章节 1：Fairwater 2 数据中心：AI 算力的十倍扩张

📝 本节摘要：

Satya Nadella（萨提亚·纳德拉）和 Scott Guthrie 接受了采访，并带领访谈者参观了微软全新的 Fairwater 2 数据中心，该中心是目前世界上最强大的设施。微软致力于每 18 到 24 个月将训练容量提升 10 倍。Fairwater 2 的网络光纤数量几乎相当于两年前 Azure 所有数据中心的总和，旨在为未来的大型训练任务聚合计算能力（flops），并能够在两个不同区域进行训练，预计实现比训练 GPT-5 时高出十倍的增长。

Start of Content

[原文] Today we are interviewing Satya Nadella.

[译文] 今天我们正在采访萨提亚·纳德拉（Satya Nadella）。

[原文] "We" being me and Dylan Patel, who is founder of SemiAnalysis.

[译文] “我们”是指我，以及 SemiAnalysis 的创始人迪伦·帕特尔（Dylan Patel）。

[原文] Satya, welcome.

[译文] 萨提亚，欢迎。

[原文] Thank you. It's great. Thanks for coming over to Atlanta.

[译文] [SN]: 谢谢。太棒了。谢谢你们来到亚特兰大。

[原文] Thank you for giving us the tour of the new facility.

[译文] 谢谢您带我们参观新设施。

[原文] It's been really cool to see.

[译文] 看到它真是太酷了。

[原文] Absolutely.

[译文] [SN]: 绝对的。

[原文] Satya and Scott Guthrie, Microsoft's EVP of Cloud and AI, give us a tour of their brand new Fairwater 2 data center, the current most powerful in the world.

[译文] 萨提亚和微软云与人工智能执行副总裁（EVP of Cloud and AI）斯科特·格斯里（Scott Guthrie）带我们参观了他们全新的 Fairwater 2 数据中心，这是目前世界上最强大的数据中心。

[原文] We've tried to 10x the training capacity every 18 to 24 months.

[译文] 我们一直努力在每 18 到 24 个月内将训练容量增加 10 倍。

[原文] So this would effectively be a 10x increase from what GPT-5 was trained with.

[译文] [Q]: 那么，这实际上将比训练 GPT-5 时所用的容量增加 10 倍。

[原文] So to put it in perspective in the number of optics, the network optics in this building is almost as much as all of Azure across all our data centers two and a half years ago.

[译文] [SN]: 所以从光纤（optics）数量的角度来看，这座建筑中的网络光纤几乎相当于我们两年前所有 Azure 数据中心（data centers）的总和。

[原文] It's got like five million network connections.

[译文] 它有大约五百万个网络连接。

[原文] You've got all this bandwidth between different sites in a region and between the two regions.

[译文] 你在同一区域内不同站点之间，以及在两个区域之间，都拥有所有这些带宽。

[原文] So is this like a big bet on scaling in the future, that you anticipate in the future that there's going to be some huge model that will

[译文] 那么，这是否是对未来扩展（scaling）的一大押注，即您预期未来会出现某个需要两个完全不同的区域来训练的巨大模型?

[原文] require two whole different regions to train?

[译文] 需要两个完全不同的区域来训练?

[原文] The goal is to be able to aggregate these flops for a large training job and then put these things together across sites.

[译文] [SN]: 目标是能够为大型训练任务聚合这些浮点运算能力（flops），然后将这些资源跨站点整合起来。

[原文] The reality is you'll use it for training and then you'll use it for data gen, you'll use it for inference in all sorts of ways.

[译文] 现实情况是，你将用它进行训练，然后用它进行数据生成（data gen），你将以各种方式用它进行推理（inference）。

[原文] It's not like it's going to be used only for one workload forever.

[译文] 它不会永远只用于一种工作负载（workload）。

[原文] Fairwater 4, which you're going to see under construction nearby, will also be on that one petabit network so that we can actually

[译文] 你们即将看到正在附近建设的 Fairwater 4，也将接入该一拍比特（petabit）网络，这样我们就可以真正地

[原文] link the two at a very high rate.

[译文] 以非常高的速率连接这两个设施。

[原文] Then we do the AI WAN connecting to Milwaukee where we have multiple other Fairwaters being built.

[译文] 接下来，我们进行人工智能广域网（AI WAN），连接到正在建设多个 Fairwater 设施的密尔沃基（Milwaukee）。

[原文] Literally you can see the model parallelism and the data parallelism.

[译文] 实际上你可以看到模型并行（model parallelism）和数据并行（data parallelism）。

[原文] It's kind of built for, essentially, the training jobs, the super pods across this campus.

[译文] 它本质上是为训练任务，即跨越这个园区的超级计算集群（super pods）而构建的。

[原文] And then with the WAN, you can go to the Wisconsin data center.

[译文] 然后通过广域网（WAN），你可以连接到威斯康星州（Wisconsin）的数据中心。

[原文] You literally run a training job with all of them getting aggregated.

[译文] 你确实可以运行一个聚合了所有这些资源进行训练的任务。

[原文] What we're seeing right here is a cell with no servers in it yet, no racks.

[译文] 我们现在看到的这里是一个单元（cell），里面还没有服务器，也没有机架（racks）。

End of Content

章节 2：基础设施设计与及时扩展的必要性

📝 本节摘要：

访谈者试图通过询问机架和单元的数量来推测数据中心的规模，引发了轻松的对话。随后，对话转向了基础设施的设计挑战。Satya 指出，硬件设计（如使用 GB200s 和 NVLink）必须与模型架构耦合优化，但考虑到芯片迭代速度（如 Vera Rubin Ultra 带来的功率密度和冷却要求的变化），构建全部符合单一规格的基础设施是危险的。因此，微软的哲学是必须“及时扩展（scaling in time）”，而不是一次性扩展后被困住。

Start of Content

[原文] How many racks are in a cell?

[译文] [Q]: 一个单元（cell）里有多少个机架（racks）?

[原文] We don't necessarily share that per se, but…

[译文] [SN]: 我们不一定透露具体数字，但是…。

[原文] That’s the reason I ask.

[译文] [Q]: 这就是我问的原因。

[原文] You'll see upstairs.

[译文] [SN]: 你会在楼上看到。

[原文] I'll start counting.

[译文] [Q]: 我会开始数。

[原文] You can start counting.

[译文] [SN]: 你可以开始数。

[原文] We'll let you start counting.

[译文] 我们会让你开始数。

[原文] How many cells are there in this building?

[译文] [Q]: 这栋楼里有多少个单元（cells）?

[原文] That part also I can't tell you.

[译文] [SN]: 那个部分我也不能告诉你。

[原文] Well, division is easy, right?

[译文] [Q]: 嗯，除法很容易，对吧?

[原文] My God, it's kind of loud.

[译文] 天哪，这里有点吵。

[原文] Are you looking at this like, "Now I see where my money is going."

[译文] 您是不是在看着它，心里想：“现在我看到我的钱花到哪里去了”。

[原文] It's like, "I run a software company. Welcome to the software company."

[译文] [SN]: 这就像是：“我经营一家软件公司。欢迎来到软件公司”。

[原文] How big is the design space once

[译文] 一旦决定使用 GB200s 和 NVLink，设计空间有多大

[原文] you've decided to use the GB200s and the NVLink?

[译文] 决定使用 GB200s 和 NVLink 之后?

[原文] How many other decisions are there to be made?

[译文] 还有多少其他决策需要制定?

[原文] There is coupling from the model architecture to what is the physical plan that's optimized.

[译文] [SN]: 模型架构与优化的物理规划之间存在耦合关系。

[原文] And it's also scary in that sense, which is, there's going to be a new chip that'll come out.

[译文] 从这个意义上说，这也很可怕，因为会有新的芯片问世。

[原文] Take Vera Rubin Ultra. That's going to have power density that's going to be so different, with cooling requirements that are going to be so different.

[译文] 拿 Vera Rubin Ultra 来说。它的功率密度（power density）将截然不同，冷却要求也将截然不同。

[原文] So you kind of don't want to just build all to one spec.

[译文] 所以你不会希望所有构建都只遵循一种规格（spec）。

[原文] That goes back a little bit to the dialogue we'll have,

[译文] 这又回到了我们即将进行的对话，即

[原文] which is that you want to be scaling in time as opposed to scale once and then be stuck with it.

[译文] 你希望是及时扩展（scaling in time），而不是一次性扩展，然后就被它困住。

End of Content

章节 3：技术革命的加速与 AI 的定位

📝 本节摘要：

Satya 将人工智能与其他技术变革（如铁路、互联网、工业化、云计算）进行比较，指出每一场革命从发现到普及的速度都在加快。尽管有些人认为 AI 是“最后的”技术革命，市场增长速度史无前例，但 Satya 保持“初期阶段”（early innings）的清醒认识。他引用图灵奖得主 Raj Reddy 的比喻，将 AI 定位为“守护天使”或“认知放大器”，视其为一种强大的工具，以衡量其对人类的效用。

Start of Content

[原文] When you look at all the past technological transitions—whether it be railroads or the Internet or replaceable parts, industrialization, the cloud, all of these things—each revolution has gotten much faster in the time it goes from technology discovered to ramp and pervasiveness through the economy.

[译文] [Q]: 当您回顾过去所有的技术转型——无论是铁路、互联网、可替换零件、工业化，还是云计算，所有这些——每一场革命从技术被发现到普及渗透到整个经济所需的时间都越来越快。

[原文] Many folks who have been on Dwarkesh's podcast believe this is the final technological revolution or transition,

[译文] 许多参加过 Dwarkesh 播客的人相信这是最后一场技术革命或转型，

[原文] and that this time is very, very different.

[译文] 并且这次与以往非常、非常不同。

[原文] At least so far in the markets, in three years we've already skyrocketed to hyperscalers doing $500 billion of capex next year, which is a scale that's unmatched to prior revolutions in terms of speed.

[译文] 至少到目前为止，在市场上，我们在三年内已经迅速发展到超大规模服务商（hyperscalers）明年将投入 5000 亿美元资本支出（capex）的程度，就速度而言，这是以往历次革命所无法比拟的规模。

[原文] The end state seems to be quite different.

[译文] 最终状态似乎截然不同。

[原文] Your framing of this seems quite different from what I would call the "AI bro" who's like, "AGI is coming."

[译文] 您对这件事的看法似乎与我称之为“AI 兄弟”的那类人——他们会说“通用人工智能（AGI）即将到来”——的看法截然不同。

[原文] I'd like to understand that more.

[译文] 我想更多地了解这一点。

[原文] I start with the excitement that I also feel for the idea that maybe after the Industrial Revolution this is the

[译文] [SN]: 我首先也有这种兴奋感，即认为这可能是继工业革命之后的

[原文] biggest thing. I start with that premise.

[译文] 最重大的事情。我是从这个前提开始的。

[原文] But at the same time, I'm a little grounded in the fact that this is still early innings.

[译文] 但与此同时，我也稍微脚踏实地，认识到这仍处于早期阶段（early innings）。

[原文] We've built some very useful things, we're seeing some great properties, these scaling laws seem to be working.

[译文] 我们已经构建了一些非常有用的东西，我们看到了一些出色的特性，这些扩展定律（scaling laws）似乎正在发挥作用。

[原文] I'm optimistic that they'll continue to work.

[译文] 我乐观地认为它们将继续有效。

[原文] Some of it does require real science breakthroughs, but it's also a lot of engineering and what have you.

[译文] 其中一些确实需要真正的科学突破，但同时它也涉及大量的工程技术等等。

[原文] That said, I also sort of take the view that even what has been happening in the last 70 years of computing has also been a march that has helped us move.

[译文] 话虽如此，我也认为，即使是过去 70 年计算领域发生的事情，也是一场有助于我们前进的征程。

[原文] I like one of the things that Raj Reddy has as a metaphor for what AI is.

