 聊到的——差不多每个月、每隔一两个月就冒出一个新的 LLM,而它多少会改变'谁领先、谁落后'的整个话题。」

"I'm curious about — you have this massive partnership with Google, very important, working with them on their TPUs. And I'd love to drill down a little bit with you on that, in particular how concerned you might be about Google taking more of the chip development and design process in-house."

「我很好奇——你们和 Google 有一桩体量巨大的合作,非常重要,一起做他们的 TPU。我想跟你往深里挖一挖,尤其是:你到底有多担心 Google 把芯片的研发和设计流程更多地收回自己手里?」

"The term 'customer-owned tooling' — is that something that keeps you up at night? And are there any provisions you've built into your partnership with them that might preclude that?"

「'客户自有工具'(customer-owned tooling)这个说法——它会不会让你夜里睡不着?你们在和他们的合作里,有没有设下什么条款来防着这一手?」

"Well, that's something we should expect in technology. We've been in semiconductors for over 20 years — in fact, that's how long I've been doing this with Broadcom."

「这是在科技行业里我们本来就该预料到的事。我们做半导体做了 20 多年——事实上,我在博通干这行也正好这么久。」

"It's what we do every year, all the time. We have multiple product divisions, each of which we are the best in what we do. And we just keep investing. Essentially, you try to out-engineer any competition."

「这就是我们年复一年、时时刻刻在做的事。我们有多个产品事业部,每一个我们都是各自领域里做得最好的。我们就是不停地投入。说到底,你要靠工程能力把任何对手都比下去。」

"Customer-owned tooling, that Google is trying to create — trying to design it themselves, with the help of perhaps some smaller partners—"

「Google 想搞的那个'客户自有工具'——就是想自己来设计,可能再找一些更小的伙伴帮忙——」

"Marvell, maybe?"

「比如 Marvell?」

"MediaTek, like that. I call them kind of like ankle biters. But they're basically working for Google to try to achieve customer-owned tooling."

「MediaTek 之类的。我管他们叫'咬脚后跟的小角色'。但说白了,他们就是替 Google 打工,想帮 Google 把'客户自有工具'这件事做成。」

"So we just compete against my own customer in a different part of it. And the whole idea is to be able to create differentiated products and technology that beat what they have."

「所以我们不过是在这盘棋的另一块地方,跟自己的客户竞争。整件事的核心就是:拿出差异化的产品和技术,压过他们手上的东西。」

"And all this is very nice, but it's been helped by the fact that the real competitor facing all this is the GPU out of Nvidia."

「这一切说起来都挺好,但真正帮了我们大忙的一点是:摆在这一切面前的真正对手,是 Nvidia 的 GPU。」

"And as long as Nvidia keeps coming up, generation after generation, with superb technology, Google has to create very equivalent technology to match that. And that's where we come up."

「只要 Nvidia 一代接一代地拿出顶级的技术,Google 就必须做出对等的技术去匹配它。而这,正是我们上场的地方。」

"The TPU process is quite involved. I'm wondering if there's anything you can do to help Google democratize that process a little bit."

「TPU 那一套流程相当复杂。我在想,你们有没有办法帮 Google 把这个流程稍微'平民化'一点?」

"How do you mean, democratize?"

「你说的'平民化'是什么意思?」

"Just in the sense of making TPUs more accessible to mainstream developers, and not just the elite AI labs."

「就是让 TPU 对主流开发者更触手可及,而不只是给那些顶尖 AI 实验室用。」

"Well, I think Google is trying to do that — by making that TPU environment available through cloud, Google Cloud, GCP."

「我觉得 Google 正在这么做——通过云把 TPU 环境开放出来,也就是 Google Cloud、GCP。」

"But a big part of that is the software: the operating system, the compilers that go with it. And that's very much in Google's purview."

「但这里头很大一块是软件:操作系统,还有配套的编译器。而这些基本都在 Google 自己的地盘里。」

"Of course, we'd like to help, but I think they've got that pretty much under control in terms of pushing that program."

「我们当然愿意帮忙,但在推进这个项目这件事上,我觉得他们基本已经掌控得挺好了。」

"I want to talk a little bit about Anthropic, and the ways that these foundation model builders are obviously in great need of capital in order to finance their needs for chips and the compute needed to build and develop."

