--- author: Bloomberg Podcasts date: '2026-11-14' guest: '' layout: post.njk source: https://www.youtube.com/watch?v=DF2i8Z4na4Y speaker: Bloomberg Podcasts tags: - key-man-risk - semiconductor-industry - export-controls - cloud-computing title: 后黄仁勋时代的英伟达：华尔街分析师如何评估关键人物风险与中国芯片政策 summary: 彭博社 Odd Lots 节目探讨了黄仁勋因丑闻离职对英伟达的影响。伯恩斯坦资深分析师 Stacy Rasgon 阐述了将英伟达评级下调至中性的原因，分析了黄仁勋离任带来的关键人物风险、长期技术路线决策不确定性，以及美国对华芯片政策和云端远程算力管控收紧对行业构成的潜在冲击。 insight: '' draft: true series: '' category: tech-trends area: tech-engineering project: [] people: - Jensen Huang - Stacy Rasgon - Joe Weisenthal - Tracy Alloway - Colette Kress - Scott Bessent companies_orgs: - Nvidia - Bernstein Research - Bloomberg - Broadcom - AMD - Anthropic products_models: - CUDA - Reuben - Hopper - Blackwell media_books: - Odd Lots status: evergreen --- ### 录制背景与英伟达的关键人物风险 **>>**: 彭博音频工作室出品，播客、广播、新闻。 [音乐] **>>**: 大家好，欢迎收听新一期的 Odd Lots 播客。[音乐] 我是 Joe Weisenthal。
Original English **>>**: Bloomberg Audio Studios podcasts, radio, news. [music] Hello and welcome to another episode of the OddLots podcast. [music] I'm Joe Weisenthal
**>>**: 我是 Tracy Alloway。
Original English **>>**: and I'm Tracy Aloway.
**>>**: Tracy，我们在 11 月 13 日星期五进行录制。英伟达事态的发展速度极其迅猛，以至于我们录制节目时必须格外小心，因为我们不知道在节目播出前还会发生什么。但我们至少知道一件事，那就是英伟达定于下周公布财报。至少这是我们目前所知的下一个可以期待的“已知的未知数”。没错，我们唯一能确定的就是日程表上确实有这么一件事。
Original English **>>**: Tracy, we're recording this Friday, November 13th. The Nvidia story is so fastm moving that we have to sort of be careful with the episodes we record because we don't know what's going to happen but by the time it comes out. But one thing we at least do know is that Nvidia is scheduled to report earnings next week. And at least that is the next sort of thing that we know we it's the it's the it's the one known unknown so to speak that we can look forward to. Right. It's the one thing we do know is that there's something on the calendar.
**>>**: 没错。我觉得这又是我们必须准确说明录制时间的一期节目。现在是纽约时间 12 点 10 分。在我们录制的同时，股价下跌了大约 4%。显然，在企业分析领域，我们经常听到人们谈论关键人物风险，虽然这很难具体衡量和深入思考，但我认为在黄仁勋（Jensen）的案例中，这可以说是最极致的关键人物风险，对吧？正是这个人创立了这家公司。我们从各种报道中获悉，他与公司本身的业务深度绑定。大约有 60 位向他直接汇报的高管，几乎没有继任计划。他似乎还深度参与了具体的融资交易等事务。所以可以想象，如果你是一名现在跟踪英伟达股票的分析师，那肯定非常具有挑战性。
Original English **>>**: That's right. I feel like this is another one of those episodes where we have to say exactly when we're recording. So, this is 12:10 New York time. Um, the stock is down about 4% as we're recording. And obviously, you know, in in corporate analyst space, we hear people talk about key man risk all the time, which is a hard thing to actually measure and think about, but I think in the case of Jensen, this is kind of the ultimate keyman risk, right? This is the guy that built the company. Uh we know from various reports that he was intensely involved in the company itself. Something like 60 direct reports, no succession plan really. It it seems like um he was very involved in individual like financing deals and things like that. And so you can imagine if you're an analyst covering Nvidia stock right now, it must be pretty challenging.
**>>**: 确实，完全如此。在我的生活中，我遇到过一些谈起自己公司的人。他们会说：“我的公司拿到了英伟达的投资，我还见到了黄仁勋。”我当时觉得：“噢，那挺酷的。”但现在回过头去听、回想起这些故事，就会觉得：“哇，他的参与度太深了。”同样一个故事，一方面有点像那个网络梗：“噢，太棒了，我们有一位深度参与业务的首席执行官。”然后又变成：“噢，我们有一位超级事无巨细的[笑声]首席执行官，这到底算好事还是坏事？”总之，如果仅从股票的角度来思考，首先存在关键人物风险；其次是公司将面临的所有法律问题，以及关于极速（Mega Speed）的一切内情，公司在什么时候知晓了什么等等；另外还有业务的基本面，不仅涉及芯片需求，还关乎英伟达在资助庞大生态系统中所扮演的角色，这个生态系统中的企业要么使用、要么采购英伟达芯片。
Original English **>>**: No, totally. I've met a few people in my life who have talked about like their companies. They're like, "My company got some investment from Nvidia. I got to meet Jensen Wong. I was like, "Oh, that's cool." But now I go back and listen and remember these hearing these stories and it's like, "Wow, he was super involved." Like that same story is like on one the it's sort of like that meme is like, "Oh, that's great. We have a super involved CEO." And then it's like, "Oh, we have a super involved [laughter] CEO. Is this good or bad?" Anyway, so if you're thinking if just thinking about it from a stock perspective, you have the keyman risk. You have all of the legal questions that the company is going to face and how you know just everything related to the sort of mega speed and what it knew when and all that stuff and then there's just the fundamentals of the business and there's chip demand but also its role in financing a big ecosystem of companies that either use and buy invidia chips.
**>>**: 嗯。显然有很多话题值得探讨。
Original English **>>**: Mhm. There's a lot to talk about obviously.
### 伯恩斯坦分析师下调评级的依据与考量 **>>**: 好的，我们别再浪费时间了。我们确实请到了一位完美的嘉宾，虽然有一段时间没请他了，但每当我们想专门探讨英伟达时，经常会与他交流。我们将与 Stacy Rasgon 展开对话。他是伯恩斯坦研究公司（Bernstein Research）的董事总经理兼资深分析师。他跟踪这家公司，覆盖半导体及半导体设备板块，对公司非常了解。如果你听过财报电话会议，很可能之前听过他的声音。那么，Stacy，非常感谢你回到 OddLots。我们太久没见了，大家都很想念你。
Original English **>>**: All right, let's not uh waste more time. We really do have the perfect guest, someone we haven't had him on in a while, but he's who we talk to frequently when we want to talk about Nvidia specifically. We're going to be speaking with Stacy Rasgen. He's a managing director and senior analyst at Bernstein Research. He covers this company. He covers semiconductor semiconductor equipment. Knows the company very well. If you listen to earnings conference calls, you've probably heard his voice before. So, uh, Stacy, thank you so much for coming back on OddLots. It's been too long. We've missed you.
**>>**: 确实很久了，能来到这里是我的荣幸，非常高兴能来。感谢再次邀请我。
Original English **>>**: It has been and it's my pleasure, my great pleasure to be here. So, thanks for having me on again.
**>>**: 当然。作为一名股票分析师，面对这样一个时刻并试图理清头绪，在职业层面上感觉有多离奇？毕竟你现在的职责就是去搞清楚真相。你目前正试图厘清多少个维度和变量？
Original English **>>**: Absolutely. How weird is this a moment as a stock analyst just figuring it out just professionally like your job is now to figure it out? How many dimensions and variables are you trying to work through?
**>>**: 对于覆盖这家公司的分析师来说，这确实是一个非常离奇的时刻，老实说也是一个令人悲伤的时刻。我的意思是，整场混乱没有任何好的解决方案或结局。坦白说，这简直是一场顶级规模的灾难，作为分析师跟进这个事件有种超现实感。这绝不是一件令人愉快的事。显然，黄仁勋的日子肯定比我们更难过，但这绝对是漫长而煎熬的一周。
Original English **>>**: This is a very weird moment as an analyst covering this company and a sad moment. Let's be honest. I mean, there's no good solutions or outcomes or anything else coming out of this entire mess. like it's just a disaster of the highest proportions frankly and and it's been sort of surreal covering this as an analyst. Um it's not been fun. Clearly you know Jensen's having a worse time than than we are but it's been a long week for sure.