[译文] 我喜欢拉杰·雷迪（Raj Reddy）对人工智能（AI）的比喻。

[原文] He's a Turing Award winner at CMU. He had this, even pre-AGI. He had this metaphor for AI, it should either be a guardian angel or a cognitive amplifier.

[译文] 他是卡内基梅隆大学（CMU）的图灵奖得主。即使在通用人工智能（AGI）出现之前，他就有这个比喻：人工智能应该要么是守护天使（guardian angel），要么是认知放大器（cognitive amplifier）。

[原文] I love that. It's a simple way to think about what this is.

[译文] 我喜欢这个说法。这是思考它本质的简单方式。

[原文] Ultimately, what is its human utility?

[译文] 最终，它对人类的效用（human utility）是什么?

[原文] It is going to be a cognitive amplifier and a guardian angel.

[译文] 它将是一个认知放大器和守护天使。

[原文] If I view it that way, I view it as a tool.

[译文] 如果我这样看待它，我就把它视为一个工具。

[原文] But then you can also go very mystical about it and say this is more than a tool.

[译文] 但你也可以对此非常神秘化，说这不仅仅是一个工具。

[原文] It does all these things, which only

[译文] 它做了所有这些事情，而这些事情直到目前为止，只有

[原文] humans did before so far.

[译文] 人类才能做到。

[原文] But that has been the case with many technologies in the past.

[译文] 但过去许多技术都是如此。

[原文] Only humans did a lot of things, and then we had tools that did them.

[译文] 只有人类才能做很多事情，然后我们有了可以完成这些事情的工具。

End of Content

章节 4：经济价值、生产力与 Satya Token

📝 本节摘要：

访谈者提出了一个关于经济价值的抽象问题：当机器开始生成有价值的“Satya tokens”（代表高度价值的输出）时，如何衡量其经济价值和利润流向？Satya 认为，真正的挑战在于经济增长。他指出，如同工业革命一样，即使技术快速普及，真正的经济增长也需要时间，因为它必须深入到工作流程（workflow）和工作成果（work artifact）的改变中。AI 的出现将使人类及其产生的“token”获得更高的杠杆作用，实现规模的 10 倍增长，并可能将工业革命 200 年的扩散时间压缩至 20-25 年。

Start of Content

[原文] We don't have to get wrapped up in the definition here, but one way to think about it is, maybe it takes five years, ten years, twenty years.

[译文] [Q]: 我们不必纠结于这里的定义，但可以这样想：也许需要五年、十年、二十年。

[原文] At some point, eventually a machine is producing Satya tokens, and the Microsoft board thinks that Satya tokens are worth a lot.

[译文] 在某个时间点，最终一台机器正在生产“萨提亚代币”（Satya tokens），而微软董事会认为这些萨提亚代币价值连城。

[原文] How much are you wasting of this economic value by interviewing Satya?

[译文] 采访萨提亚，您浪费了多少这种经济价值?

[原文] I could not afford the API costs of Satya tokens.

[译文] 我负担不起萨提亚代币的应用程序编程接口（API）费用。

[原文] Whatever you want to call it, are the Satya tokens a tool or an agent, whatever.

[译文] 无论你想怎么称呼它，萨提亚代币是一个工具还是一个代理（agent），都无所谓。

[原文] Right now, if you have models that cost on the order of dollars or cents per million tokens, there's just an enormous room for margin expansion there, where a million tokens of Satya are worth a lot.

[译文] 现在，如果你拥有的模型每百万个代币的成本在几美元或几美分量级，那么这里就有巨大的利润扩张空间，而一百万个萨提亚代币价值巨大。

[原文] Where does that margin go and what level of that margin is Microsoft involved in is the question I have.

[译文] 我想知道的是，这些利润流向哪里，微软参与了其中多大比例的利润。

[原文] In some sense this goes back again to, essentially, what's the economic growth picture going to really look like?

[译文] [SN]: 从某种意义上说，这又回到了本质上关于经济增长前景将真正是什么样的问题。

[原文] What's the firm going to look like?

[译文] 公司的形态将是什么样?

[原文] What's productivity going to look like?

[译文] 生产力将是什么样?

[原文] That to me is where, again, if the Industrial Revolution created… After 70 years of diffusion is when you started seeing the economic growth.

[译文] 对我来说，这又回到了工业革命所创造的……是在扩散（diffusion）了 70 年之后，你才开始看到经济增长。

[原文] That's the other thing to remember.

[译文] 这是另一件需要记住的事情。

[原文] Even if the tech is diffusing fast this time around, for true economic growth to appear it has to diffuse to a point where the work, the work artifact, and the workflow has to change.

[译文] 即使这次技术扩散速度很快，要出现真正的经济增长，它必须扩散到工作、工作成果（work artifact）和工作流程（workflow）都必须发生变化的程度。

[原文] So that's one place where I think the change management required for a corporation to truly change is something we shouldn't discount.

[译文] 所以我认为，一家公司要实现真正的变革所需要的变革管理（change management），是我们不应该低估的。

[原文] Going forward, do humans and the tokens they produce get higher leverage, whether it's the Dwarkesh or the Dylan tokens of the future?

[译文] [Q]: 展望未来，人类以及他们生产的代币——无论是未来的 Dwarkesh 代币还是 Dylan 代币——是否会获得更高的杠杆作用（leverage）?

[原文] Think about the amount of technology.

[译文] [SN]: 想想技术的数量。

[原文] Would you be able to run SemiAnalysis or this podcast without technology?

[译文] 如果没有技术，你能够运营 SemiAnalysis 或这个播客吗?

[原文] No chance, at the scale that you have been able to achieve, there’s no chance.

[译文] [Q]: 没有机会，以你目前达到的规模来看，没有机会。

[原文] So the question is, what's that scale?

[译文] [SN]: 所以问题是，这个规模是什么?

[原文] Is it going to be 10x'ed with something that comes through? Absolutely.

[译文] 它会被某些到来的东西提高 10 倍吗（10x'ed）？绝对会。

[原文] Therefore, whether you're ramped to some revenue number

[译文] 因此，无论你是否达到了某个收入数字

[原文] or you're ramped to some audience number or what have you, that I think is what's going to happen.

[译文] 或者达到了某个受众数字或其他什么，我认为这才是将要发生的事情。

[原文] The point is, what took 70 years, maybe 150 years for the Industrial Revolution, may happen in 20 years, 25 years.

[译文] 关键是，工业革命花费了 70 年，也许是 150 年的事情，可能在 20 年、25 年内发生。

[原文] I would love to compress what happened in 200 years of the Industrial Revolution into a 20-year period, if we're lucky.

[译文] 如果我们幸运的话，我希望将工业革命 200 年内发生的事情压缩到 20 年内。

End of Content

章节 5：AI 时代 SaaS 业务模式的挑战与市场扩张

📝 本节摘要：

访谈者提出，AI 时代的高昂商品成本（COGS）正在颠覆低增量成本的软件即服务（SaaS）商业模式。作为最大的 SaaS 公司之一，微软如何应对这一挑战？Satya 认为，业务模式的杠杆（如订阅、交易、消费）将保持相似，关键在于定价。他以 Office 365 向云端迁移为例，解释虽然云服务带来了商品成本（COGS），但它极大地扩张了市场（例如，印度的小客户也能负担碎片化的服务器成本），因此整体上抵消了成本压力，反而更具盈利性。

Start of Content

[原文] Microsoft historically has been perhaps the greatest software company, the largest software-as-a-service company.

[译文] [Q]: 微软历来也许是最大的软件公司，最大的软件即服务（software-as-a-service, SaaS）公司。

[原文] You've gone through a transition in the past

[译文] 您过去经历了一次转型

[原文] where you used to sell Windows licenses and disks of Windows or Microsoft, and now you sell subscriptions to 365.

[译文] 在这次转型中，您过去出售 Windows 许可证（licenses）和 Windows 或 Microsoft 的光盘，而现在您出售 365 的订阅服务。

[原文] As we go from that transition to where your business is today, there's also a transition going on after that.

[译文] 当我们从那次转型走向您今天的业务时，在那之后还有一次转型正在发生。

[原文] Software-as-a-service has incredibly low incremental cost per user.

[译文] 软件即服务（SaaS）的每用户增量成本（incremental cost per user）极低。

[原文] There's a lot of R&D, there's a lot of customer acquisition costs.

[译文] 存在大量的研发（R&D）投入，存在大量的客户获取成本。

[原文] This is sort of why, not Microsoft, but the SaaS companies have underperformed massively in the markets, because the COGS of AI is just so high, and that just completely

[译文] 这就是为什么，不是微软，而是许多 SaaS 公司在市场上表现严重不佳的原因，因为人工智能的商品成本（COGS）太高了，这完全

[原文] breaks how these business models work.

[译文] 打破了这些商业模式的运作方式。

[原文] How do you, as perhaps the greatest software-as-a-service company, transition Microsoft to this new age where COGS matters a lot and the incremental cost per user is different?

[译文] 作为也许是最伟大的软件即服务（SaaS）公司，您如何将微软转型到这个商品成本（COGS）非常重要、并且每用户增量成本不同的新时代?

[原文] Because right now you're charging like, "Hey, it's 20 bucks for Copilot."

[译文] 因为现在您的收费方式像是：“嘿，Copilot 收 20 美元”。

[原文] It's a great question because in some sense with the business models themselves, the levers are going to remain similar.

[译文] [SN]: 这是一个很好的问题，因为从某种意义上说，商业模式本身的杠杆（levers）将保持相似。

[原文] If you look at the menu of models starting from consumer all the way, there will be some ad unit,

[译文] 如果你看看从消费者到全范围的模型菜单，将会有一些广告单元（ad unit），

[原文] there will be some transaction, there will be some device gross margin for somebody who builds an AI device.

[译文] 会有一些交易（transaction），对于制造人工智能设备的人来说，会产生一些设备毛利率（device gross margin）。

[原文] There will be subscriptions, consumer and enterprise, and then there'll be consumption.

[译文] 会有订阅（subscriptions），针对消费者和企业，然后会有消费（consumption）。

[原文] So I still think those are all the meters.

[译文] 所以我仍然认为这些都是衡量标准（meters）。

[原文] To your point, what is a subscription?

[译文] 针对你的观点，订阅是什么?

[原文] Up to now, people like subscriptions because they can budget for them.

[译文] 截至目前，人们喜欢订阅是因为他们可以为此做预算。

[原文] They are essentially entitlements to some consumption rights that come encapsulated in a subscription.

[译文] 它们本质上是对包含在订阅中的某些消费权利的授权（entitlements）。

[原文] So I think that in some sense becomes a pricing decision.

[译文] 所以我认为从某种意义上说，这变成了一个定价决策。

[原文] How much consumption you are entitled to is, if you look at all the coding subscriptions, kind of what they are, right?

[译文] 你被授权消费多少，如果你看看所有编码订阅，它们大致就是这样，对吧?

[原文] Then you have the pro tier, the standard tier, and what have you.

[译文] 然后你有专业层级（pro tier），标准层级（standard tier），等等。

[原文] So I think that's how the pricing and the margin structures will get tiered.

[译文] 所以我认为定价和利润结构将由此分级。

[原文] The interesting thing is that at Microsoft, the good news for us is we are in that business across all those meters.

[译文] 有趣的是，对于微软来说，好消息是我们在所有这些衡量标准上都涉及该业务。

[原文] At a portfolio level, we pretty much have consumption, subscriptions, to all of the other consumer levers as well.

[译文] 在产品组合层面，我们几乎拥有消费、订阅，以及所有其他面向消费者的杠杆。

[原文] I think time will tell which of these

[译文] 我认为时间会告诉我们，这些模式中哪一种

[原文] models make sense in what categories.

[译文] 在哪些类别中是有意义的。

[原文] One thing on the SaaS side, since you brought it up, which I think a lot about.