「我想聊聊 Anthropic,以及这些基础模型公司的处境——他们显然极度需要资本,来为自己在芯片和算力上的需求埋单,以便去搭建和研发。」

"Can you talk a little bit about the way that relationship is being structured and financed, and the provisions that you're having to take to protect yourself against downside?"

「你能不能讲讲,这段关系是怎么设计、怎么融资的?你们又不得不做了哪些安排,来保护自己、防范下行风险?」

"How concerned are you about the end-user demand materializing and enabling them to meet their commitments?"

「你有多担心:终端用户的需求最终能不能兑现,好让他们有能力履行对你们的承诺?」

"I, frankly — [clears throat] — I think in this particular case, we were partners with Google to provide compute capacity through TPUs, which we co-designed with Google, to Anthropic."

「我,坦白说——(清嗓子)——我觉得在这个具体的案子里,我们是和 Google 做伙伴,一起通过 TPU 向 Anthropic 提供算力,而那些 TPU 是我们和 Google 共同设计的。」

"And frankly, there's no shortcut about it. When we first started, almost a year ago now, it was basically a leap of faith — that Anthropic, and generative AI, and Anthropic through its business model of addressing enterprises with code assistance, coding tools, would make a difference."

「而且老实讲,这事没有捷径。我们最初动手,差不多是一年前了,那基本上就是一次'信念的一跃'——赌 Anthropic、赌生成式 AI,赌 Anthropic 那套'用代码助手、编程工具去服务企业'的商业模式,能真正带来改变。」

"We went in on that leap of faith. And I guess today, that was a great bet."

「我们就凭着这份信念一跃跳了进去。现在回头看,我想那是一次绝佳的下注。」

"They do seem to be very ascendant at the moment, given the fact that they have leapfrogged ahead of OpenAI on their valuation, filing confidentially for an IPO."

「他们眼下确实势头很猛——毕竟在估值上已经反超了 OpenAI,还悄悄递交了 IPO 的保密申请。」

"So have you not had conversations with investors who might be worried about you getting left holding the bag if things don't materialize the way Anthropic anticipates them?"

「那你有没有和投资者聊过——他们会不会担心:万一事情没像 Anthropic 预期的那样兑现,最后砸在你们手里?」

"Well, as I said, we were making a bet on Anthropic — us and Google, too, jointly doing that. And I guess we're not just making a bet on Anthropic and its business model."

「就像我说的,我们是在押注 Anthropic——我们和 Google 一起下的注。而且我想,我们押的不只是 Anthropic 和它的商业模式。」

"We're also making a bet that generative AI would be something that would take off — take off in a fairly substantial way, particularly among enterprises, the way Anthropic goes. And it sure did. And that's pretty good reason for it."

「我们同时也在赌:生成式 AI 会真正起飞——而且是相当大规模地起飞,尤其是在企业市场,按 Anthropic 走的那条路。结果它确实起飞了。这也算是个挺好的理由。」

"When it started a year ago, I probably knew a fraction of what I know today. And what I know today, I'll probably know a fraction of what I'll know in 6 months' time. These things are changing very fast."

「一年前刚起步时,我懂的大概只是今天我所懂的一个零头。而今天我懂的,恐怕又只是六个月后我会懂的一个零头。这些东西变得太快了。」

"But the tools we see coming out of Anthropic — and the same applies to OpenAI and Gemini and all of Google — they are just phenomenal in a sense of what they can achieve to help productivity."

「但我们看到从 Anthropic 出来的这些工具——OpenAI、Gemini、以及整个 Google 也一样——它们在提升生产力这件事上能做到的,简直是惊人的。」

"In Broadcom, we use it. We use those tools for engineering design, code assistance. And that's pretty remarkable — what can be achieved in terms of enhancing productivity, more than enhancing productivity — I think even making product designs that are probably better than you do with our engineers, and these are very good engineers that we have."

「在博通,我们就在用。我们拿这些工具来做工程设计、代码辅助。这相当了不起——它在提升生产力上能做到的程度,甚至不止是提升生产力——我觉得它做出来的产品设计,可能比我们工程师自己做的还要好,而我们的工程师是非常出色的。」

"Talk a little bit more about how you're advising your staff to use these different tools, and have you found yourself having to put limits on that?"

「再多讲讲:你是怎么建议员工使用这些不同工具的?你有没有发现,自己不得不给它设一些限制?」

"We hear about the era of 'token maxing' — which I hate that term — but the idea of letting people just use tokens without limits. Have you had to put limits on that, and what does that look like?"