**>>**: 所以你实际上在本周早些时候决定下调英伟达的评级。采取这一行动的催化剂是什么？因为在头条新闻仍在满天飞的时候，这可是一个相当重大的决定。
Original English **>>**: So you actually decided to downgrade Nvidia earlier this week. What was the catalyst for taking that move? Because that's a a pretty big call when the headlines are still flying, you know.
**>>**: 让我先退一步回顾一下。围绕这件事的新闻动态一开始零零星星地浮现，随后在周五晚上基本明朗化，当时有报道称他与这位名叫 Alice Wang 的人士存在关系。报道指出，美国情报部门担心她可能是中国的特工。顺便说一句，我当时并不一定信服，至今我也未必确信她一定是个间谍。
Original English **>>**: Let me step back actually. So the the news flow around this started uh I mean it'd been coming up in dribbs and drabs and it sort of crystallized on Friday evening when um the article started coming out that that he had had a relationship with this uh individual Alice Wayne. It was coming out that the US intelligence community was concerned that she might be an asset for China. By the way, I will say I was not necessarily convinced at that point. I'm still not necessarily convinced that that she was necessarily a spy,
**>>**: 对吧？她可能只是一个喜欢接近权力的女商人，走着瞧吧。但这就是周五开始传出的消息。我们在周日晚上发表了一份报告，概述了此类风险。在那时，我的看法是他或许还能留任。同样地，我对间谍说法的确信度相对较低，现在也仍未完全确信。而且她是客户，这算是一种灰色地带，但或许还可以妥善处理。需要明确的是，我并没有对他任何个人层面的行为做出评判，对这些进行道德审判不是我的工作。但我周日晚上的看法是，其实没有人真正希望看到他离开。这始终是风险所在。正如我们在那篇报告中概述的那样，这里的关键人物风险显然是其中最大的组成部分。我想我们稍后会更详细地探讨这一点。但是的，我周日的看法是，股东不想看到他离任，董事会当时也不想看到他离任，让他离开不符合任何人的最大利益。所以我当时认为他能挺过去。对我来说，转折点在于随后的补充披露，涉及他据称与下属之间存在的关系。不幸的是，这直接成了致命一击。你绝不能和自己的下属发生关系，随后还就此对董事会撒谎。在我多年来覆盖的公司中，就有其他首席执行官正是因为类似原因被解雇。坦白讲，有些被解雇的案例在这方面的情节甚至远没有这次所见到的严重。因此，当这些消息传出时，我们必须做出抉择：我们是否认为他被免职的可能性已大大增加？我当时的结论是肯定的。而就在第二天，我们确实收到了他们将他免职的确认消息。接下来我必须决定该如何对待这只股票。这正是讨论的核心。我当时开玩笑说，我和我的团队花在讨论该怎么做上的时间，远多于实际撰写下调报告的时间。需要说明的是，我们将股票评级从跑赢大盘（Outperform）下调至与大盘同步（Market Perform），并将目标价减半。写出这份报告没花多少时间，因为其实并没有太多复杂的模型分析要做，对吧？但你必须决定该采取什么评级动作。因此，从这个衡量标准来看，这大概是我不得不写出的最艰难的一次降级报告，因为你永远不想做出这种决定。正如我所说，我喜欢黄仁勋，我喜欢英伟达，我也喜欢它的故事。我过去一直是数据中心的铁多头，现在依然是。但你必须权衡这些因素：维持原评级的理由——我也对此有一定共鸣——在于股票估值极其便宜。它之前就已经很便宜了，眼下显然变得更加便宜。我们并没有下调财务预测数字。我实际上认为，其业务在近期的发展轨迹并未面临巨大风险。我不认为客户眼下会因为这件事而减少明年从他们那里采购数据中心产品。你可以认为当前的估值已经消化了大量的坏消息。这是一个论点。另一方面，我认为你会看到越来越多来自公司内部及外部的新闻爆料和盘中突发利空。非常明显的是，美国政府内部似乎有势力希望看到他们出局。坦率地说，我认为挑起这起事件的对华鹰派人士，最终大概也无法得到他们原本期望的局面。即便你是一个对华鹰派，这也不是一个理想的结果。但你将会看到越来越多要求开展调查的呼声。问题在于，我认为董事会和英伟达会发现放走他是一个错误，但遗憾的是，他自身的行为让他已经不可能留任。如果继续留任他，可能会是更大的错误，对吧？鉴于我们后续可能会看到的更多消息，我认为他们不得不尽快斩断这种联系。现在，从这些调查中流出的任何问题，至少你还可以辩称这不是英伟达的问题，而是黄仁勋个人的问题。我正在观望后续发展。比如他们如果发出传票调取他的短信和邮件，只要对任何人的历史深挖细究，你总能挖出点什么，而无论他们挖出什么，我猜都会被渲染成极其恶劣的状态。是的，我们从现在的一些头条新闻中已经看到了这种苗头。在面临所有这些情况的前提下，我实在难以向我的客户建言持有这只股票。此外，还得加上新任首席执行官带来的不确定性，无论是内部提拔还是外部空降，新 CEO 会秉持何种视角？他们在供应链上做出了大量决策，而众所周知，其中许多决策都建立在黄仁勋个人的握手协议与私人关系之上，换人后如何维持这些关系？我不知道。因此我最终得出的结论是，如果我无法坐在客户面前坚定建议“你们绝对应该逢低买入”，我就很难维持对它的买入评级。这就是我们下调评级的原因。但团队在动笔前确实讨论了数小时。实际动笔只用了 20 分钟，从这个角度来说并不是一篇冗长繁琐的分析。真是令人沮丧。
Original English **>>**: right? may just be like a business woman who likes to be close to power like we'll see. But that was the news that started coming out on on Friday. We published something on on Sunday evening sort of outlining those kind of risks. And at that point my view was that he could probably stick around. Again, I was less convinced. I still am not entirely convinced on the espionage aspect. And she was a customer, which is kind of a gray area, but maybe you can manage it. And and to be clear, like I'm not making any judgments on any of his personal activities. like that's not my my job to to to judge that. But my view at the time on Sunday evening was nobody really wanted to see him leave. That was always the risk. And like we sort of outlined in that piece, the key man risk here was clearly the biggest piece of this. And I assume we'll we'll talk a little more in detail on that. But yeah, my view on Sunday was that look shareholders didn't want to see him gone. The board at that point didn't want to see him gone and it wasn't anybody's best interest to see him gone. So I thought he could hang around. The catalyst for me was the the secondary disclosures about the relationships that he that he apparently had with his subordinates. And unfortunately, that's just kind of the kiss of of death. Like, you just you can't sleep with your subordinates and then lie to the board about it. That just I I've had other CEOs in my coverage over the years fired for exactly that. And and and frankly, even for things that were less egregious along those lines than what we saw here. And so when that came out, we sort of had to make a decision. you know, do we think that it's much more likely now that he gets dismissed? And and my conclusion at that point was yes. And the next day we actually got confirmation that that they did in fact dismiss him. And so then I had to decide like what do what do you do with the stock? And and this is where the discussion is. And and I was joking that we spent more time with me and my team discussing what to do versus writing the actual downgrade. And and to be clear, we downgraded the stock to from from out perform to market perform. We cut our target price in half. And it didn't take long to write it. there's not a whole lot of analysis to do, right? But you have to decide what to do and and so the it was a it was probably the most difficult downgrade though from from that standard that I've ever had to write cuz you know you don't you never want to do it and like I said I I love Jensen and I love Nvidia and I I love the story and I' I've been a very big data center bull and I still am but you sort of have to weigh to weigh this stuff like the argument to keep the call and and I have some sympathy for it is the stock is incredibly cheap. It was already cheap. It's it's gotten cheaper clearly at this point. We didn't cut our numbers. I actually don't think near-term, you know, trajectory for their business is really at a at a ton of risk. I don't think customers at at this point are buying less data center stuff from them next year because of this. And you can argue that this valuation that you're pricing in a lot of bad news. So that that's one argument. On the other hand, I think you're going to see more and more like news flow and tape bombs coming up both from inside the company as well as externally. It's very clear that there seem to be elements in the US government that that want them gone. And frankly, the China hawks that I think are probably the ones that kick this off are probably not going to get the scenario that they were hoping for. This is this is not a good scenario even if you're if if you're a China hawk. But you're going to see more and more calls for investigation. And by the problem is is I think it's the board and Invid will find that it was a mistake to let him go, but unfortunately he made it impossible for him to stay. It would it may have been even a bigger mistake to keep him, right? and and given the news we're likely to find. They they I think they had to sever that tie as quickly as possible. Um and now anything that comes out of these investigations, at least you can argue it's not Nvidia, it it was Jensen. I'm waiting to see what happens. You know, when they I don't I don't know, they subpoena his, you know, his text and his emails and you dig hard enough into into anybody's like like history, you're going to find something and whatever they find, my guess is going to be spun in the absolute worst possible light. Yeah, we've already seen that with some of the headlines now. And I just find it very difficult to advocate to my clients to own the stock in in front of of all of that. And and and now I fold on top of that the uncertainty around the CEO and what kind of a perspective is any new CEO going to have whether it's internal or external and they've made lots of decisions in the supply chain and you know that many of them are based on Jensen's handshake and and his personal relationships and how do you maintain that? I I don't know. So I I came down on the side if I can't sit in front of a client and advocate I I think absolutely you should be buying this dip. It it's it's hard for me to maintain the the the buy on it. So that that's why we downgraded. But it was I we discussed this for hours among the team before writing it. Writing it took 20 minutes, right? It's it's it's not it's not a long thing to to have to to go through on that standpoint. What a bummer.