[译文] 既然你提到了 SaaS 方面，有一点我思考了很多。

[原文] Take Office 365 or Microsoft 365.

[译文] 以 Office 365 或 Microsoft 365 为例。

[原文] Having a low ARPU is great, because here's an interesting thing.

[译文] 拥有较低的每用户平均收入（ARPU）是很棒的，因为有一个有趣的事情。

[原文] During the transition from server to cloud, one of the questions we used to ask ourselves is, "Oh my God, if all we did was just basically move the same users who were using our Office licenses and our Office servers at the time to the cloud, and we had COGS, this is going to not only shrink our margins but we'll be fundamentally a less profitable company."

[译文] 在从服务器到云的转型过程中，我们过去经常问自己一个问题：“天哪，如果我们所做的只是把当时使用我们的 Office 许可证和 Office 服务器的同一批用户转移到云上，并且我们有了商品成本（COGS），这不仅会压缩我们的利润率，还会使我们从根本上成为一家盈利能力较差的公司”。

[原文] Except what happened was the move to the cloud expanded the market like crazy.

[译文] 然而发生的情况是，向云的迁移疯狂地扩大了市场。

[原文] We sold a few servers in India, we didn't sell much.

[译文] 我们在印度只卖了几台服务器，卖得不多。

[原文] Whereas in the cloud suddenly everybody in India also could afford fractionally buying servers, the IT cost.

[译文] 然而在云中，突然间印度的每个人也能负担得起碎片化地购买服务器，即信息技术（IT）成本。

[原文] In fact, the biggest thing I had not realized, for example, was the amount of money people were spending buying storage underneath SharePoint.

[译文] 事实上，我当时没有意识到的一件最大的事情是，例如，人们花钱购买 SharePoint 下面的存储（storage）。

[原文] In fact, EMC's biggest segment may have been storage servers for SharePoint.

[译文] 事实上，EMC 最大的业务部门可能就是 SharePoint 的存储服务器。

[原文] All that sort of dropped in the cloud because nobody had to go buy.

[译文] 所有这些在云中都消失了，因为没有人需要去购买。

[原文] In fact, it was working capital, meaning basically, it was cash flow out.

[译文] 事实上，那是营运资本（working capital），基本上意味着现金流出。

[原文] So it expanded the market massively.

[译文] 所以它大规模地扩大了市场。

End of Content

章节 6：AI 编码市场的爆炸式增长与竞争态势

📝 本节摘要：

Satya 认为，AI 带来的市场扩张效应将复制云计算的成功。他以编码领域为例，指出 Microsoft 历经数十年构建的 GitHub 和 VS Code，其编码助手（Copilot）在短短一年内就达到了巨大规模。尽管 GitHub Copilot 曾占据近乎垄断地位，但现在面临 Claude Code、Cursor 等新竞争者。Satya 欢迎这种竞争，认为新竞争者的出现（而不是过时的公司如 Borland）是市场正在扩张的最佳信号，并预言编码 AI 领域将成为比知识工作（knowledge work）更大的类别，即“软件工厂”类别。

Start of Content

[原文] This AI thing will be that.

[译文] [SN]: 人工智能这件事也将是如此。

[原文] If you take coding, what we built with GitHub and VS Code over decades, suddenly the coding assistant is that big in one year.

[译文] 如果你看看编码（coding）领域，我们用 GitHub 和 VS Code 几十年来建立的基础，突然之间，编码助手在一年内就发展到了如此巨大的规模。

[原文] That I think is what's going to happen as well, which is the market expands massively.

[译文] 我认为这也将发生，即市场会大规模扩张。

[原文] There’s a question of, the market will expand, but will the parts of the revenue that touch Microsoft expand?

[译文] [Q]: 问题是，市场会扩张，但与微软相关的收入部分也会扩张吗?

[原文] Copilot is an example.

[译文] Copilot 就是一个例子。

[原文] If you look earlier this year, according to Dylan's numbers, GitHub Copilot revenue

[译文] 如果你看今年早些时候，根据迪伦（Dylan）的数据，GitHub Copilot 的收入

[原文] was like $500 million or something like that and there were no close competitors.

[译文] 大约是 5 亿美元左右，并且没有势均力敌的竞争对手。

[原文] Whereas now you have Claude Code, Cursor, and Copilot with around similar revenue, around a billion.

[译文] 然而现在你有 Claude Code、Cursor 和 Copilot，它们的收入都在 10 亿美元左右，相差不大。

[原文] Codex is catching up around $700–800 million.

[译文] Codex 正在迎头赶上，收入约为 7 亿至 8 亿美元。

[原文] So the question is, across all the surfaces that Microsoft has access to, what is the advantage that Microsoft's equivalents of Copilot have?

[译文] 那么问题是，在微软所能接触到的所有领域中，微软的 Copilot 等价物具有什么优势?

[原文] By the way, I love this chart.

[译文] [SN]: 顺便说一句，我喜欢这张图表。

[原文] I love this chart for so many reasons.

[译文] 我喜欢这张图表，原因有很多。

[原文] One is we're still on the top.

[译文] 第一，我们仍然位居榜首。

[原文] Second is all these companies that are listed here are all companies that have been born in

[译文] 第二是这里列出的所有这些公司都是在过去

[原文] the last four or five years.

[译文] 四五年内诞生的公司。

[原文] That to me is the best sign.

[译文] 对我来说，这是最好的迹象。

[原文] You have new competitors, new existential problems.

[译文] 你有了新的竞争对手，新的生存问题。

[原文] When you say, who's it now?

[译文] 当你说，现在是谁?

[原文] Claude's going to kill you, Cursor is going to kill you, it's not Borland.

[译文] 是 Claude 要打败你，Cursor 要打败你，而不是 Borland（一家过时的软件公司）。

[原文] Thank God. That means we are in the right direction.

[译文] 谢天谢地。这意味着我们的方向是正确的。

[原文] This is it. The fact that we went from nothing to this scale is the market expansion.

[译文] 就是这样。我们从零达到这个规模，这就是市场扩张。

[原文] This is like the cloud-like stuff.

[译文] 这就像云计算（cloud-like）的东西。

[原文] Fundamentally, this category of coding and AI is probably going to be one of the biggest categories.

[译文] 从根本上说，编码和人工智能这个类别可能会成为最大的类别之一。

[原文] It is the software factory category.

[译文] 它是软件工厂（software factory）类别。

[原文] In fact, it may be bigger than knowledge work.

[译文] 事实上，它可能比知识工作（knowledge work）更大。

End of Content

章节 7：Agent HQ：利用 GitHub 平台构建代理控制中心

📝 本节摘要：

Satya 承认竞争是艰难的，但微软已经成功利用了其现有优势（如 GitHub）进入 AI 编码市场。尽管 Copilot 竞争激烈，GitHub 平台本身却在持续增长。微软的竞争策略是利用 GitHub 的基础功能（Git、issues、actions）并推出新概念“Agent HQ”（代理总部），其中包含一个“Mission Control”（任务控制）系统。这个系统能让用户通过一个订阅管理和指挥多个 AI 代理（包括 Codex、Claude、Grok 等），这些代理在独立分支中工作，从而提供必要的控制平面（control plane）和可观测性（observability），解决多智能体协作问题。

Start of Content

[原文] I want to keep myself open-minded about it.

[译文] [SN]: 我想对此保持开放的心态。

[原文] We're going to have tough competition.

[译文] 我们将面临激烈的竞争。

[原文] That's your point, which is a great one.

[译文] 你的观点很好。

[原文] But I'm glad we have parlayed what we had into this and now we have to compete.

[译文] 但我很高兴我们已经将已有的优势转化（parlayed）到这个领域，现在我们必须竞争。

[原文] On the competing side, even in the last quarter we just finished, we did our quarterly announcement and I think we grew from 20 to 26 million subs.

[译文] 在竞争方面，即使在我们刚刚结束的上一季度，我们发布了季度公告，我认为我们的订阅（subs）用户从 2000 万增长到了 2600 万。

[原文] I feel good about our sub growth and where the direction of travel on that is.

[译文] 我对我们的订阅增长及其发展方向感觉良好。

[原文] But the more interesting thing that has happened is, guess where all the repos of all these other

[译文] 但更有趣的事情是，猜猜所有这些其他竞争对手的所有代码仓库（repos）都去了哪里

[原文] guys who are generating lots and lots of code go?

[译文] 那些正在生成大量代码的公司都去了哪里?

[原文] They go to GitHub. GitHub is at an all-time high in terms of repo creation, PRs, everything.

[译文] 它们都去了 GitHub。GitHub 在代码仓库创建、拉取请求（PRs）等方面都处于历史最高水平。

[原文] In some sense we want to keep that open, by the way.

[译文] 顺便说一句，从某种意义上说，我们希望保持这种开放性。

[原文] That means we want to have that.

[译文] 这意味着我们想要拥有这个。

[原文] We don't want to conflate that with our own growth.

[译文] 我们不想将这一点与我们自己的增长混为一谈。

[原文] Interestingly enough, we are getting one developer joining GitHub a second or something, that is the stat, I think.

[译文] 有趣的是，我认为我们的统计数据显示，每秒钟大约有一名开发人员加入 GitHub。

[原文] And 80% of them just fall into some GitHub Copilot workflow, just because there are.

[译文] 其中 80% 的人只是因为存在 GitHub Copilot 工作流程而加入。

[原文] By the way, many of these things will even

[译文] 顺便说一句，其中许多东西甚至会

[原文] use some of our coding code review agents, which are by default on, just because you can use it.

[译文] 使用我们默认开启的一些编码代码审查代理（agents），只因为你可以使用它们。

[原文] We'll have many, many structural shots at this.

[译文] 我们将有很多很多结构性的机会。

[原文] The thing that we're also going to do is what we did with Git.

[译文] 我们还将做的一件事，就像我们对 Git 所做的那样。

[原文] The primitives of GitHub, starting with Git, to issues, to actions, these are powerful, lovely things because they kind of are all built around your repo.

[译文] GitHub 的基本要素（primitives），从 Git 开始，到问题（issues），再到操作（actions），这些都是强大而美好的东西，因为它们都围绕着你的代码仓库（repo）构建。

[原文] We want to extend that. Last week at GitHub Universe, that's kind of what we did.

[译文] 我们想要扩展它。在上周的 GitHub Universe 大会上，我们就是这样做的。

[原文] We said Agent HQ was the conceptual thing that we said we're going to build out.

[译文] 我们说“代理总部”（Agent HQ）是我们计划构建的概念性事物。

[原文] This is where, for example, you have a thing called Mission Control.

[译文] 例如，在这里，你有一个被称为“任务控制”（Mission Control）的东西。

[原文] You go to Mission Control, and now I can fire off.

[译文] 你去到任务控制（Mission Control），现在我就可以启动（fire off）。

[原文] Sometimes I describe it as the cable TV of all these AI agents because I'll have, essentially packaged into one subscription, Codex, Claude, Cognition stuff, anyone's agents, Grok, all of them will be there.

[译文] 有时我将它描述为所有这些人工智能代理（AI agents）的有线电视（cable TV），因为本质上，我将把 Codex、Claude、Cognition 的东西、任何人的代理、Grok，所有这些都打包在一个订阅中。

[原文] So I get one package and then I can literally go issue a task and steer them so they'll all be working in their independent branches.

[译文] 这样我就获得了一个软件包，然后我就可以真正地发布任务并引导它们，让它们都在各自独立的（independent branches）分支中工作。

[原文] I can monitor them. I think that's going to be one of

[译文] 我可以监控它们。我认为这将是

[原文] the biggest places of innovation, because right now I want to be able to use multiple agents.

[译文] 最大的创新点之一，因为现在我希望能够使用多个代理。

[原文] I want to be able to then digest the output of the multiple agents.

[译文] 我希望能够消化多个代理的输出。

[原文] I want to be able to then keep a handle on my repo.