「我们常听到所谓'token 拉满'(token maxing)的时代——这个词我特别讨厌——意思就是让大家不设上限地随便烧 token。你有没有不得不对此设限?具体又是什么样子?」

"Frankly, on a bigger picture, we aren't at that stage yet. So maybe it's because we're behind other guys who got to that point and start to throttle what they do."

「老实说,从大局看,我们还没到那个阶段。所以也许恰恰是因为我们落后于那些已经走到那一步、开始给用量踩刹车的家伙。」

"Now, a big part of it is: when you use a tool like Opus 4.7, and the engineers start using it, they go through a learning process — even how to use it. You can't get a tool to be very productive at the start, but as you keep using it, and you use it long enough, you get very good at it. And then the tools become much more productive."

「而这里头很大一块在于:当你用像 Opus 4.7 这样的工具,工程师们一开始上手,是要经历一个学习过程的——连'怎么用'都要学。一件工具在最初是没法很高产的,但你一直用、用得够久,你就会变得很在行。到那时,工具的产出就会高得多。」

"And as it gets much more productive — just think about it — you can get one great, very senior engineer to produce an application design in one week that would otherwise take 10 engineers, whom we pay $300,000 a year, three months to produce the same thing."

「而一旦它变得高产得多——你想想看——你可以让一个顶尖的、非常资深的工程师,在一周内做出一个应用设计,而这在过去要 10 个工程师、每人年薪 30 万美元、花三个月才能做出同样的东西。」

"The return on investment is still pretty compelling. So to say we're throttling is probably a question of return on investment and what application you're using. But it's also being able to be trained to use it better and better."

「这笔投资回报仍然相当有说服力。所以说我们要不要限流,归根结底大概是个 ROI 的问题,以及你到底在用它跑什么应用。但这同时也是一个'能不能被训练得越用越好'的问题。」

"Last year you announced a landmark agreement to work on AI chips with OpenAI. Could you give us an update on where things stand in bringing that to fruition — in particular addressing what another publication wrote about maybe some snags there, and your wanting Microsoft to agree to buy a certain percentage of the chips? How are things going with finalizing that agreement?"

「去年你们宣布了一项里程碑式的协议,要和 OpenAI 一起做 AI 芯片。能不能给我们讲讲进展,离落地还有多远?尤其是回应一下另一家媒体写到的、可能存在的一些卡点——比如你们希望微软同意按一定比例来采购这些芯片。这份协议敲定得怎么样了?」

"Oh no. By the way, creating chips with partners — best example, as you said earlier, my great relationship with Google. While they try to do custom-owned tooling, it's very, very friendly in a way. We all have the same goals — competition. We have a great and deep partnership with these guys, as we do with most of our… partners, I call it."

「哦,没有的事。顺便说一句,和伙伴一起造芯片——最好的例子,就像你刚才说的,是我和 Google 的良好关系。他们一边想搞'客户自有工具',这中间其实又非常、非常友好。我们的目标是一致的——竞争。我们和这些人有着深厚而牢固的合作,跟我们大多数……我称之为'伙伴'的客户一样。」

"But it's a journey for them. So the first generation of chip would tend to be replaced by a second generation, which gets better and better and better. So for them, it's a journey towards creating custom AI accelerator silicon — to optimize on the workload, to replace what is the de facto now, which tends to be Nvidia."

「但对他们来说这是一段旅程。第一代芯片往往会被第二代取代,然后一代比一代更好、更好、更好。所以对他们而言,这是一条通往'打造定制 AI 加速芯片'的路——为特定的工作负载做优化,去替换掉如今那个事实标准,而那通常就是 Nvidia。」

"We do that indeed, but we only have a few customers — to be exact, six. Each of them on a journey at a different stage towards creating their own silicon which they will use. And obviously, Google is way ahead, being much earlier."

「我们确实在做这件事,但我们只有寥寥几个客户——精确地说,六个。每一个都走在这条'造自己要用的芯片'的路上,只是处在不同阶段。而很明显,Google 遥遥领先,因为它起步早得多。」

"OpenAI is one of them we've engaged with, and we've now engaged three over 2 years. It's pretty remarkable how much they've achieved, because we have the AI accelerator silicon working very well in the labs, in the data centers, and we are on track to go into production late this year. And that's it — off we go to the races."