### 地区营收划分与潜在走私及远程访问风险 **>>**: 让我们从股票的角度来拆解其中几个问题。既然你对财务预测数字有信心，认为不会出现大规模的订单取消，那我们来谈谈对亚洲或东南亚的销售有多重要。当我们讨论估值时，我们会谈到 2027 年的前瞻展望，或是对 2026 年的回顾。假设这里出现了实质性的限制措施，或者假设存在对走私行为睁一只眼闭一只眼的情况，以及围绕这些问题的各种质疑，未来必须采取行动；同时，英伟达向亚洲市场（与其说是中国市场本身，不如说是广义亚洲市场）销售的方式发生了重大变化。这对滞后指标有多重要？对前瞻指标有多重要？其中有多少真正面临实质性风险，无论是由于政策变化还是内部合规风控的收紧？
Original English **>>**: Let's just take a few of these questions from the stock perspective. Sort of break it down. you're confident in the numbers. You don't think there's going to be some big cancellation of orders, but let's just talk about how important are sales into Asia or Southeast Asia. Yeah. When we talk about valuations, we talk about the 2027 outlook or the backwards looking 2026 and you me like let's say there are meaningful curbs here or let's say you know there was some blind eye towards smuggling or any questions around this and something has to happen and there are meaningful changes in how Nvidia can sell into I should say the Asian market rather than the Chinese market per se. How important is that to the uh sort of lagging numbers and how important is this market to the forward numbers and how much of that is actually at material risk either because some change in policy or internal controls clamping down?
**>>**: 是的，毫无疑问。遗憾的是，这是一个很难回答的问题。公司确实按地理区域披露收入，他们目前采用这种方式。他们不久前做了一项调整，改为按客户公司总部所在地的地理位置来报告。举例来说，如果你看调整之前，他们向新加坡和马来西亚的销售额非常高，这些是货物的出运目的地，但实际下订单的公司却位于其他地方。
Original English **>>**: Yeah, you bet. So it it's a tough question to answer unfortunately. So the company does report revenues by geography and and they do this now. They made a switch a little while ago to report by the location of the headquarters of the companies. So for example, if you look prior, they had like very high sales going into Singapore and Malaysia and these these were like the location of of the shipments, but the companies that were doing that ordering were located elsewhere.
**>>**: 所以他们实际上做出了改变，并且近期许多半导体公司都做出了这一转变，即按照客户总部的所在地来披露地理分布。因此在这个基准下，像马来西亚和新加坡，基本上在他们报告的分类中：美国、中国台湾、包含香港在内的中国大陆以及“其他”，而“其他”就包含了新加坡、马来西亚以及所有其他地区。
Original English **>>**: So they've actually made a shift and and actually many semiconductor companies recently have made this shift to report geographical breakdown by the location of the headquarters of the customer. So on that basis like Malaysian and Singapore basically any they report no US, Taiwan, China including Hong Kong and and other and other would be Singapore and Malaysia and all these other ones.
**>>**: 这么说来，甲骨文（Oracle）在马来西亚的数据中心，在当前的财务报告规则下，实际上会显示为对美国的销售，从地缘政治的角度来看，这可以说极具误导性。
Original English **>>**: So an Oracle data center in Malaysia and there was I don't know if like that would under the current reporting regime that would actually show up as a sale to the US which from a geopolitical standpoint is arguably very misleading.
**>>**: 也许是，也许不是。我认为可以辩称，真正拥有并运营该数据中心的主体才是最重要的，因为我假设甲骨文——拉里·埃里森（Larry Ellison）总不至于把 Rubin 架构芯片藏在身后偷偷溜进深圳吧，对吧？
Original English **>>**: Well may maybe or maybe not. I mean, I could argue that the whoever is actually owning and running the data center is actually really the most important thing because I'm assuming Oracle's, you know, I'm assuming Larry Ellison's not sneaking into Shenzen, you know, with with with a Vir Rubin shoved up as behind it. Right.
**>>**: 这画面感真强。[笑声]
Original English **>>**: There's an image. [laughter]
**>>**: 我对此并不一定反对。而且再说一次，他们并不是唯一这样报告收入的公司。作为旁注，我退一步说，在很多情况下，半导体行业按出货地划分地理区域反而可能产生误导。举个例子，我随口举例，以高通（Qualcomm）这样的公司为例，他们会显示出极高的对华销售额。为什么？因为智能手机是在哪里组装的？是在中国组装的。
Original English **>>**: I don't necessarily object to this. And again, they're not the only ones that are reporting revenues like this. Many, as an aside, I should step back. The the the geographical split by shipments in many cases can be misleading for semis. So, for example, I'm I'm gonna I'll make it up. Take company like a Qualcomm, they will show very high sales into China. Why? Because where are the smartphones assembled? They're assembled in China.
**>>**: 那么他们实际上是把零部件运到哪里呢？出运到当地。所以即使是高通出货给苹果用于 iPhone 的零部件，在严格按地理出运地统计的情况下，很多时候也会显示为中国区域的收入，即便实际上并非如此。所以很有意思。
Original English **>>**: So where are they actually shipping the parts? They're shipping. So even a shipment to Apple from Qualcomm that's going into an iPhone in many cases under a a a strict geographical. So it would show up as Chinese revenue even though really it's it's not. So interesting. I
**>>**: 我实际上认为，所有权结构以及由谁实际运营这些设施非常关键。
Original English **>>**: I actually think the the ownership and and who's running these things is important.