[译文] 我希望能够掌控我的代码仓库（repo）。

[原文] If there's some kind of a heads-up display that needs to be built and then for me to quickly steer and triage what the coding agents have generated, that to me, between VS Code, GitHub, and all of these new primitives we'll build as Mission Control with a control plane.

[译文] 如果需要构建某种平视显示器（heads-up display），让我能够快速引导并分类处理（triage）这些编码代理生成的内容，对我来说，这将在 VS Code、GitHub 和我们作为任务控制（Mission Control）用控制平面（control plane）构建的所有这些新基本要素之间实现。

[原文] Observability… Just think about everyone who is going to deploy all this.

[译文] 可观测性（Observability）……想象一下，所有将部署这一切的人。

[原文] It will require a whole host of observability of what agent did what at what time to what code base.

[译文] 这将需要针对“哪个代理在什么时间对哪个代码库做了什么”的整套可观测性。

[原文] I feel that's the opportunity.

[译文] 我认为这就是机会。

End of Content

章节 8：市场份额下降与市场规模扩张的平衡

📝 本节摘要：

访谈者指出，虽然编码 AI 市场在一年内实现了 10 倍的增长，但微软的市场份额已从接近 100% 降至 25% 以下。Satya 认为，关键在于市场规模的扩张：拥有一个巨大市场中的可观份额，比拥有一个小市场中的高份额更有价值。他以云计算转型为例，证明微软即使在新的超大规模（hyperscale）市场中份额低于客户端-服务器时代，但业务规模仍以数量级增长。他强调，只要市场在创造更多价值，存在多个赢家，微软通过 Agent HQ 等创新就能够保持竞争力。

Start of Content

[原文] At the end of the day your point is well taken, which is we better be competitive and innovate.

[译文] [SN]: 归根结底，你的观点是正确的，那就是我们最好保持竞争力和创新。

[原文] If we don't, we will get toppled.

[译文] 如果不这样做，我们就会被推翻。

[原文] But I like the chart, at least as long as we're on the top, even with competition.

[译文] 但我喜欢这张图表，至少只要我们保持领先地位，即使存在竞争。

[原文] The key point here is sort of that GitHub will keep growing regardless of whose coding agent wins.

[译文] [Q]: 这里的关键点是，无论哪个编码代理获胜，GitHub 都会继续增长。

[原文] But that market only grows at say 10, 15, 20%, which is way above GDP.

[译文] 但那个市场可能只以 10%、15%、20% 的速度增长，这远高于国内生产总值（GDP）。

[原文] It's a great compounder.

[译文] 这是一个很好的复合增长点。

[原文] But these AI coding agents have

[译文] 但是这些人工智能编码代理（AI coding agents）已经

[原文] grown from say $500 million run rate at the end of last year—which was just GitHub Copilot—to now where the current run rate across GitHub Copilot, Claude Code, Cursor, Cognition, Windsurf, Replit, OpenAI Codex… That’s run rating at $5–6 billion now for the Q4 of this year.

[译文] 从去年底的 5 亿美元年化运行率（run rate）（当时只有 GitHub Copilot）增长到现在的 GitHub Copilot、Claude Code、Cursor、Cognition、Windsurf、Replit、OpenAI Codex 等的总运行率，在今年第四季度达到 50 亿至 60 亿美元。

[原文] That's 10x.

[译文] 这是 10 倍的增长。

[原文] When you look at the TAM of software agents, is it the $2 trillion of wages you pay people, or is it something beyond that?

[译文] 当您看待软件代理的潜在市场规模（TAM）时，它是因为您支付给人们的 2 万亿美元薪水，还是超越了这一点?

[原文] Because every company in the world will now be able to develop software more?

[译文] 因为世界上每家公司现在都能更多地开发软件?

[原文] No question Microsoft takes a slice of that.

[译文] 毫无疑问，微软会分得其中一部分。

[原文] But you've gone from near 100%, or certainly way above 50%, to sub-25% market share in just one year.

[译文] 但你们在仅仅一年内，市场份额就从接近 100%，或肯定远高于 50%，降至 25% 以下。

[原文] What is the confidence that people can get that Microsoft will keep winning?

[译文] 人们从哪里获得信心，认为微软会继续获胜?

[原文] It goes back a little bit, Dylan, to that there's no birthright here, that we should have any confidence other than to say we should go innovate.

[译文] [SN]: 迪伦（Dylan），这又回到了这里没有与生俱来的权利（birthright），我们不应该有任何信心，除非我们说我们应该去创新。

[原文] Knowing the lucky break we have, in some sense, is that this category is going to be a lot bigger than anything we had high share in.

[译文] 从某种意义上说，我们拥有的幸运机会在于，这个类别将比我们曾拥有高份额的任何领域都大得多。

[原文] Let me say it that way.

[译文] 让我这样说。

[原文] You could say we had high

[译文] 你可以说我们在

[原文] share in VS Code, we had high share in the repos with GitHub, and that was a good market.

[译文] VS Code 中拥有高份额，我们在 GitHub 的代码仓库中拥有高份额，那是一个不错的市场。

[原文] But the point is that even having a decent share in what is a much more expansive market…

[译文] 但关键是，即使在一个更加广阔的市场中拥有一个体面的份额……

[原文] You could say we had a high share in client-server server computing.

[译文] 你可以说我们在客户端-服务器（client-server）计算中拥有高份额。

[原文] We have much lower share than that in hyperscale.

[译文] 我们在超大规模（hyperscale）市场中的份额比那低得多。

[原文] But is it a much bigger business? By orders of magnitude.

[译文] 但它是一个更大的业务吗？是的，大了几个数量级（orders of magnitude）。

[原文] So at least it's existence proof that Microsoft has been okay even if our share position has not been as strong as it was, as long as the markets we are competing in are creating

[译文] 所以，这至少证明了微软是没问题的，即使我们的市场份额不如以前那么强劲，只要我们正在竞争的市场正在创造

[原文] more value.

[译文] 更多价值。

[原文] And there are multiple winners.

[译文] 而且存在多个赢家。

[原文] That's the stuff. But I take your point that ultimately it all means you have to get competitive.

[译文] 就是这样。但我接受你的观点，即最终这一切都意味着你必须变得有竞争力。

[原文] I watch that every quarter.

[译文] 我每个季度都会关注这一点。

[原文] That’s why I'm very optimistic about what we're going to do with Agent HQ, turning GitHub into a place where all these agents come.

[译文] 这就是为什么我对我们将用“代理总部”（Agent HQ）所做的事情，将 GitHub 变成所有这些代理聚集的地方，感到非常乐观。

[原文] As I said, we'll have multiple shots on goal on there.

[译文] 正如我所说，我们将有多次进球机会（multiple shots on goal）。

[原文] It need not be… Some of these guys can succeed along with us, so it doesn't need to be just one winner and one subscription.

[译文] 这不一定需要……其中一些公司可以和我们一起成功，所以不一定只有一位赢家和一个订阅。

End of Content

章节 9：价值分配：模型（Model）的商品化与支架（Scaffolding）的优势

📝 本节摘要：

访谈者提出了一个核心问题：随着 AI 模型越来越智能、越来越自主（autonomous），是否所有价值都会迁移到模型本身（由模型公司赚取所有利润），而作为用户界面（UI）或流程管理层的“支架”（scaffolding）将变得不那么重要？Satya 认为，时间会给出答案，但他的观点是，激励结构最终会清晰化。他提出了“模型商品化”的论点，尤其考虑到开源模型和数据流动性（data liquidity）。他认为，如果微软赢得了处理智能问题的“锯齿状”（jaggedness）和“受阻”（hobbling）的支架层，他们就可以通过数据流动性实现向模型层的垂直整合，从而避免成为“赢家的诅咒”（winner's curse）。

Start of Content

[原文] I guess the reason to focus on this question is that it's not just about GitHub,

[译文] [Q]: 我想关注这个问题的原因是，它不仅关乎 GitHub，

[原文] but fundamentally about Office and all the other software that Microsoft offers.

[译文] 而是从根本上关乎 Office 和微软提供的所有其他软件。

[原文] One vision you could have about how AI proceeds is that the models are going to keep being hobbled and you'll need this direct visible observability all the time.

[译文] 关于人工智能将如何发展，一种愿景是模型将继续受到限制（hobbled），你需要一直保持这种直接可见的可观测性。

[原文] Another vision is that over time these models which are now doing tasks that take two minutes, in the future, they'll be doing tasks that take 10, 30 minutes.

[译文] 另一种愿景是，随着时间的推移，这些模型现在完成需要两分钟的任务，未来将完成需要 10 分钟、30 分钟的任务。

[原文] In the future, maybe they're doing days worth of work autonomously.

[译文] 未来，它们或许能自主地（autonomously）完成几天的工作量。

[原文] Then the model companies are charging thousands of dollars maybe for access to, really,

[译文] 那么模型公司可能会收取数千美元的费用，用于访问，实际上是

[原文] a coworker which could use any UI to communicate with their human and migrate between platforms.

[译文] 一个可以使用任何用户界面（UI）与其人类交流并在不同平台之间迁移的同事。

[原文] If we’re getting closer to that, why aren't the model companies that are just getting more and more profitable, the ones that are taking all the margin?

[译文] 如果我们越来越接近那种情况，为什么那些利润越来越高的模型公司，不是那些攫取所有利润的公司?

[原文] Why is the place where the scaffolding happens, which becomes less and less relevant as the AI becomes more capable, going to be that important?

[译文] 为什么发生支架（scaffolding）的场所（随着人工智能能力增强，它将变得越来越不重要）会如此重要?

[原文] That goes to Office as it exists now versus coworkers that are just doing knowledge work.

[译文] 这关系到目前存在的 Office 与那些仅从事知识工作的同事之间的关系。

[原文] That's a great point. Does all the value

[译文] [SN]: 这是一个很好的观点。是否所有价值

[原文] migrate just to the model?

[译文] 都只迁移到了模型（model）?

[原文] Or does it get split between the scaffolding and the model?

[译文] 还是在支架（scaffolding）和模型之间分配?

[原文] I think that time will tell.

[译文] 我认为时间会证明一切。

[原文] But my fundamental point also is that the incentive structure gets clear.

[译文] 但我的基本观点是激励结构会变得清晰。

[原文] Let’s take information work, or take even coding.

[译文] 让我们以信息工作，或者甚至是编码为例。

[原文] Already in fact, one of my favorite settings in GitHub Copilot is called auto, which will just optimize.

[译文] 事实上，在 GitHub Copilot 中，我最喜欢的设置之一就是名为“自动”（auto）的设置，它将自动优化。

[原文] In fact I buy a subscription and the auto one will start picking and optimizing for what I am asking it to do.

[译文] 事实上，我购买一个订阅，这个“自动”功能就会开始选择和优化我要求它做的事情。

[原文] It could even be fully autonomous.

[译文] 它甚至可以是完全自主的（fully autonomous）。

[原文] It could arbitrage the tokens available across multiple models to go get a task done.

[译文] 它可以套利（arbitrage）于多个模型之间可用的代币，来完成一项任务。

[原文] If you take that argument, the commodity there will be models.

[译文] 如果你接受这个论点，那么那里的商品（commodity）将是模型。

[原文] Especially with open source models, you can pick a checkpoint and you can take a bunch of your data and you're seeing it.

[译文] 特别是对于开源模型，你可以选择一个检查点（checkpoint），你可以使用你的大量数据，你正在看到这一点。

[原文] I think all of us will start, whether it's from Cursor or from Microsoft, seeing some in-house models even.

[译文] 我认为我们所有人，无论是 Cursor 还是微软，甚至都会开始看到一些内部（in-house）模型。

[原文] And then you'll offload most of your tasks to it.