「OpenAI 就是我们已经在合作的其中一个,两年多来我们已经和其中三家展开了合作。他们取得的进展相当惊人,因为我们的 AI 加速芯片在实验室里、在数据中心里都跑得非常好,而且我们正按计划在今年年底投入量产。就是这样——我们要开跑了。」

"So there's nothing holding back that partnership at this stage — whether it's needing Microsoft to make any kinds of agreements or anything like that?"

「所以现阶段没有任何东西在拖这桩合作的后腿?——比如需要微软先达成什么协议之类的?」

"No, not at all."

「没有,完全没有。」

"Great. I want to talk about M&A, because historically Broadcom has used M&A in a relatively substantial way. Does the AI revolution sort of temper the need for you to supplement growth using mergers and acquisitions? Are you entering a post-M&A phase, or is that the wrong way to think about it?"

「很好。我想聊聊并购,因为历史上博通相当倚重并购。这场 AI 革命,是不是多少削弱了你们靠并购来补充增长的必要性?你们是不是正在进入一个'后并购'阶段——还是说这么想本身就不对?」

"No, you hit it right on, Tom. It's an interesting question. It's a very valid question, because it crossed my mind and that of my board's on an ongoing basis — not regularly, as we do M&A. So I'll be very flip and give you a straight answer."

「不,你一下就说中了,Tom。这是个有意思的问题,也是个非常站得住脚的问题——因为它时不时就会在我脑子里、在我董事会脑子里冒出来,不像我们做并购那样是例行公事。那我就干脆点,给你个直白的答案。」

"In the last 2 years, between '24 and '26 this year, I would be doubling my revenues. I would create over $50 billion per year annualized in revenue. I'm looking around — what can I buy that even comes close to that? And that's the tricky part."

「过去这两年,从 '24 到今年 '26,我的营收会翻一番。我一年会新增 500 多亿美元的年化营收。我环顾四周——我能买到什么、哪怕勉强接近这个数?这才是棘手的地方。」

"I mean, it's a distraction. All M&As are a distraction — to acquire, go to regulators; further distraction to integrate — another year."

「我是说,它是一种分心。所有并购都是分心的事——去收购、去过监管;然后整合起来又是一重分心——又搭进去一年。」

"Meanwhile, organically, this phenomenon we call generative AI and our ability to ship them picks and shovels — the compute capacity into this demand, which is almost insatiable — makes it very hard to choose M&A over focusing and succeeding in generative AI compute."

「与此同时,靠内生增长,这个我们称之为生成式 AI 的现象,加上我们向他们供应'镐头和铲子'的能力——把算力送进这股几乎永不满足的需求里——让我很难在'并购'和'专注并做好生成式 AI 算力'之间,选择前者。」

"Is there any area where you think you might want to make an exception to this — whether it's maybe photonics, just to name one area?"

「有没有哪个领域,你觉得可能会破一次例?——比如说光子学(photonics),随口举一个?」

"Photonics? Or you mean optics?"

「光子学?还是你是指光学(optics)?」

"I'm just throwing it out."

「我就是随口一抛。」

"Anything. Yeah. I've run a business model for 20 years. I've tried very hard, I know, to avoid bright shiny objects."

「什么都行。是啊。我这套生意做了 20 年。我一直很努力,我清楚——努力去躲开那些'闪闪发光的新玩具'。」

"I want to ask you some of the same questions that I've asked some of the other keynote speakers about — when you look at the price sensitivity around token usage, and the price sensitivity around when businesses are prioritizing spending on this, we've seen a lot of companies start to cut back on / throttle use of tokens, etc."

「我想问你几个我也问过其他主题演讲嘉宾的问题——当你审视围绕 token 用量的价格敏感度,以及企业在给这块支出排优先级时的价格敏感度,我们已经看到不少公司开始削减、限制 token 的用量,等等。」

"As we talked about earlier, we've also seen indications that certain Chinese large language models are starting to gain a lot of traction, particularly given how efficiently they can run — whether it's DeepSeek, the list goes on, Alibaba's Qwen. Are you seeing that at all from your vantage point?"

「就像我们前面聊到的,我们也看到一些迹象:某些中国的大语言模型开始获得相当的增长势头,尤其是它们跑起来非常高效——无论是 DeepSeek,还是往下数的一长串,阿里巴巴的 Qwen。从你的视角看,你有没有观察到这一点?」

"From the one I've heard — I can't say we're doing it here in Broadcom, but one I've heard a lot of, and I assume you hear from the same source: a lot of CIOs of companies talking about, 'Hey, we're trying to find ways to reduce token cost and all that.'"