**>>**: 好的。所以我并不反对这种划分方式。但在当前的拆分口径下，马来西亚和可以说整个“其他”板块，大概只占收入的 1%，规模非常小。对中国客户的直接销售，即包含香港在内的中国区，我认为上季度大约占其销售额的 8%。顺便说一句，这其中还包含了很大一部分的游戏和汽车业务收入。在数据中心产品方面，他们基本上几乎没有直接向中国出货。他们实际上确实获得了美国政府的一些许可去销售较旧的型号，但实际上中国政府并没有真正允许他们出运。这引发了我们稍后可以展开讨论的更广泛的对华政策议题，但我认为我们后面会谈到。眼下他们并没有直接向中国运送太多数据中心产品。这主要受限于中方原因，哪怕英伟达拿到了美国的许可，中方也不允许进口。因此绝大部分——我随口说一下数字，但我认为大致准确——美国客户占 60% 以上。剩下的部分，我认为在 20% 出头，我可以去查证一下，但我认为是 20% 左右，来自中国台湾的客户。所以我的推测是，如果真有物资实际被走私到中国大陆，那可能不是来自美国客户，对吧？大概率也不是来自中国客户，因为你本来就没卖给他们多少东西，数额极小。而马来西亚和新加坡客户的规模非常小。因此，那只可能涉及中国台湾的客户。没错。我看到外界流传着许多数字，有人认为其规模可能占到收入的 20%。我不知道究竟有多大，我怀疑没那么大，但中国台湾板块占营收的百分比确实在这个区间。所以或许有人会辩称他们并非真正的中国台湾客户，而是中国大陆的白手套，对吧？我不知道。我的猜测是大概不会到 20%，但可能也不是零。我的意思是，你永远无法彻底杜绝走私，但它确实存在一定的数额。我不认为能达到 20%，但确实有一定规模。不过这里还有另一个维度的因素，我认为目前正逐渐引起政府的关注，那就是他们希望打击中国境内的用户在远程调用算力。
Original English **>>**: Okay. And so I I don't object to the to the to the split. But on the current split, the Malaysian and say basically the other block, it's like 1% of the revenue. It's very small. The direct sales to Chinese customers. So China including Hong Kong, I think last quarter was roughly 8% of their sales. And and and by that will include a big chunk of their gaming and auto revenues as well. In terms of data center stuff, they're basically shipping effectively nothing directly into China. They actually do have some license from the US government to ship older parts, but in in fact the the that the Chinese government is not really allowing them to ship. And this gets into a whole broader discussion we can have around China policy that I I think we'll get to, but but right now they're not really shipping much data center stuff directly into China. And it's mostly because of China. China won't allow them even where Nvidia has the permission to do it from the US. So the vast majority I I I'm going to make up the numbers, but I I think these are roughly correct. US customers are over 60%. And then the remainder, which I believe is low 20s, I could look it up, but but I think it's in the low 20s, is Taiwanese customers. So my assumption would be that if there's like stuff that is actually getting smuggled into China, it's probably not from the US customers, right? It's probably not from the Chinese customers. You're not selling the stuff to them anyway. It's not very much. And the Malaysian and Singaporean customers is very slow. So it's got to be the Taiwanese customers. Right. I've seen a lot of numbers floating around. Some people would suggest, oh, it's like up to 20% of the revenue. I don't know how big it is. I doubt it's that big, but like the Taiwanese block is is in that ballpark in terms of a percentage of the revenue. So maybe you could argue they're not really chaiwanese customers. They're, you know, China fronts, right? In I don't know. My guess is it's probably not 20%. It's probably not zero. I mean, you can never stop smuggling entirely, but it it's some amount. I don't I don't think it's 20%, but it's some amount. There's another piece of this though which is and and this is now getting some traction I think with the government is is they want to crack down on China users within China using capacity remotely
**>>**: 对，这就是算力租赁的问题。
Original English **>>**: right this is the rental issue
**>>**: 目前，中国境内的公司从境外获取云服务是合法的，而他们想要叫停这种模式。我不知道这块业务规模有多大，我并不清楚……
Original English **>>**: so right now it is legal for a company in China to access you know cloud services from out outside of China and they want to put a stop to that I don't know how big that is I don't know I don't
**>>**: [音乐] [音乐]
Original English **>>**: [music] [music]
### 黄仁勋的运作模式与技术路线交接风险 **>>**: 你对黄仁勋在公司内的整体参与度有何印象？我们同样听到不少传闻，说很多交易在某种程度上是他与客户之间的个人握手协议。他在英伟达的日常工作状态是怎样的？
Original English **>>**: What's your impression of Jensen's overall involvement in the company? And again, we hear these stories that like a lot of the deals are sort of personal handshake deals between him and and you know, customers. What was his dayto-day like at Nvidia?
**>>**: 遗憾的是，我现在对他日常生活的了解[笑声]，恐怕比我想知道的还要多一些。他总是在四处奔波。他是公司的核心代表。他的职责有点像兼任首席技术专家、首席战略官，同时几乎还扮演着首席销售官和核心布道者的角色，对吧？从某种意义上说。
Original English **>>**: Well, I guess we know a little more about his day-to-day now, unfortunately, [laughter] than I wanted to. Um, I he's he's always running around. I mean, he is the he's the face. I mean, his job it's it's kind of, you know, he's he's sort of the chief technologist and this chief strategist and he's also like almost like the chief sales guy and the chief storyteller, right? In some sense,
**>>**: 没错。在公司内部他也像是一位人工智能的传教士。正如你所提到的，他拥有一个极其扁平的组织架构，绝大多数核心骨干都直接向他汇报。他制定所有关键决策，而从积极的角度来看，这意味着他们的行动可以极其敏捷。现在我们看到各种各样的人站出来说，他们曾经遇到过巨大的难题、处于水深火热之中，给黄仁勋发了封邮件，结果黄仁勋在七分钟左右的时间里就把问题解决了。由于他拥有这种程度的直接掌控力，他们的运转节奏是绝大多数公司无法企及的。我认为外界的许多交易确实是口头承诺；或者即使不是口头协议，很多客户和供应商之所以扩充产能，大概也是因为信任黄仁勋，是他信服了他们放手一搏去扩建产能。我认为新任管理者接手后，将极难去履行并维系这些承诺。我倒不是真正担心日常运营。他们那里有一批极其聪明的人才，有优秀的工程师，在这方面我现在并不真正担心。黄仁勋所拥有的这种直接掌控度，在他掌舵时运作得非常出色，但我认为一旦他离开，肯定会引发诸多问题。
Original English **>>**: right? And he's kind of been an evangelist like for AI internally. I mean, look, he as as you've you've said, he's got a very flat organization. most of the the key folks report to him directly. He makes all of the key decisions and and by the that's a positive sense that they can move very very quickly. We're seeing all these folks come out of the woodwork now about, you know, oh, they had some huge problem and they were in dire straits and they emailed Jensen and Jensen, you know, fixed the problem in seven minutes or whatever it was. They could move at a pace that most companies could not because he had that degree of direct control. I think a lot of these deals out there are handshakes or if they're not, I mean, you've got customers and suppliers that are presumably adding capacity because they trust Jensen and Jensen's convinced them to take that leap of faith and and and build it. And I think those are going to be very difficult things for a new co to come in and fulfill. Like I'm not really worried about the operations. I mean, they've got very smart people there and they've got great technologists and I'm not really worried about that at this point. said like the the degree of direct control that Jensen had it it worked very very well while he was there it's going to cause problems I think once he's gone for sure
**>>**: 在技术层面，暂且不论运营能否平稳延续。作为一名技术决策者，多年来必须做出重大的抉择，对吧？当我们思考未来一两年的 AI 或技术路线图时，或许这些已经明确成型，但后续显然还有重大决策必须制定。黄仁勋在把握行业趋势方向、做出正确重大抉择方面，显然拥有出色的过往记录。你如何看待英伟达在这方面的中期风险与能力？
Original English **>>**: on technology like okay there's who knows how smooth the sort of operations will continue as a technologist over the years you have to make some big calls right and so when you think like okay maybe with the AI or the tech roadmap for the next year or two is understood but like presumably there going to be big calls that have to be make made in future and he clearly has a good had an amazing track record of uh you know knowing where the puck was going and making getting some big calls right. How do you think about that sort of medium-term risk and the capacity of Nvidia?
**>>**: 这正是我所担忧的。同样，我并不太担心明年的收入，甚至不担心 2028 年的收入，但想想技术路线图：他们现在正在出货 Reuben，Reuben 架构已经成型；Reuben Ultra 听起来也已经定型；接下来大概会是 Feynman 架构，我不知道 Feynman 的成型度有多高；但在 Feynman 之后会是什么？他们需要做出什么样的抉择？未来的架构会是什么样子？
Original English **>>**: That's exactly what I'm worried about. So again, I'm not worried about so much the revenue next year, maybe even 28, but I mean you think about the road map. They're shipping Reuben now. Reuben is baked. Reuben Ultra sounds like it's baked. Like Fineman, I guess will be next. I don't know how fully baked that that is. And then what comes after Fineman and what what are the kind of calls they what what's the architecture look like?
**>>**: 是的。
Original English **>>**: Yeah.