[译文] 然后你将把大部分任务卸载给它。

[原文] So one argument is if you win the scaffolding—which today is dealing with all the hobbling problems or the jaggedness of these intelligence problems, which you kind of have to—if you win that,

[译文] 所以一种论点是，如果你赢得了支架层（scaffolding）——它目前正在处理所有这些智能问题的“受阻”（hobbling problems）或“锯齿状”（jaggedness）问题，而你不得不处理这些问题——如果你赢得了它，

[原文] then you will vertically integrate yourself into the model just because you will have the liquidity of the data and what have you.

[译文] 那么你就会垂直整合（vertically integrate）到模型中，仅仅因为你拥有数据的流动性（liquidity of the data）等等。

[原文] There are enough and more checkpoints that are going to be available.

[译文] 将会有足够多可用的检查点。

[原文] That's the other thing. Structurally, I think there will always be an open source model that will be fairly capable in the world that you could then use, as long as you have something that you can use that with, which is data and a scaffolding.

[译en] 这是一个结构性的问题。我认为世界上将永远存在一个相当有能力的开源模型，只要你拥有可以与之配合使用的东西——数据和支架。

[原文] I can make the argument that if you're a model company, you may have a winner's curse.

[译文] 我可以提出这样的论点：如果你是一家模型公司，你可能会遭受“赢家的诅咒”（winner's curse）。

[原文] You may have done all the hard work, done unbelievable innovation, except it's one copy away from that being commoditized.

[译文] 你可能做了所有艰苦的工作，完成了令人难以置信的创新，但它距离被商品化（commoditized）只差一个复制（one copy away）的距离。

[原文] Then the person who has the data for grounding and context engineering, and the liquidity of data can then go take that checkpoint and train it.

[译文] 然后，拥有用于基础（grounding）和上下文工程（context engineering）的数据以及数据流动性的人，就可以拿走那个检查点并进行训练。

[原文] So I think the argument can be made both ways.

[译文] 所以我认为这个论点可以从两个方面来看待。

End of Content

章节 10：模型层的利润扩张与微软的全栈混合策略

📝 本节摘要：

访谈者引用 Anthropic 的例子，指出模型公司的毛利率（gross margins）在提高，这似乎反驳了模型商品化的论点。Satya 承认这一现实，但他解释了微软如何通过将模型深度整合到应用中来创造价值，而非仅仅是用户界面（UI）包装。他以 Excel Agent 为例，指出微软利用 GPT 家族的知识产权（IP），将模型置于中间层（middle tier），使其“原生理解”Excel 的工具和逻辑。Satya 总结了微软的混合战略：在超大规模层支持多个模型（包括开源、OpenAI），利用 OpenAI IP 进行七年创新，同时开发自有的 MAI 模型，并在所有关键业务（如安全、编码）中构建“模型优先”（model-forward）的应用支架，确保在竞争环境中保持优势。

Start of Content

[原文] Unpacking what you said, there's two views of the world.

[译文] [Q]: 剖析您的说法，这里存在两种世界观。

[原文] One is that there are so many different models out there. Open source exists. There will be differences between the models that

[译文] 一是市面上有如此多不同的模型。开源存在。模型之间将存在的差异

[原文] will drive some level of who wins and who doesn't.

[译文] 将在一定程度上决定谁输谁赢。

[原文] But the scaffolding is what enables you to win.

[译文] 但支架（scaffolding）才是让你获胜的关键。

[原文] The other view is that, actually, models are the key IP.

[译文] 另一种观点是，实际上，模型才是关键知识产权（IP）。

[原文] And everyone's in a tight race and there's some, "Hey, I can use Anthropic or OpenAI."

[译文] 每个人都在激烈竞争，有人会说：“嘿，我可以使用 Anthropic 或 OpenAI”。

[原文] You can see this in the revenue charts.

[译文] 你可以在收入图表中看到这一点。

[原文] OpenAI's revenue started skyrocketing once they finally had a code model with similar capabilities to Anthropic, although in different ways.

[译文] OpenAI 的收入在他们最终拥有一个与 Anthropic 能力相似（尽管方式不同）的编码模型后，开始飙升。

[原文] There's the view that the model companies are the ones that garner all the margin.

[译文] 观点是模型公司才是获取所有利润（margin）的公司。

[原文] Because if you look across this year, at least at Anthropic, their gross margins on inference went from well below 40% to north of 60% by the end of the year.

[译文] 因为如果你看今年，至少在 Anthropic，他们的推理（inference）毛利率（gross margins）从远低于 40% 上升到年底的 60% 以上。

[原文] The margins are expanding there despite more Chinese open source models than ever.

[译文] 尽管中国的开源模型比以往任何时候都多，但那里的利润仍在扩张。

[原文] OpenAI is competitive, Google is competitive, X/Grok is now competitive.

[译文] OpenAI 具有竞争力，Google 具有竞争力，X/Grok 现在也具有竞争力。

[原文] All these companies are now competitive, and yet despite this, the margins have expanded at the model layer significantly.

[译文] 所有这些公司现在都具有竞争力，然而尽管如此，模型层面的利润率仍显著扩张。

[原文] How do you think about that?

[译文] 您如何看待这个问题?

[原文] It's a great question. Perhaps a few years ago people were saying, "Oh, I could just wrap a

[译文] [SN]: 这是一个很好的问题。或许几年前人们还在说：“哦，我可以直接封装一个

[原文] model and build a successful company."

[译文] 模型，然后建立一家成功的公司”。

[原文] That has probably gotten debunked just because of the model capabilities, and the tools used, in particular.

[译文] 这一点很可能已经被推翻了，特别是由于模型的能力以及所使用的工具。

[原文] But the interesting thing is, when I look at Office 365, let's take even this little thing we built called Excel Agent.

[译文] 但有趣的是，当我看到 Office 365 时，我们来看一看我们构建的这个名为 Excel Agent（Excel 代理）的小东西。

[原文] It's interesting.

[译文] 这很有趣。

[原文] Excel Agent is not a UI-level wrapper.

[译文] Excel Agent 不是一个用户界面（UI）层面的包装器。

[原文] It's actually a model that is in the middle tier.

[译文] 它实际上是位于中间层（middle tier）的一个模型。

[原文] In this case, because we have all the IP from the GPT family, we are taking that and putting it into the core middle tier of

[译文] 在这种情况下，因为我们拥有来自 GPT 家族的所有知识产权（IP），我们正在将其放入 Office 系统的核心中间层

[原文] the Office system to teach it what it means to natively understand Excel, everything in it.

[译文] 以教会它如何原生理解 Excel 及其内部的一切。

[原文] It's not just, "Hey, I just have a pixel-level understanding."

[译文] 它不仅仅是：“嘿，我只是具备像素级理解”。

[原文] I have a full understanding of all the native artifacts of Excel.

[译文] 我对 Excel 的所有原生工件（native artifacts）都有完整的理解。

[原文] Because if you think about it, if I'm going to give it some reasoning task, I need to even fix the reasoning mistakes I make.

[译文] 因为如果你仔细想想，如果我要给它一些推理任务，我甚至需要修正我犯下的推理错误。

[原文] That means I need to not just see the pixels, I need to be able to see, "Oh, I got that formula wrong," and I need to understand that.

[译文] 这意味着我不仅需要看到像素，我还需要能够看到：“哦，我那个公式算错了”，并且我需要理解这一点。

[原文] To some degree, that's all being done not at the UI wrapper level with some prompt,

[译文] 在某种程度上，所有这些都不是通过某个提示（prompt）在用户界面（UI）包装器层面完成的，

[原文] but it's being done in the middle tier by teaching it all the tools of Excel.

[译文] 而是在中间层通过教会它 Excel 的所有工具来完成的。

[原文] I'm giving it essentially a markdown to teach it the skills of what it means to be a sophisticated Excel user.

[译文] 我基本上是给了它一个 Markdown（格式）来教它成为一个成熟的 Excel 用户所需的技能。

[原文] It's a weird thing that it goes back a little bit to the AI brain.

[译文] 这很奇怪，它有点回归到人工智能大脑（AI brain）的概念。

[原文] You're building not just Excel, business logic in its traditional sense.

[译文] 你构建的不仅仅是 Excel，不仅仅是其传统意义上的业务逻辑。

[原文] You're taking the Excel business logic in the traditional sense and wrapping essentially a cognitive layer to it, using this model which knows how to use the tool.

[译文] 你正在拿传统的 Excel 业务逻辑，并用这个知道如何使用工具的模型，给它包上一个认知层（cognitive layer）。

[原文] In some sense, Excel will come with an analyst

[译文] 从某种意义上说，Excel 将捆绑（bundled in）一名分析师

[原文] bundled in and with all the tools used.

[译文] 以及所有使用的工具。

[原文] That's the type of stuff that will get built by everybody.

[译文] 这就是每个人都会构建的那类东西。

[原文] So even for the model companies, they’ll have to compete.

[译文] 所以即使对于模型公司来说，他们也必须竞争。

[原文] If they price stuff high, guess what, if I'm a builder of a tool like this, I'll substitute you.

[译文] 如果他们定价很高，你猜怎么着，如果我是这种工具的构建者，我就会取代（substitute）你。

[原文] I may use you for a while.

[译文] 我可能会使用你一段时间。

[原文] So as long as there's competition…

[译文] 所以只要有竞争……

[原文] There's always a winner-take-all thing.

[译文] 总是存在赢家通吃（winner-take-all）的事情。

[原文] If there's going to be one model that is better than everybody else with massive distance, yes, that's a winner-take-all.

[译文] 如果有一个模型比其他所有模型都好，且优势巨大，那么，是的，那就是赢家通吃。

[原文] But as long as there's competition where there are

[译文] 但只要存在竞争，有

[原文] multiple models, just like hyperscale competition, and there's an open source check, there is enough room here to go build value on top of models.

[译文] 多个模型，就像超大规模（hyperscale）竞争一样，并且有开源（open source check）的制衡，这里就有足够的空间在模型之上构建价值。

[原文] At Microsoft, the way I look at it is that we are going to be in the hyperscale business, which will support multiple models.

[译文] 在微软，我的看法是我们将参与超大规模业务，它将支持多个模型。

[原文] We will have access to OpenAI models for seven more years, which we will innovate on top of.

[译文] 我们将拥有 OpenAI 模型七年多的访问权限，我们将在其之上进行创新。

[原文] Essentially, I think of ourselves as having a frontier-class model that we can use and innovate on with full flexibility.

[译文] 从本质上讲，我认为我们自己拥有一个前沿级别（frontier-class）的模型，我们可以充分灵活地使用并在其上进行创新。

[原文] And we'll build our own models with MAI.

[译文] 我们也将利用 MAI（Microsoft AI）构建我们自己的模型。

[原文] So we will always have a model level.

[译文] [SN]: 所以我们永远会有一个模型层（model level）。

[原文] And then we'll build—whether it's in security, whether it's in knowledge work, whether it's in coding, or in science—our own application scaffolding, which will be model-forward.

[译文] 然后我们将在安全、知识工作、编码或科学等领域构建我们自己的应用支架（application scaffolding），这将是“模型优先”（model-forward）的。

[原文] It won't be a wrapper on a model, but the model will be wrapped into the application.

[译文] 它不会是模型上的一个包装器，而是模型将被包装到应用程序中。

End of Content

章节 11：向 AI 代理基础设施业务的转型

📝 本节摘要：

访谈者担忧，如果未来的 AI 模型能够像人类一样使用电脑、查看 Excel 公式并自由迁移数据，那么 Office 软件的集成将变得不重要。Satya 纠正了这一观点，他将 Excel 视为 AI 分析师的工具。他区分了两种未来：一种是人类驾驶工具（如 Copilot），另一种是完全自主的 AI 代理（fully autonomous agent）。在后一个世界中，微软的终端用户工具业务（end-user tools business）将转型为支持 AI 代理工作的基础设施业务。AI 代理需要计算资源、存储、归档、发现和身份管理等所有传统的 M365 基础层，这催生了为代理配置 Windows 365 资源的新需求。

Start of Content

[原文] I have so many questions about the other things you mentioned.