「就我听说的——我不能说我们博通这边在这么干,但有一件事我听得很多,我猜你也是从同一个渠道听来的:很多公司的 CIO 在讲,'嘿,我们正想办法把 token 成本降下来,诸如此类。'」

"And maybe they are ahead of the game where we are, but my sense of enterprise usage of AI — we're still at the early innings of this game."

「也许他们比我们所处的位置更领先一步,但就我对企业使用 AI 的感觉——我们还处在这盘棋的很早期的局。」

"I think we're still trying to learn what we can get out of this phenomenal tool that we have created and continue to enhance and improve with each succeeding generation."

「我觉得我们还在摸索:从这件我们亲手造出来、并且每一代都在不断增强、不断改进的了不起的工具里,我们究竟能榨出多少东西。」

"From my personal viewpoint, as I said earlier, and from the experience running Broadcom — yeah, I don't want my guys to go nuts overspending on tokens. On the other side, if you can create value out of it, and you have a return on investment that you otherwise could not, why stop?"

「从我个人的角度,像我前面说的,也从经营博通的经验看——是的,我不想我的人发了疯似的在 token 上乱花钱。但反过来,如果你能从中创造价值,拿到一个本来根本不可能有的投资回报,那凭什么停?」

"And this is not going to a video arcade when we were kids and play a game, and at the point when the monster is about to eat you, you're going to run out of ammunition, so you feed more tokens. It's almost there, but it's not quite."

「这可不是我们小时候去游戏厅打游戏——眼看怪物就要把你吃掉,你的弹药也快打光了,于是你再投几个币('喂'更多 token)。有点像那个意思,但还不完全是。」

"I'd be remiss if we didn't talk about the other businesses that Broadcom is known for — networking in particular. Your gear has become increasingly important to the modern AI era. We're starting to see other companies like Cisco wanting to play in the areas that you have dominated, and I'm curious what you think of the prospects for them to create competition for you there."

「要是不聊聊博通另一块出名的业务——尤其是网络(networking)——那就是我的失职了。你们的设备在当今的 AI 时代变得越来越重要。我们开始看到像 Cisco 这样的公司,想挤进你们一直主导的地盘。我很好奇,你怎么看他们在那里对你们形成竞争的前景?」

"Well, we're very fortunate. I use it truthfully, not tongue in cheek. We have been in semiconductors for 20 years, and we have accumulated a portfolio of about 17 semiconductor product divisions in specific core areas where we are literally number one, or we're not in."

「这个嘛,我们非常幸运。我这话是当真说的,不是挖苦。我们做半导体做了 20 年,攒下了大约 17 个半导体产品事业部,分布在一些特定的核心领域里——在这些领域,我们要么就是名副其实的第一,要么干脆就不做。」

"We're number one, and we continue to invest. And about five, six of those divisions happen to be very important in creating AI chips and AI clusters. So all became that way, and one of them is networking, as you say — for switching gear — and that's taking off like a rocket."

「我们是第一,而且我们持续投入。这些事业部里,大概有五六个恰好在打造 AI 芯片和 AI 集群上极其重要。就这么一路走成了今天的样子,其中一个就是网络,正如你说的——交换设备——而这一块正像火箭一样起飞。」

"But we also try to be, as we always have been when we sell our products, very — I like to call it strategic. When we look at a customer: are they using it for a particular purpose? Are they sustainable? Do they want not just one generation, but the next and the next and the next — sustainable?"

「但我们也一直努力做到——就像我们卖产品时一贯的那样——非常……我喜欢用'战略性'这个词。我们看一个客户时会问:他们是不是用于某个特定的目的?他们可持续吗?他们要的是不是不止一代,而是下一代、再下一代、再下一代——能一直走下去?」

"And so we prioritize our products to those kind of strategic customers. And demand, as I said, for AI and AI networking today is just massive. So it's okay for us that, with such strength and the way we look at businesses, the overflow — I'm happy for it to go to Cisco."

「所以我们会把产品优先给这类战略客户。而正如我说的,如今对 AI、对 AI 网络的需求,大得惊人。所以对我们来说,凭着这样的实力和我们看待生意的方式,那些溢出来、我们吃不下的部分——我很乐意让它流向 Cisco。」