**>>**: 他们会采用什么样的内存架构？会搭载多少 HBM？是否会采用其他方案？天知道。这些都是必须做出的重大转折性抉择。正如你所言，他在这方面拥有极佳的历史战绩。这就是我的顾虑所在：他是那个能坐下来深谋远虑构想未来 5 年、10 年的人，是那个能够洞察未来潜在拐点的人。英伟达今天之所以如此具有统治力，是因为他在 10 年前做出的决策。
Original English **>>**: What kind of memory architecture are they using? How much HBM is going to be there? Are they going to use something like who knows? These are big major inflections that will have to be called. And to your point, he's had a very good track record. And so that's my issue like like he's the one that sits there and and thinks out like 5 years, 10 years. He's the one that's seeing around those corners. The reason they are so dominant today is because of decisions that he made, you know, 10 years ago.
**>>**: 比如你在 10 年前做出的哪些决策，在你看来如今获得了丰厚的回报？
Original English **>>**: Like what like what was the decision he made 10 years ago that you would say is paying
**>>**: 例子有很多，单单是对 CUDA 的投入就足以说明。虽然 10 年前他们倒没有陷入绝境，但在 15 年前他们确实经历过极度艰难的时期，当时他向这个软件生态系统砸入了巨额资金，而那个时候我认为几乎没有人真正想要或在意这个东西。但看看它带来了什么：它构筑了一条巨大的护城河。当这些应用场景最终爆发时，他正好在正确的时间出现在了正确的地点。外界很多其他 CEO 大概都不会做出那样的决策。人们可能忘记了，虽然这只股票现在的表现极其惊人，但多年以来它一直只是一只 10 到 20 美元的股票。顺便提一句，按当前的股本计算，这两个数字还得除以 40，因为他们在那之后进行过一次 1 拆 40 的拆股。那么算下来是多少？
Original English **>>**: I mean just just for this has been plenty but I mean investing in CUDA I mean to the extent I mean they were in dire I mean they weren't in dire straits 10 years ago 15 years ago they they they were right and he was investing a ton of money into this software ecosystem that at the time like I don't think anybody really wanted or cared about right and but look look what it's got it's it's built a massive moat and he was in the right place at the right time when these these applications kind of came to light and plenty of other CEOs those out there probably would not have made that call. I mean, and if people forget, you know, the stock's been I mean, it's been amazing, but I mean, it was a it was a $10 to20 stock like for for years and years. And by the way, on the current share count, you got to divide both of those numbers by 40. They've done a 40 for one stock split since then. So, I mean, what what is that?
**>>**: 让我心算一下，折合现在的股本大约是 25 到 50 美分。
Original English **>>**: I do my math. 25 to 50 cents on on today's share count basis.
**>>**: 在年复一年的时间里都处于那个水平。当时他在大家都劝他收手的情况下持续进行这些投资，但他没有停步。因此，我不知道在未来 10 年的趋势研判中，他能洞察到什么先机，但现在必须由其他人来做出这些抉择，而不是他了。如果可以的话，我宁愿让他来做决定。抛开关键人物风险以及英伟达能否在失去黄仁勋后继续保持以往的增长速度不谈，当前正在浮现或者说加速的风险之一，是对华鹰派在芯片政策上的进一步抬头。我必须指出，由于我们是在 11 月 13 日星期五进行录制，今天早上的重大突发新闻之一，就是我们看到财政部长斯科特·贝森特（Scott Bessent）声称……
Original English **>>**: It would have been for for years and years and years while he was making these kinds of investments when people were telling him to stop and and he and he didn't. And so I don't know what kind of like like pucks he or where he sees the puck going over the next 10 years, but you're gonna have somebody else making that call now, not not him. I' I'd rather have him making it if I could, you know. So setting aside the sort of keyman risk and whether Nvidia will continue to grow at the pace it's been able to grow under Jensen is one of the risks that's emerging now or I guess accelerating just the the China hawks even more on the rise when it comes to chip policy and I have to say so again we're recording this on Friday November 13th one of the big items of breaking news this morning was we saw Treasury Secretary Scott Bessant like
### 对华芯片政策争端与行业生态 **>>**: 称这都是卢特尼克（Lutnick）的错，因为极速（Mega Speed）本就应该被列入黑名单。
Original English **>>**: saying that well this is Lutnik's fault. because Mega Speed should have been on the blacklist.
**>>**: 是的。依我看，整件事很大程度上可能肇始于对华鹰派的运作——虽然这有些阴谋论的色彩，但感觉很多端倪都由他们挑起。再说回 Alice Wang，她甚至不需要真的是一名间谍，她只需要能被合乎情理地描绘成间谍就足够了。即使其他丑闻都没有被曝光，假设他保住了职位，那固然很好，但他未来还能就中美人工智能政策公开发声吗？大概不能了，对吧？因此这或许就是整出事件最初的意图。他此前被传唤参加听证吹风，有三名共和党议员传唤了他要求汇报情况，而如今民主党人也介入了进来，我推测民主党人对他与特朗普走得如此之近非常不悦。因此政府内部有很多人可能联手针对他。同样，我认为眼下的结局大概也不是他们原本所期望的，我希望这不是他们想要的结果，因为这并非一个好的局面。但他日后在为对华政策发声倡导方面肯定会被迫噤声，因为我认为对华政策注定会变得更加严厉。不过我必须说，在这件事上我更倾向于站在黄仁勋的立场。他一贯的主张是——我们暂且把极速公司搁置一边，稍后可以再谈，但他总体的对华政策主张是：允许我们向他们出售较旧型号的芯片。此前有各种新闻炒作他们如何向中国出售最先进的芯片，其实并非如此。出售的是 H200 和 H20，这些已经是三到四代前的技术了。Hopper 架构早在 2022 年就推出了，所以已经是几代前的产品。但其核心逻辑在于，正如卢特尼克本人也曾表示的，我想他说过我们希望让中国对我们的技术产生依赖，对吧？我措辞不会用得那么尖锐，但道理确实如此。你希望他们围绕我们的技术来构建其本土的 AI 生态系统，这样我们才能洞悉进展并保持掌控。如果限制这些销售，就会促使他们去自主研发，而这也正是他们目前在做的事。如今英伟达拥有销售部分 H200 等芯片的许可，但由于中方为了扶持本土厂商而不允许进口，他们并没有实际售出。这就是为什么我觉得本周早些时候的一些指控讲不通：据称 Alice Wang 曾试图促使他就增加对华销售进行游说。首先，他本来就一直在这么做；其次，中国方面竟然会希望她游说争取中方目前根本不允许英伟达做的事情，这根本说不通。因此在对华政策上，我赞同黄仁勋的立场。我不认为我们应该把 Blackwell 或 Reuben 出售给中国，但销售较旧的产品我认为曾经是个好主意，不过现在恐怕为时已晚了。
Original English **>>**: Yeah. So, look, my my guess is is a lot of this may have started with these are the conspiracy theories, but I mean, feels like a lot of this started with the China Hawks. And again, even with with with Alice Wang, like she doesn't need to be a spy, she just needs to be able to be credibly painted as a spy. And if none of the other stuff had come out, like let's say he had kept his job, like that would be great. But could he ever be like a a a public voice on China AI policy ever again? Probably not, right? So that might have been the original goal of of all of this was to get because I I don't there's plenty of China and you know he was called uh for a briefing. They wanted to come and do a brief. It was three Republicans that that called him, right? I'm I'm in for for that. And then now we got the Democrats coming in who my guess is are not super happy with how close he's been to to Trump. And so you got a lot of folks in the government that may have been aligning against him. And again, I think this outcome is probably not what they would have wanted. I would I would hope that this is not what they would have wanted. Like this is not a great outcome. But but yeah, I think he would have been muted in terms of being advocate for China policy because I do think China policy is going to get going to get tougher. Now, I got to say though, I I'm more in Jensen's camp on this. His whole view has been and I'll put Mega CB aside for a moment. We can talk about Mega Speed, but his general China policy has been let's allow allow ourselves to sell them the older parts. So, there's been all this news about how they're selling the most advanced stuff in China. Like, no, they're not. These are H200s and H20s. This is like three or four generations old. Hopper showed up in 2022 is actually when it showed up. So, it's several generations old. But the idea is let and even Lutnik has said this. I think he said we wanted to get China addicted to our technology, right? So, I wouldn't have put it that harshly, but it's kind of true. You'd like them to build their local AI ecosystem around our technology so we can know what's going on and keep it under control. And if you limit those sales, you start to encourage them to build their own. And which is exactly what they're doing. And today, like again, Nvidia has licenses to sell H some H200s and everything and they're not selling them because China won't allow them to do it because China wants to encourage their their local players. And that's why I find a lot of this stuff, you know, there were some allegations like earlier in the week that supposedly like like Alice Wang was trying to get him to lobby for more sales into China. So like number one, he's doing that anyways and and number two, they want China wants her to lobby for something that they're already not allowing Nvidia to do. doesn't make any sense. So, I'm on Jensen's side on the on on the China policy. I don't think we'd be we'd be selling Blackwells or Reubens to China, but older stuff, I think it would have been a good idea. I think it's too late now, though.