[译文] [Q]: 对于您提到的其他事情，我有很多问题。

[原文] But before we move onto those topics, I still wonder whether this is not forward-looking on AI capabilities, where you're imagining models like they exist today.

[译文] 但在我们转到那些话题之前，我仍然怀疑这是否没有前瞻性地看待人工智能的能力，您可能只是在想象今天存在的模型。

[原文] It takes a screenshot of your screen, but it can't look inside each cell and what the formula is.

[译文] 它截取了你的屏幕截图，但它不能查看每个单元格的内部以及公式是什么。

[原文] I think the better mental model here is just imagining that these models will be able to use a computer as well as a human.

[译文] 我认为这里更好的心智模型（mental model）是想象这些模型将能够像人类一样使用计算机。

[原文] A human knowledge worker who is using Excel can look into the formulas, can use alternative software, can migrate data between Office 365 and another piece of software if the migration is necessary, et cetera.

[译文] 使用 Excel 的人类知识工作者可以查看公式，可以使用替代软件，如果需要，可以在 Office 365 和另一个软件之间迁移数据，等等。

[原文] That's kind of what I'm saying.

[译文] [SN]: 我说的就是这个意思。

[原文] But if that's the case, then the

[译文] [Q]: 但如果真是这样，那么

[原文] integration with Excel doesn't matter that much.

[译文] 与 Excel 的集成就不那么重要了。

[原文] No, no, don't worry about the Excel integration.

[译文] [SN]: 不，不，别担心 Excel 的集成。

[原文] After all, Excel was built as a tool for analysts.

[译文] 毕竟，Excel 是作为分析师（analysts）的工具而构建的。

[原文] Great. So whoever is this AI that is an analyst should have tools that they can use.

[译文] 很好。那么，无论这个人工智能分析师是谁，都应该拥有他们可以使用的工具。

[原文] They have the computer. Just the way a human can use a computer. That's their tool.

[译文] 他们有计算机。就像人类可以使用计算机一样。那是他们的工具。

[原文] The tool is the computer.

[译文] 工具就是计算机。

[原文] So all I’m saying is that I'm building an analyst as essentially an AI agent, which happens to come with an a priori knowledge of how to use all of these analytical tools.

[译文] 所以我只是说，我正在将分析师本质上构建为一个人工智能代理（AI agent），它恰好自带如何使用所有这些分析工具的先验知识（a priori knowledge）。

[原文] Just to make sure we're talking about the same thing, is it a thing that a human like me using Excel…

[译文] [Q]: 只是为了确保我们讨论的是同一件事，这是否是像我这样的人使用 Excel……

[原文] No, it's completely autonomous.

[译文] [SN]: 不，它是完全自主的（completely autonomous）。

[原文] So we should now maybe lay out what I think the future of the company is.

[译文] 所以我们现在或许应该阐明我认为公司的未来是什么。

[原文] The future of the company would be the tools business in which I have a computer, I use Excel.

[译文] 公司的未来将是工具业务（tools business），其中我拥有一台计算机，我使用 Excel。

[原文] In fact, in the future I'll even have a Copilot, and that Copilot will also have agents.

[译文] 事实上，未来我甚至会有一个 Copilot，那个 Copilot 也会有代理。

[原文] But it's still me steering everything, and everything is coming back.

[译文] 但仍然是我在引导一切，并且所有事情都在反馈回来。

[原文] That's one world.

[译文] 这是一个世界。

[原文] The second world is the company just literally provisions a computing resource for an AI agent,

[译文] 第二个世界是公司直接为人工智能代理配置（provisions）计算资源，

[原文] and that is working fully autonomously.

[译文] 并且它完全自主地工作。

[原文] That fully autonomous agent will have essentially an embodied set of those same tools that are available to it.

[译文] 那个完全自主的代理将拥有一套本质上可供其使用的、具体化（embodied）的相同工具。

[原文] So this AI tool that comes in also has not just a raw computer, because it's going to be more token-efficient to use tools to get stuff done.

[译文] 所以这个加入进来的人工智能工具也不仅仅拥有原始计算机，因为它使用工具来完成任务会更具代币效率（token-efficient）。

[原文] In fact, I kind of look at it and say that our business, which today is an end-user tools business, will become essentially an infrastructure business in support of agents doing work.

[译文] 事实上，我有点将其视为，并认为我们今天的终端用户工具业务（end-user tools business），将本质上转变为支持代理工作的基础设施业务（infrastructure business）。

[原文] It's another way to think about it.

[译文] 这是另一种思考方式。

[原文] In fact, all the stuff we built underneath M365 still is going to be very relevant.

[译文] 事实上，我们在 M365 之下构建的所有东西仍然会非常重要。

[原文] You need some place to store it, some place to do archival, some place to do discovery, some place to manage all of these activities, even if you're an AI agent.

[译文] 即使你是一个人工智能代理，你也需要存储（store）的地方、进行归档（archival）的地方、进行发现（discovery）的地方、管理所有这些活动的地方。

[原文] It's a new infrastructure.

[译文] 这是一种新的基础设施。

[原文] To make sure I understand, you're saying theoretically a future AI that has actual computer use—which all these model companies are working on right now—could use, even if it's not partnered with Microsoft or under our umbrella, Microsoft software.

[译文] [Q]: 确保我理解了，您是说理论上未来的、能够实际使用计算机的人工智能——所有这些模型公司目前正在致力于此——即使它没有与微软合作或不在我们的保护伞下，也可以使用微软的软件。

[原文] But you're saying, if you're working with our infrastructure, we're going to give them lower-level access that makes it more efficient for you to do the same things you could have otherwise done anyways?

[译文] 但是您是说，如果你使用我们的基础设施，我们将为它们提供更底层的访问权限（lower-level access），使得它们能更高效地完成本来就能完成的同样事情?

[原文] 100%.

[译文] [SN]: 100%。

[原文] What happened is we had servers, then there was virtualization, and then we had many more servers.

[译文] 发生的情况是我们有了服务器，然后有了虚拟化（virtualization），然后我们有了更多的服务器。

[原文] That's another way to think about this.

[译文] 这是另一种思考这个问题的方式。

[原文] Don't think of the tool as the end thing.

[译文] 不要把工具看作是最终的东西。

[原文] What is the entire substrate underneath that tool that humans use?

[译文] 隐藏在人类使用的工具之下的整个底层基础（substrate）是什么?

[原文] That entire substrate is the bootstrap for the AI agent as well,

[译文] 整个底层基础也是人工智能代理的引导程序（bootstrap），

[原文] because the AI agent needs a computer.

[译文] 因为人工智能代理需要一台计算机。

[原文] In fact, one of the fascinating things where we're seeing a significant amount of growth is all these guys who are doing these Office artifacts and what have you, as autonomous agents and so on want to provision Windows 365.

[译文] 事实上，我们看到大量增长的其中一个引人入胜的现象是，所有这些正在做 Office 工件（artifacts）等等的、作为自主代理（autonomous agents）的公司，都希望配置（provision） Windows 365。

[原文] They really want to be able to provision a computer for these agents.

[译文] 他们真的希望能够为这些代理配置一台计算机。

[原文] Absolutely.

[译文] 绝对如此。

[原文] That's why we're going to have essentially an end-user computing infrastructure business, which is going to just keep growing because it's going to grow faster than the number of users.

[译文] 这就是为什么我们本质上会拥有一个终端用户计算基础设施业务（end-user computing infrastructure business），它将持续增长，因为它将比用户数量增长得更快。

End of Content

章节 12：从“每用户”到“每代理”的计费模式

📝 本节摘要：

Satya 认为，未来的计费模式将从“每用户”（per-user）转向“每代理”（per-agent）。为每个代理配置一台计算机、身份（identity）、安全和管理层（management layers）将成为关键。访谈者提到 AI 的主要价值驱动力在于数据迁移（例如将 Excel/Word 数据迁移到 SQL 数据库），这可能导致 Office 生态系统的使用减少。Satya 承认混合世界将持续很长时间，但即使在“代理与代理”的完全优化世界中，仍需要存储、电子发现（e-discovery）、可观测性和身份系统等核心底层设施，而这正是微软的优势。

Start of Content

[原文] That's one of the other questions people ask me, "Hey, what happens to the per-user business?"

[译文] [SN]: 这是人们问我的另一个问题：“嘿，每用户（per-user）业务会发生什么”?

[原文] At least the early signs maybe, the way to think about the per-user business is not just per user, it's per agent.

[译文] 至少早期的迹象可能是，思考每用户业务的方式不只是按用户，而是按代理（per agent）。

[原文] And if you say it's per user and per agent, the key is what's the stuff to provision for every agent?

[译文] 如果你说它是按用户和按代理计算的，那么关键是为每个代理配置什么?

[原文] A computer, a set of security things around it, an identity around it.

[译文] 一台计算机，围绕它的一套安全配置，一个身份（identity）。

[原文] All those things, observability and so on, are the management layers.

[译文] 所有这些东西，可观测性等等，都是管理层（management layers）。

[原文] That's all going to get baked into that.

[译文] 所有这些都将被融入其中。

[原文] The way to frame it—at least the way I currently think about it and I’d like to hear your view—is that these model companies are all building environments to train their models to use Excel or Amazon shopping or whatever it is, book flights.

[译文] [Q]: 至少我目前思考这个问题的方式（我想听听您的看法）是：这些模型公司都在构建环境，来训练它们的模型使用 Excel 或亚马逊购物，或者预订航班之类的。

[原文] But at the same time, they're also training these models to do migration.

[译文] 但与此同时，它们也在训练这些模型进行迁移（migration）。

[原文] Because that is probably the most immediately valuable thing: converting mainframe-based systems to standard cloud systems, converting Excel databases into real databases with SQL, or converting what is done in Word and Excel to something that is more programmatic and

[译文] 因为这可能是最立即可见的有价值的事情：将基于大型机（mainframe-based）的系统转换为标准云系统，将 Excel 数据库转换为带有结构化查询语言（SQL）的真实数据库，或者将 Word 和 Excel 中完成的工作转换为更具编程性（programmatic）的

[原文] more efficient in a classical sense that can be done by humans as well.

[译文] 并在传统意义上更有效率的东西，这也是人类可以做到的。

[原文] It's just not cost-effective for the software developer to do that.

[译文] 只是软件开发人员做这件事不划算。

[原文] That seems to be what everyone is going to do with AI, for the next few years at least, to massively drive value.

[译文] 这似乎是未来几年每个人都会利用人工智能做的事情，以大规模地驱动价值。

[原文] How does Microsoft fit into that if the models can utilize the tools themselves to migrate to something?

[译文] 如果模型可以利用工具本身迁移到其他事物，微软如何适应这种情况?

[原文] Yes, Microsoft has a leadership position in databases and in storage and in all these other categories, but the use of an Office ecosystem is going to

[译文] 是的，微软在数据库、存储和所有其他类别中都处于领先地位，但是 Office 生态系统的使用将会

[原文] be significantly less just like the use of a mainframe ecosystem could be potentially less.

[译文] 显著减少，就像大型机生态系统的使用可能会减少一样。

[原文] Now mainframes have grown for the last two decades actually, even though no one talks about them anymore.

[译文] 尽管现在没有人再谈论大型机，但实际上在过去的二十年里，大型机一直在增长。

[原文] They've still grown.

[译文] 它们仍在增长。

[原文] 100%, I agree with that.

[译文] [SN]: 100%，我同意这一点。

[原文] How does that flow?

[译文] [Q]: 这种情况如何发展?

[原文] At the end of the day, there is going to be a significant amount of time where there's going to be a hybrid world, because people are going to be using the tools that are going to be working with agents that have to use tools, and they have to communicate with each other.