**>>**: 我在这个问题上没有预设立场，也缺乏定论，但请为现状做个辩护，或者像人们常说的那样构建一个“强力反驳假说”（Steelman）……
Original English **>>**: Look, I have no opinion on this question and I don't know, but justify the status quo where you can't sell the most in U or Steelman as people like to say,
**>>**: 强力论证。
Original English **>>**: Steelman,
**>>**: 也许可以说“硅人论证”（Silicon Man），或者“芯片人论证”（Chipman）。请为这个论点提供最强有力的辩护：为什么我们一方面不能向中国出售最先进的 GPU，另一方面却可以卖给一家在马来西亚运营的新加坡公司，并允许中国 AI 企业远程访问。这在很多人看来完全是自相矛盾的。
Original English **>>**: maybe Silicon Man, maybe we should say Silicon or Chipman. chipman the argument that we shouldn't be selling the most advanced GPUs into China, but that it's okay to sell them to a Singaporean company with operations in Malaysia that allows Chinese AI companies remote access. That strikes a lot of PEOPLE AS INCOHERENT.
**>>**: 没错，完全如此。这是一个完全独立的问题。正如我刚才所说，看起来他们即将对远程访问展开打击。平心而论，我并不清楚当前云计算开支中究竟有多大比例是被中国境内的客户所使用，我不知道，我们将拭目以待。是的，我认为这确实在逻辑上缺乏前后一致性，除非你只允许中国境内的人员远程调用 Hopper 架构。如果你允许中国境内的人员远程使用大批量的 Blackwell 和 Reuben 算力，那确实是一个可能需要、也本应该被堵住的漏洞，对吧？但这是另一个层面的辩论，不同于“我们是否应该允许向中国出售低端硬件”。无论采取直接交付还是远程访问的形式，我们是否该允许中国获取低端产品？如果核心关切是我们不希望中国接触到最新、最强的技术，那显然我们不该让他们接触到。但这其实是两个略有差异的问题。
Original English **>>**: NO, ABSOLUTELY. And and look, this is a whole other issue. And so, like I said earlier, it looks like they're going to be cracking down on remote access. And to be fair, I do not know how much of the current spending like on on cloud compute is is being used by by Chinese customers in China. I I don't know like we'll we'll see. And and yeah, I I don't think that that is is necessarily, you know, a consistent point of view unless you're only allowing folks in China to use, you know, hoppers remotely. you know, if you're allowing folks in China to use to to remotely access black welds and Reubins in size, then yeah, that that's a hole that maybe need maybe needs to be plugged and maybe maybe should have been, right? But that's a different debate than should we be allowing sales of like like lower-end stuff into into China. Should we be allowing access whether it's direct or or remote, frankly, of lower-end stuff into into China? If the concern is we don't want to have China access to the latest and greatest, like clearly we shouldn't. They shouldn't have access to the latest and greatest. But they're two slightly different issues.
### 下周财报看点与长期竞争格局 **>>**: 正如 Joe 开篇所言，英伟达接下来的“已知的未知数”就是下周公布的财报。没错。你届时具体会重点关注什么？如果说眼下最大的风险在于关键人物风险以及美国对华政策的未来走向，我们究竟能从财报中获知什么？
Original English **>>**: So as Joe said in the beginning, the next known unknown coming up for Nvidia is results next week. Yeah. What are you actually looking out for there? Like if the big risk now is, you know, it's the keyman risk and it's the future trajection of China policy in the US like what are we actually going to learn from earnings?
**>>**: 首先我想说，财报数据在技术层面上最好是无可挑剔的[笑声]，对吧？上帝保佑数据千万别出差错。正如我所说，我不认为短期业务轨迹会因此出现实质性改变，我不这么认为。当然我们可以探讨长期前景。一些竞争对手的股价近期大幅上涨，正是基于一种预期，即英伟达长期技术路线图的削弱会增强其他对手的实力，这或许不无道理，我们可以探讨。如果短期内出现剧烈变动，那将会非常棘手。他们必须在近中期发展路线上提供确定性，给予投资者信心与信念，证明现有的增长轨迹没有发生偏离。曾任首席财务官、现任临时首席执行官的科莱特（Colette Kress）必须传递信心，表明大船稳固且仍在平稳前行。关于这次事件的来龙去脉，他们将面临大量尖锐质询，还会收到许多关于人才留存等方面的提问。挺有意思的是，虽然我不记得具体数字，但他们员工中有相当大一部分人的身家达到了几千万甚至上亿美元……
Original English **>>**: Yeah. I mean, so look, first I would say the numbers better be technically perfect, [laughter] right? God forbid something's off in the numbers. Now, like I said, I don't think you're going to see any material changes to the the near-term trajectory as a result of this. I I don't I don't think so. Again, we can talk about the longer term. You know, there there's other other names that have been running a lot more like some of their competitors have been running a lot on the belief that a weaker Nvidia road map over the long term strengthens it, which which may be true. We we can we can debate that I I would be surprised if there were change. If there were changes in the near- term, that would be very problematic. I mean I mean clearly look I I guess they'll have to give some certainty on that near-term trajectory near to medium-term trajectory and give investors confidence and conviction that at least that the current trajectory is not changing. I think they're going to I don't know what they're going to say around you know there she's Kleta who's the who was the CFO and is now the interimm CEO is going to have to give like some confidence that like the ship is stable and still sailing right. They're going to get a lot of tough questions on how this went down. They're going to get a lot a lot of questions on talent retention and things like that. And you know, it's it's funny. You know, I I can't remember the number, but like a huge portion of their employees have net worths that are in the tens or even hundreds of millions of dollars
**>>**: 而且他们按理说完全不需要再出来工作，他们留在那里或许就是因为黄仁勋还在。如果他走了，情况会变成怎样？我们如何能确信核心人才团队依然稳固？她将面临所有这些类型的质问，我们将看看她如何作答。这肯定会是一场极具看点的电话会议。下周发生的另一件事是 Anthropic 的 S-1 上市招股书披露。如果你阅读那份招股书，他们强调了自身采用多芯片架构的事实，芯片对他们而言是同质化的，TPU、Trainium 芯片以及 AMD 的 GPU 都只是这一庞大同质化计算技术栈的一部分。我认为他们从技术角度明确提出这一点很有意思，而且我们知道 Anthropic 在芯片出口政策上也极其鹰派。暂且抛开所有丑闻不谈，从竞争地位来看，无论是 TPU 还是其他芯片方案，即便完全不考虑黄仁勋的个人风波，英伟达的护城河目前看起来是否依然坚不可摧？
Original English **>>**: and they presumably don't need to work and they're there maybe because Jensen's there and like what what happens to that? Like, so how do we get confidence that that the talent base is still going to be? She's she's going to get all these kinds of questions and and we'll see what she can say. It's going to be an interesting call for sure. The other thing that's happening next week is the anthropic S1. And if you read that S1, they make a point about the fact that they use multi-chips and that chips are homogeneous to them and that TPUs and tranium chips and GPUs from AMD are all just part of this big homogenized stack. I think it's interesting they say that is a technical thing and we also know that Anthropic specifically is very hawkish on chip export policy. from a competitive position whether we're talking about TPUs whether we're talking about others set aside all the scandals and everything else how salailable is the Nvidia moat looking right now even if we weren't talking about Jensen scandals
**>>**: 我不会对 Anthropic 发表任何评价，我要非常明确地向合规部门声明把这一话题搁置，绝不发表意见。但总体而言，关于客户采用多元化供应商的趋势，确实有多家企业在这么做。举例来说，OpenAI 与英伟达签有大单，但他们与博通（Broadcom）也有相当大的合作，甚至与 AMD 达成了重大协议；Meta 同样如此，还有其他厂商。谷歌显然自主研发 TPU，他们虽然仍采购 GPU，但主要依托 TPU。
Original English **>>**: I'm not going to say anything about anthropic I'm going to put that aside very clearly to my compliance people totally not going to say anything but in general the idea of customers using multiple sources um there are a number of them so for example open AI They've got a big deal with Nvidia. They also have a pretty big deal with Broadcom. They've even got a big deal with AMD. Same thing with Meta and there are others. Google clearly, I mean, they do their own T. They still buy GPUs, but they they do TPUs.