[译文] [SN]: 归根结底，将在很长一段时间内存在一个混合世界（hybrid world），因为人们将使用工具，这些工具将与必须使用工具的代理一起工作，而且它们必须相互通信。

[原文] What's the artifact I generate that

[译文] 我生成的工件是什么

[原文] then a human needs to see?

[译文] 然后人类需要看到什么?

[原文] All of these things will be real considerations in any place, the outputs, inputs.

[译文] 所有这些事情，输出、输入，在任何地方都将是实际的考虑因素。

[原文] I don't think it'll just be about, "Oh, I migrated off."

[译文] 我不认为它仅仅是关于：“哦，我已经迁移出去了”。

[原文] The bottom line is that I have to live in this hybrid world.

[译文] 底线是我必须生活在这个混合世界中。

[原文] But that doesn't fully answer your question because there can be a real new efficient frontier where it's just agents working with agents and completely optimized.

[译文] 但这并没有完全回答你的问题，因为可能存在一个全新的高效前沿（efficient frontier），其中只是代理与代理一起工作，并且完全优化。

[原文] Even when agents are working with agents, what are the primitives that are needed?

[译文] 即使代理与代理一起工作，需要哪些基本要素（primitives）?

[原文] Do you need a storage system?

[译文] 你需要一个存储系统吗?

[原文] Does that storage system need to have e-discovery?

[译文] 那个存储系统需要具备电子发现（e-discovery）功能吗?

[原文] Do you need to have observability?

[译文] 你需要具备可观测性吗?

[原文] Do you need to have an identity system that is going to use multiple models with all having one identity system?

[译文] 你是否需要一个身份系统（identity system），它将使用多个模型，并且所有模型都拥有一个统一的身份系统?

[原文] These are all the core underlying rails we have today for what are the Office systems or what have you.

[译文] 所有这些都是我们今天拥有的核心底层轨道（rails），用于 Office 系统或其他系统。

[原文] And that's what we will have in the future as well.

[译文] 这也是我们未来将拥有的。

[原文] You've talked about databases.

[译文] 你谈到了数据库。

[原文] I mean man, I would love all of Excel to have a database backend.

[译文] 我的意思是，我希望所有的 Excel 都能有一个数据库后端（database backend）。

[原文] I would love for all that to happen immediately.

[译文] 我希望所有这些都能立即发生。

[原文] And that database is a good database.

[译文] 并且那个数据库是一个好的数据库。

[原文] Databases in fact will be a big thing that will grow.

[译文] 事实上，数据库将是会大幅增长的大业务。

[原文] If I think about all of the Office artifacts being structured better, the ability to do the joins between structured and unstructured better because of the agentic world, that will grow the underlying infrastructure business.

[译文] 如果我考虑到所有 Office 工件得到更好的结构化，以及由于代理世界（agentic world）而更好地实现结构化和非结构化数据之间的连接（joins），这将推动底层基础设施业务的增长。

End of Content

章节 13：模型研发策略：GPT 家族合作与 MAI 的差异化发展

📝 本节摘要：

访谈者质疑微软自主研发的 MAI 模型（Microsoft AI）排名较低，并提问微软如何确保持有领先模型。Satya 阐述了微软的模型战略：第一，在未来七年内最大化利用 OpenAI 模型，通过强化学习微调（RL fine-tuning）和中期训练（mid-training runs）等方式，利用微软独特的专有数据资产在其上创新，增加价值。第二，MAI 团队的目标是建立世界级的超级智能团队（superintelligence team），避免与 GPT 家族进行重复性（duplicative）的工作，而是专注于成本优化和差异化能力。他提到 MAI 已经发布了排名靠前的图像模型和专注于产品优化的音频模型，并计划开发一个融合音频、图像和文本能力的“全能模型”（omni-model），旨在通过实际研究为未来五到八年的超级智能突破做准备。

Start of Content

[原文] It happens that the consumption of that is all being driven by agents.

[译文] [SN]: 碰巧的是，所有这些消费都是由代理驱动的。

[原文] You could say all that is just-in-time generated software by a model company.

[译文] 你可能会说，所有这些都是模型公司即时（just-in-time）生成的软件。

[原文] That could also be true. We will be one such model company too.

[译文] 这也可能是真的。我们也将是这样一家模型公司。

[原文] We will build in... The competition could be that we will build a model plus all the infrastructure and provision it, and then there will be competition between a bunch of those folks who can do that.

[译文] 我们将构建……竞争可能是我们构建一个模型加上所有基础设施并进行配置（provision），然后一群能够这样做的人之间会展开竞争。

[原文] Speaking of model companies, you say not only will you have the infrastructure, you'll have the model itself.

[译文] [Q]: 谈到模型公司，您说您不仅拥有基础设施，还将拥有模型本身。

[原文] Right now, Microsoft AI's most recent model that was released two months ago is 36 in Chatbot Arena.

[译文] 目前，微软人工智能（Microsoft AI）两个月前发布的最新模型在 Chatbot Arena 中排名第 36 位。

[原文] You obviously have the IP rights to OpenAI.

[译文] 你们显然拥有 OpenAI 的知识产权（IP rights）。

[原文] To the extent you agree with that, it seems to be behind.

[译文] 在您同意的范围内，这似乎是落后了。

[原文] Why is that the case, especially given the fact that you theoretically have the right to fork OpenAI's monorepo or distill their models, especially if it's a big part of your strategy that you need to have a leading model company?

[译文] 为什么会出现这种情况，特别是考虑到理论上您有权分叉（fork）OpenAI 的单一代码库（monorepo）或精炼（distill）他们的模型，尤其如果您的战略中很重要的一部分是您需要拥有一家领先的模型公司?

[原文] First of all, we are absolutely going to use the OpenAI models to the maximum across all of our products.

[译文] [SN]: 首先，我们绝对会在我们所有的产品中最大限度地使用 OpenAI 模型。

[原文] That's the core thing that we're going to continue to do all the way for the next seven years, and not just use it but then add value to it.

[译文] 这是我们在接下来的七年里将一直持续做的核心事情，不仅是使用它，还要在此基础上增加价值。

[原文] That's where the analyst and this Excel agent, these are all things that we will

[译文] 这就是分析师和这个 Excel 代理的用武之地，所有这些都是我们将要

[原文] do where we'll do RL fine-tuning.

[译文] 做的事情，我们将在其中进行强化学习微调（RL fine-tuning）。

[原文] We'll do some mid-training runs on top of a GPT family where we have unique data assets and build capability.

[译文] 我们将在 GPT 家族之上进行一些中期训练（mid-training runs），因为我们拥有独特的数据资产（data assets），并以此构建能力。

[原文] With the MAI model, the way that I think we’re going to think about it is that the good news here with the new agreement is we can be very, very clear that we're going to build a world-class superintelligence team and go after it with a high ambition.

[译文] 对于 MAI 模型，我认为我们的想法是，新协议带来的好消息是我们可以非常、非常明确地表明，我们将建立一个世界级的超级智能团队（superintelligence team），并以雄心壮志去追求它。

[原文] But at the same time, we're also going to use this time to be smart about how to use both these things.

[译文] 但与此同时，我们也将利用这段时间，巧妙地运用这两者。

[原文] That means we will, on one end,

[译文] 这意味着，一方面，我们将

[原文] be very product-focused, and on the other end, be very research-focused.

[译文] 非常专注于产品（product-focused），另一方面，又非常专注于研究（research-focused）。

[原文] Because we have access to the GPT family, the last thing I want to do is use my flops in a way that is just duplicative and doesn't add much value.

[译文] 因为我们可以访问 GPT 家族，我最不希望做的事情是以一种只是重复（duplicative）且没有增加太多价值的方式来使用我的浮点运算能力（flops）。

[原文] I want to be able to take the flops that we use to generate a GPT family and maximize its value, while my MAI flops are being used for… Let's take the image model that we launched, which I think is at number nine in the image arena.

[译文] 我希望能够利用我们用于生成 GPT 家族的浮点运算能力并使其价值最大化，同时我的 MAI 浮点运算能力用于……比如我们发布的图像模型，我认为它在图像竞技场（image arena）中排名第九。

[原文] We're using it both for cost optimization, it's on Copilot,

[译文] 我们用它来进行成本优化，它用于 Copilot，

[原文] it's in Bing, and we're going to use that.

[译文] 它在 Bing 中，我们将使用它。

[原文] We have an audio model in Copilot.

[译文] 我们在 Copilot 中有一个音频模型。

[原文] It's got personality and what have you.

[译文] 它具备个性和其他功能。

[原文] We optimized it for our product.

[译文] 我们针对我们的产品对其进行了优化。

[原文] So we will do those.

[译文] 所以我们会做这些。

[原文] Even on the LMArena, we started on the text one and it debuted at like 13.

[译文] 即使在 LMArena 上，我们以文本模型起步，首次亮相时排名大约在第 13 位。

[原文] By the way, it was done only on around 15,000 H100s.

[译文] 顺便说一句，它只用了大约 15,000 个 H100 芯片。

[原文] It was a very small model.

[译文] 这是一个非常小的模型。

[原文] So it was, again, to prove out the core capability, the instruction following, and everything else.

[译文] 所以，它再次是为了验证核心能力、指令遵循（instruction following）以及其他一切。

[原文] We wanted to make sure we could match what was state of the art.

[译文] 我们想确保我们能够匹配最先进（state of the art）的水平。

[原文] That shows us, given scaling laws, what we are capable of doing if we gave more flops to it.

[译文] 这向我们展示了，根据扩展定律（scaling laws），如果我们投入更多的浮点运算能力（flops），我们能够做到什么。

[原文] The next thing we will do is an omni-model where we will take the work we have done in audio, what we have done in image, and what we have done in text.

[译文] 我们下一步将做的是一个全能模型（omni-model），我们将整合我们在音频、图像和文本方面所做的工作。

[原文] That will be the next pit stop on the MAI side.

[译文] 这将是 MAI 方面的下一个目标。

[原文] So when I think about the MAI roadmap, we are going to build a first-class superintelligence team.

[译文] 所以当我考虑 MAI 路线图时，我们将建立一个一流的超级智能团队。

[原文] We are going to continue to drop, and do it in the open, some of these models.

[译文] 我们将继续发布其中一些模型，并以开放的方式进行。

[原文] They will either be used in our products, because they're going to be latency-friendly, cost-friendly, or what have you, or they'll have some special capability.

[译文] 它们要么被用于我们的产品中，因为它们将对延迟（latency-friendly）和成本（cost-friendly）有利，或者具备一些特殊能力。

[原文] And we will do real research in order to be ready for the next five, six, seven, eight breakthroughs that are all needed on this march towards superintelligence—while exploiting the advantage we have of having the GPT family that we can work on top of as well.

[译文] 我们将进行真正的研究，以便为迈向超级智能（superintelligence）进程中所需的未来五、六、七、八项突破做好准备——同时利用我们拥有 GPT 家族的优势，我们也可以在其之上进行工作。

End of Content

章节 14：模型持续学习的风险与多模型架构的必要性

📝 本节摘要：

访谈者提出，如果模型达到人类水平，并具有“持续学习”（continuously learn on the job）的能力，将会产生指数级的反馈循环，可能导致单一模型的赢家通吃（intelligence explosion），如果微软不是领先者，则其平台优势将削弱。Satya 认为，现实中不存在单一模型主导一切的局面，而是像数据库市场一样，存在多个模型。他强调持续学习带来的“数据流动性”网络效应不会在所有领域、地理和细分市场同时发生。对于超大规模服务商来说，必须构建一个能够支持多种模型家族和架构（如 MoE）的灵活基础设施，以避免因模型迭代（如架构调整）导致资本浪费。行业结构将迫使企业进行专业化，而非追求垂直垄断。

Start of Content

[原文] Say we roll forward seven years, you no longer have access to OpenAI models.