**>>**: 如今有越来越多的客户出现，谷歌现在也开始在二级市场上向其他客户销售 TPU 了。
Original English **>>**: You've got more and more customers now that are starting that Google is now starting to sell TPUs to like on a on a secondary basis as well.
**>>**: 这种情况正在发生。就目前而言，我总体的看法是，更核心的问题在于市场蛋糕是否在持续变大。英伟达此前占据了该市场的大部分份额，随着其他玩家的涌入，英伟达的市占率大概率会下降。但关键在于这块蛋糕究竟够不够大。如果蛋糕足够大，或许容得下所有人；如果蛋糕不够大，那大家都会陷入困境，对吧？如果你看一下对明年的数据预测，假设在 2027 年不发生重大变故的前提下，英伟达目前的收入增长指引至少在 70% 以上，这将使公司的体量逼近 7000 亿美元。博通则暗示其 2027 年的数据中心收入将同比翻番达到约 1150 亿美元，并在 2028 年再次翻番至 2300 亿美元，如果供应能跟上，根据需求前景甚至可能做得更多。再看看 AMD，倒推计算的话，他们暗示其 GPU 销售额可能同比接近翻三倍，规模可能会进入 400 多亿美元的区间。甚至高通也在切入这个市场，他们表示要做 50 亿美元规模。此外还有联发科（MediaTek）、美满电子（Marvell）等其他厂商，大家目前都过得不错。但在宏观层面上，是否存在更多替代选项？确实存在。AMD 的技术路线图虽然推进需要时间，但正在逐步改善；客户对 TPU 的使用在增加；像博通这样的厂商为定制芯片争取到了更多客户；而至于 AMD 本身，大家常把英伟达的 CUDA 称为软件生态系统，我认为这确实是护城河。不过 AMD 和其他厂商如今也谈到，据传 AI 在编写软件方面正变得越来越出色，企业正尝试利用 AI 来辅助完成代码的跨平台移植。这又回到了黄仁勋身上：他所做出的决策不仅在硬件端，也一直在软件端，确保 CUDA 的护城河坚不可摧。英伟达内部还有其他优秀的人才，比如被传为潜在接班人选的 Ian Buck，人们常说是他发明了 CUDA，他是 CUDA 背后的核心推动力，目前负责超大规模数据中心业务。他们内部有顶尖的技术专家可以维持业务运转，但软件平台未来走向的重大抉择仍需有人拍板。作为客户，在未来两三年或许不必担心，但从长远来看，是否需要更严肃地考虑分散采购风险？也许需要。这也是为什么过去一周左右，在这一连串新闻流的发酵下，部分竞争对手的股价表现强劲。
Original English **>>**: Um, so that is happening and look, my my general view for now has been that it it's not the more the more important question so far has been is the opportunity bigger? or is there not? You know, Nvidia's had like the bulk of this market and presumably you'll get other players coming in and that market share will go down. But is the pie big or is it not? If it's big, there's probably room for everybody and if not, everybody would be screwed, right? And if you look at the numbers for next year just for I guess coming on this for 2027 you assuming no changes Nvidia currently is guided for revenue growth of at least 70% which would put them in the ballpark in spitting distance of 700 billion like for the company. Broadcom has suggested that their data center revenues will double year-over-year in 27 to like 115 billion and will double again um in 2028 to 230 and again with with demand outlook to do even more if if they can if they can get get there. Um you look at at AMD AMD is if you back into it they're kind of suggesting that their GPU sales will probably almost triple year-over-year and you're probably in the into the low 40 billions for AMD. Even Qualcomm is getting into this market like they're going to do they said they're gonna do five billion right and you've got others MediaTek and Marll and and all these others that they sell everybody's doing fine right now but in general are there more like alternative options there are and look you know a AMD is their road map is it it takes time but it's getting better and again you're seeing more customers using TPUs and you know guys like Broadcom have had more customers for their custom chips AMD themselves you know everybody talks about Nvidia CUDA as as Cuda is their software ecosystem. I think it is a moat. Um, but AMD has talked and others have talked about now AI is apparently getting good at writing software from what we've heard and using AI actually to to to help to port things over like companies are experimenting with that. And again, this this gets back to Jensen, you know, part of the decisions he's been making are not just on the hardware side, they've also been on the software side, you know, and so making sure that that CUDA mode stays wrong. I mean, they've got other people there that are good. You got Ian Buck who's been floated as a name for potential successor. He he's sort of like in people say that he invented Cuda. He's kind of that was a driving force behind CUDA and he runs the the he runs the hypers scale business over there. They've got good technical folks over there who can keep this going but again the some of those big decisions on where where even the software platforms going still have to be made and and if I'm a customer you know again again not next year or even the year after but like longer term do I maybe need to think a little harder about potentially diversifying my footprint? maybe and this is why some of their competitors have have have been strong in the in the last week or so in in the in the wake of all of this news flow.
### “AI央行”与资本开支风险 **>>**: 我还有一个最后的问题，关于人们所称的“AI 中央银行”，以及英伟达芯片需求之所以如此旺盛，在多大程度上是因为英伟达以某种形式提供了供应商融资。从这只股票的角度来看，这构成了多大的风险？
Original English **>>**: I have one last question and this is just about what people call the central bank of AI and the degree to which demand for Nvidia chips is partly as high as it is because it's vendor financed in one way or another from Nvidia. How much risk is that when you think about the stock?
**>>**: 我不确定。但他们确实拥有“AI 银行”的声誉，而且我认为这在很大程度上是由黄仁勋的个人意志所推动的。
Original English **>>**: Yeah, I I don't know. Um but again they do have this reputation of being the the bank of AI and again I think a lot of this is is being was being driven by the force of Jensen's will.
**>>**: 是的。
Original English **>>**: Yeah.
**>>**: 已经有报道传出——我们也略有耳闻——科莱特可能并没有完全赞同他们所推进的某些庞大资本动作，不像黄仁勋那么积极。顺便提一句，昨天有报道流传已久的消息，这并不令人震惊，她是首席财务官，把控风险本就是她的职责。但这让人不禁深思。未来肯定会迎来一位新的正式首席执行官，届时管理层的风险偏好会是怎样？我不知道，但这是一个值得提出的好问题。这同样是悬在外部的未知数之一。在调用公司资本方面，黄仁勋显然属于极度激进派，不仅用于兜底和融资支持，还大量用于锁定供应链。外界常常抛出各种天文数字，说他们耗资数千亿美元，但必须记住，英伟达流出的大多数实际物理现金支出并不是用于兜底和融资，而是用于确保供应链产能，锁定内存、晶圆和 CoWoS 先进封装产能等等。而其中很多合作同样建立在他个人的伙伴关系与私人交情之上。
Original English **>>**: You know there's already been stories coming out which we we kind of heard that like potentially Colette was not maybe as completely on board with some of the the magnitude of what they're doing as maybe Jensen was. And by the way that was reported yesterday that that's been floating around for a while and it's it's not Chucky. She was a CFO. That's her job, right? But one has to wonder. It'll be very interesting. They're going to get a new permanent CEO at some point and and what will the appetite be? I I don't know, but it's a good question to ask. These are again some of some of those unknowns that are out there. I I don't know. I'd say Jensen was clearly on the more bullish side of of committing company capital for this. Not not just for, you know, back stops and things, but I mean also for supply. People throw in all these numbers. I mean, they spent hundreds of billions of dollars. You have to remember, but most of the actual physical cash outlay that's gone out from medieval stuff is not back stops and financing. It's to secure supply. to secure memory and wafers and co-s and that kind of stuff, right? And and I mean a lot of that was I they're based on his own partnerships and relationships as well.
**>>**: 好的，Stacy Rasgon，我们今天就先聊到这里。事态发展极为迅猛，但很高兴能在财报公布前进行一番前瞻梳理。[音乐] 非常感谢你重返 Odd Lots。
Original English **>>**: All right, Stacy Rasgan, we'll have to leave it there. Very fast unfolding story, but glad we could get a preview in uh before the earnings. Thank you [music] so much for coming back on OddLock.
**>>**: 随时乐意效劳，这是我的荣幸。
Original English **>>**: Oh, you bet. My pleasure. Anytime.
**>>**: [音乐] [音乐]
Original English **>>**: [music] [music]
### 总结讨论：员工去留、华盛顿内斗与地缘博弈 **>>**: 留下了很多重大的疑问。Tracy，Stacy 提到的一点我认为很有意思：英伟达那些身家暴富的员工现在会作何打算？他们会不会觉得“现在正是退休的好时机，现在正适合去享受生活”？我不认为科技圈人士会普遍抱着“我已经赚够了、准备享受人生”的心态，但或许有少部分人会在这里停下脚步，直接选择离职。
Original English **>>**: Lots of big questions. Tracy, one thing that Stacy mentioned that I think is interesting is what what happens now to all of the really rich Nvidia employees. Do they just like, you know what, maybe now is a good time to retire. Maybe now is a good time to just enjoy life. I don't think tech people have a reputation for being like, you know what, I'm good. Like enjoy. I'm ready. Hey, I'm ready to just enjoy life. Maybe a few of them will take a pause here and just just walk out the door.
**>>**: 这也是我针对 Anthropic IPO 一直在问的问题。这对于旧金山来说基本上是一场巨额的流动性套现盛宴。人们认为持有那些股票的人拿到钱后实际上会做些什么？似乎没人知道，因为似乎没人真正了解科技狂人们除了开发技术之外，生活里还想做些什么。
Original English **>>**: This has been a question that I've been asking about the anthropic IPO, like this huge liquidity event for San Francisco basically. And what do people think that everyone who owns that stock is actually going to do with the money? And no one seems to know because no one know like no one really seems to know what tech bros want to do with their lives other than develop tech.
**>>**: 是的，核心就在这里。如果在 Anthropic，有人甚至会用某种词汇——我不说那个词，但有人称其为某种狂热组织。所以就算成了亿万富翁，大家依然会继续工作。也许这是一种积极意义上的狂热。但如果此前英伟达存在对黄仁勋的崇拜，而他现在离任了，那么身家上亿的员工何必继续留守？
Original English **>>**: Yeah, that's the Yeah, that's the thing. Like if you're like at anthropic there's you know people I'm not going to say the word but other people call that a cult. So it's like all right it doesn't matter you become billionaires you just keep working it cult in a good way maybe but but like if there was a cult of Jensen and he's gone then it's like yeah like why why stay if you're worth 100 million maybe not
**>>**: 另外他个人的持股会发生什么变化？他会选择抛售吗？眼下他大概有充分的理由这么做。但我认为整件事给我带来的一个宏观层面的启示是，华盛顿特区对华政策的鹰派基调似乎将进一步升级。仅从舆论表象来看，哪怕缺乏对所有指控的实际事实核实，感觉这种势头已经开始显现。因此观察英伟达将如何应对会非常耐人寻味，特别是考虑到黄仁勋一直是主张“开拓中国市场对我们未来有利”的芯片领军人物之一。
Original English **>>**: also what happens to his shareholdings alone right is he going to sell those he would probably have a decent reason to at the moment but I also think like the one takeaway sort of macro takeaway from all of this for me is that it feels like the China hawkishness is going to ratchet up in DC. Like it just based on the optics alone, even without like actual um factual verification of all these claims, it feels like we're starting to see that. And so that like it'll be interesting to see how Nvidia actually reacts to that, especially given that Jensen is one of the chips guys who has been making the argument that like actually diversification into the China market would be good for us in the future.
**>>**: 不，这确实将是一件大事。今天早上看到贝森特将卢特尼克推出来背锅，白宫内部围绕中国芯片问题的内斗显然已经彻底公开化。那些主张向中国销售芯片的人必须牢记一点：外界可能认为“我们不应该向中国出售芯片，因为我们不想让他们的 AI 模型赶上我们”，但实际上我们并没有那么多的筹码。中国可以说：“好啊，稀土等关键原材料断供。”所以这些关系极其微妙脆弱、很容易偏离轨道，因为中国在很多其他领域显然对美国拥有大量筹码。如果在这一环节稍微打破平衡，你很难预料会引发怎样的连锁反应。但显而易见的是，对华鹰派抓住了这一契机，正试图推动美国出口管制政策发生重大转向。
Original English **>>**: No, this is this is going to be huge. And uh seeing like Bessant throw Lutnik under the bus this morning. It was obviously this inter sort of interessing fight within the White House on the sort of China chips question that's now clearly spilled out into the public in a big way. And I think the people who argue like no let's sell into China. You have to reme, you know, one thing I think that's important to remember is like, you know, people may say, "We don't want to sell chips into China because we don't want their AI models to catch up with ours." We don't have like a ton of like leverage, though. You know, China could say, "All right, fine. No more rare earths, etc." So, it's like these are like very precarious, like, you know, unstable. These are relationships that are easily knocked off course because China obviously has plenty of leverage over the US and many other areas. So like you you you you you knock that a little bit and you don't really know how it's going to spill over, but clearly clearly the the China hawks have like seized a moment here and are trying to like get a big change in the way the US conducts export policy.
**>>**: 确实如此。而且贝森特不仅仅是把卢特尼克推下车，他还让司机倒车多碾压了几次，接着又拦住救护车不让救，这完全是一场极其夸张的公开清算。好的。[笑声] 我们今天就聊到这里吧？
Original English **>>**: Absolutely. And Besson didn't just like throw Lutnik under the bus. He had the bus driver like back up and run him over a few more times and then block the ambulance and like this was a throwing under the bus in epic proportions. Yeah. All right. [laughter] Shall we leave it there?
**>>**: 好的，就到这里吧。
Original English **>>**: Yeah, let's leave it there.
**>>**: 感谢收听新一期的 Odd Lots 播客。我是 Tracy Alloway，你可以在推特上关注我 @TracyAlloway。
Original English **>>**: This has been another episode of the All Thoughts podcast. I'm Tracy Aloway. You can follow me at Tracy Aloway.
**>>**: 我是 Joe Weisenthal，可以关注我的推特 @TheStalwart。关注我们的制作人 Carmen Rodriguez（@CarmenArmen）、Dasho Bennett（@DashoBennett）、Kale Brooks（@Kellbrooks）以及 Kevin Lozano（@KevinLloydLozano）。当然，你也可以在我们的 Discord 社群里 24 小时全天候[音乐]探讨所有这些话题。Tracy 和我每天都为 Odd Lots 新闻通讯撰稿，欢迎前往 bloomberg.com/oddplots 订阅查阅。
Original English **>>**: And I'm Jill Weisenthal. You can follow me at the stalwart. Follow our producers Carmen Rodriguez at Carmen Armen Dasho Bennett atbot Kalebrooks at Kellbrooks and Kevin Lozano at Kevin Lloyd Lozano. And of course, you can chat about all of these topics 247 [music] in our Discord. And Tracy and I write every day for the OddLotss newsletter, which you can find and subscribe to at bloomberg.com/odlotss.
**>>**: 如果你喜欢这次对话，如果你喜欢我们从跟踪该公司的分析师视角探讨英伟达的未来，请在各大播客平台为我们留下好评。如果你正在观看视频版本，请点赞、[音乐]留言或订阅频道。请记住，如果你是彭博订阅用户，可以在 Apple Podcasts 的 Bloomberg 频道免广告收听我们所有的节目，只需按指引操作即可。
Original English **>>**: And if you enjoyed this conversation, if you like it when we talk about the future of Nvidia from the perspective of the analysts that cover it, then please leave us a positive review on your favorite podcast platform. And if you're watching on video, then please [music] like or leave a comment, or better yet, subscribe. And remember, if you're a Bloomberg subscriber, you can listen to all of our episodes absolutely adree. All you need to do is [music] go to the Bloomberg channel on Apple Podcasts and follow the instructions there.
**>>**: 感谢收听。
Original English **>>**: Thanks for joining.
**>>**: [音乐]
Original English **>>**: [music]
**>>**: 我知道。
Original English **>>**: I know