[译文] [Q]: 假设我们快进七年，您不再拥有对 OpenAI 模型的访问权限。

[原文] What does Microsoft do to make sure they are leading, or have a leading AI lab?

[译文] 微软会怎么做来确保他们处于领先地位，或者拥有一个领先的人工智能实验室?

[原文] Today, OpenAI has developed many of the breakthroughs, whether it be scaling or reasoning.

[译文] 今天，OpenAI 已经开发了许多突破，无论是扩展（scaling）还是推理（reasoning）。

[原文] Or Google's developed all the breakthroughs like transformers.

[译文] 或者像 Google 开发了诸如 transformer 等所有突破。

[原文] But it is also a big talent game.

[译文] 但这也是一场人才（talent）的巨大竞争。

[原文] You've seen Meta spend north of $20 billion on talent.

[译文] 你看到 Meta 在人才上花费了超过 200 亿美元。

[原文] You've seen Anthropic poach the entire Blueshift reasoning team from Google last year.

[译文] 你看到 Anthropic 去年从 Google 挖走了整个 Blueshift 推理团队。

[原文] You've seen Meta poach a large reasoning and post-training team from Google more recently.

[译文] 你看到 Meta 最近从 Google 挖走了一个大型推理和后期训练团队。

[原文] These sorts of talent wars are very capital intensive.

[译文] 这类人才战争是非常资本密集（capital intensive）的。

[原文] Arguably, if you're spending $100 billion on infrastructure, you should also spend X amount of money on the people

[译文] 可以说，如果您在基础设施上花费 1000 亿美元，您也应该在人才上花费 X 金额

[原文] using the infrastructure so that they're more efficiently making these new breakthroughs.

[译文] 以让使用基础设施的人能够更高效地实现这些新的突破。

[原文] What confidence can one get that Microsoft will have a team that's world-class that can make these breakthroughs?

[译文] 人们如何获得信心，相信微软将拥有一支能够取得这些突破的世界级团队?

[原文] Once you decide to turn on the money faucet—you're being a bit capital efficient right now, which is smart it seems, to not waste money doing duplicative work—but once you decide you need to, how can one say, "Oh yeah, now you can shoot up to the top five model?"

[译文] 一旦您决定开启资金水龙头（money faucet）——您目前看起来是在进行资本效率（capital efficient）较高的做法，不浪费钱去做重复性工作——但一旦您决定需要这样做，人们如何能说：“哦，是的，现在你们可以一跃成为排名前五的模型”?

[原文] At the end of the day, we're going to build

[译文] [SN]: 归根结底，我们将构建

[原文] a world-class team and we already have a world-class team that's beginning to be assembled.

[译文] 一个世界级的团队，我们已经有一支正在组建中的世界级团队。

[原文] We have Mustafa coming in, we have Karen.

[译文] 我们有穆斯塔法（Mustafa）加入，我们有凯伦（Karen）。

[原文] We have Amar Subramanya who did a lot of the post-training at Gemini 2.5 who's at Microsoft.

[译文] 我们有阿玛尔·苏布拉马尼亚（Amar Subramanya），他在 Gemini 2.5 做了大量的后期训练工作，他现在在微软。

[原文] Nando, who did a lot of the multimedia work at DeepMind, is there.

[译文] 南多（Nando），他在 DeepMind 做了大量的多媒体工作，也在那里。

[原文] We're going to build a world-class team.

[译文] 我们将组建一个世界级的团队。

[原文] In fact, later this week even, Mustafa will publish something with a little more clarity on what our lab is going to go do.

[译文] 事实上，就在本周晚些时候，穆斯塔法将发布一些更清晰的内容，说明我们的实验室将要做什么。

[原文] The thing that I want the world to know, perhaps,

[译文] 也许我希望世界知道的是，

[原文] is that we are going to build the infrastructure that will support multiple models.

[译文] 我们将构建支持多种模型（multiple models）的基础设施。

[原文] Because from a hyperscale perspective, we want to build the most scaled infrastructure fleet that's capable of supporting all the models the world needs, whether it's from open source or obviously from OpenAI and others.

[译文] 因为从超大规模（hyperscale）的角度来看，我们希望构建最具规模的基础设施集群（infrastructure fleet），能够支持世界所需的所有模型，无论是来自开源还是显然来自 OpenAI 和其他公司。

[原文] That's one job.

[译文] 这是一项工作。

[原文] Secondly, in our own model capability, we will absolutely use the OpenAI model in our products and we'll start building our own model.

[译文] 其次，在我们自己的模型能力方面，我们绝对会在我们的产品中使用 OpenAI 模型，并且我们将开始构建我们自己的模型。

[原文] And we may—like in GitHub Copilot where Anthropic is used—even have other frontier models that are

[译文] 而且我们可能——就像在使用了 Anthropic 的 GitHub Copilot 中一样——甚至拥有其他

[原文] going to be wrapped into our products, as well.

[译文] 将被整合到我们产品中的前沿模型（frontier models）。

[原文] I think that’s how each time…

[译文] 我认为每一次都是这样……

[原文] At the end of the day, the eval of the product as it meets a particular task or a job is what matters.

[译文] 归根结底，产品在完成特定任务或工作时的评估（eval）才是最重要的。

[原文] We'll start back from there into the vertical integration needed, knowing that as long as you're serving the market well with the product, you can always cost-optimize.

[译文] 我们将从那里出发，进行所需的垂直整合（vertical integration），要知道只要你用产品很好地服务于市场，你就总能进行成本优化。

[原文] There's a question going forward.

[译文] [Q]: 展望未来，有一个问题。

[原文] Right now, we have models that have this distinction between training and inference.

[译文] 现在，我们的模型在训练（training）和推理（inference）之间存在区别。

[原文] One could argue that there's a smaller and smaller

[译文] 有人可能会认为，不同模型之间的差异越来越小

[原文] difference between the different models.

[译文] 差异越来越小。

[原文] Going forward, if you're really expecting something like human-level intelligence, humans learn on the job.

[译文] 展望未来，如果您确实期望出现类似人类水平的智能，人类是在工作中学习的。

[原文] If you think about your last 30 years, what makes Satya tokens so valuable?

[译文] 如果您想想您过去的 30 年，是什么让萨提亚代币如此有价值?

[原文] It's the last 30 years of wisdom and experience you've gained in Microsoft.

[译文] 是您在微软积累的过去 30 年的智慧和经验。

[原文] We will eventually have models, if they get to human level, which will have this ability to continuously learn on the job.

[译文] 最终我们将拥有模型，如果它们达到人类水平，它们将具备这种在工作中持续学习（continuously learn on the job）的能力。

[原文] That will drive so much value to the model company that is ahead, at least in my view, because you have copies of one model broadly deployed through

[译文] 至少在我看来，这将为领先的模型公司带来巨大的价值，因为你将一个模型的副本广泛部署在

[原文] the economy learning how to do every single job.

[译文] 经济的各个角落，学习如何完成每一项工作。

[原文] And unlike humans, they can amalgamate their learnings to that model.

[译文] 而且与人类不同，它们可以将它们的学习成果融合（amalgamate）到那个模型中。

[原文] So there's this sort of continuous learning exponential feedback loop, which almost looks like a sort of intelligence explosion.

[译文] 因此，存在着这种持续学习的指数反馈循环（exponential feedback loop），这几乎看起来像是一种智能爆炸（intelligence explosion）。

[原文] If that happens and Microsoft isn't the leading model company by that time…

[译文] 如果这种情况发生，而那时微软不是领先的模型公司……

[原文] You're saying that well, we substitute one model for another, et cetera. Doesn’t that then matter less?

[译文] 您说我们可以用一个模型代替另一个模型，等等。那么这是否就不那么重要了?

[原文] Because it's like this one model knows how to do every single job in

[译文] 因为这就像是这个模型知道如何完成经济中的每一项工作

[原文] the economy, the others in the long tail don't.

[译文] 而长尾中的其他模型则不知道。

[原文] Your point, if there's one model that is the only model that's most broadly deployed in the world and it sees all the data and it does continuous learning, that's game set match and you stop shop.

[译文] [SN]: 你的观点是，如果有一个模型是世界上部署最广泛的唯一模型，它看到所有数据并进行持续学习，那就是大局已定（game set match），你就可以关门大吉了。

[原文] The reality that at least I see is that in the world today, for all the dominance of any one model, that is not the case.

[译文] 至少我所看到的现实是，在今天的世界中，尽管任何一个模型都占据主导地位，但情况并非如此。

[原文] Take coding, there are multiple models.

[译文] 以编码为例，存在多个模型。

[原文] In fact, everyday it's less the case.

[译文] 事实上，这种情况每天都在减少。

[原文] There is not one model that is getting deployed broadly.

[译文] 并不是只有一个模型被广泛部署。

[原文] There are multiple models that are getting deployed.

[译文] 有多个模型正在被部署。

[原文] It's like databases. It's always the thing, "Can one database be the one that is just used everywhere?" Except it's not.

[译文] 这就像数据库。总是有这样的问题：“有没有一个数据库可以到处使用？”但事实并非如此。

[原文] There are multiple types of databases that are getting deployed for different use cases.

[译文] 有多种类型的数据库被部署用于不同的使用案例（use cases）。

[原文] I think that there are going to be some network effects of continual learning—I call it data liquidity—that any one model has.

[译文] 我认为持续学习会产生一些网络效应（network effects）——我称之为数据流动性（data liquidity）——任何一个模型都会拥有。

[原文] Is it going to happen in all domains? I don't think so.

[译文] 它会在所有领域（domains）都发生吗？我不这么认为。

[原文] Is it going to happen in all geos? I don't think so.

[译文] 它会在所有地理区域（geos）都发生吗？我不这么认为。

[原文] Is it going to happen in all segments? I don't think so.

[译文] 它会在所有细分市场（segments）都发生吗？我不这么认为。

[原文] It'll happen in all categories at the same time? I don't think so.

[译文] 它会在所有类别同时发生吗？我不这么认为。

[原文] So therefore I feel like the design space is so large that there's plenty of opportunity.

[译文] 因此，我觉得设计空间（design space）如此之大，以至于存在大量的机会。

[原文] But your fundamental point is having a capability which is at the infrastructure layer, model layer, and at the scaffolding layer, and then being able to compose these things not just as a vertical stack, but to be able to compose each thing for what its purpose is.

[译文] 但你基本的观点是：在基础设施层、模型层和支架层都具备能力，然后能够组合这些东西，而不仅仅是作为一个垂直堆栈（vertical stack），而是能够根据每个事物的目的来组合它们。

[原文] You can't build an infrastructure that's optimized for one model.

[译文] 你不能构建一个只为单一模型优化（optimized）的基础设施。

[原文] If you do that, what if you fall behind?

[译文] 如果你那样做了，万一你落后了怎么办?

[原文] In fact, all the infrastructure you built will be a waste.

[译文] 事实上，你建造的所有基础设施都将是浪费。

[原文] You kind of need to build an infrastructure that's capable of supporting multiple families and lineages of models.

[译文] 你需要构建一个能够支持多种家族和谱系模型（multiple families and lineages of models）的基础设施。

[原文] Otherwise the capital you put in, which is optimized for one model architecture, means you're one tweak away, some MoE-like breakthrough that happens, and your entire network topology goes out of the window.

[译文] 否则，你投入的资本，如果只针对一种模型架构进行优化，就意味着只要进行一次微调（tweak），发生一些 MoE（混合专家模型）般的突破，你整个网络拓扑结构（network topology）就会报废。

[原文] That's a scary thing.

[译文] 那是很可怕的事情。

[原文] Therefore you kind of want the infrastructure

[译文] 因此，你希望基础设施

[原文] to support whatever may come in your own model family and other model families.

[译文] 能够支持你自己的模型家族和其他模型家族中可能出现的任何东西。

[原文] You've got to be open.

[译文] 你必须保持开放。

[原文] If you're serious about the hyperscale business, you've got to be serious about that.

[译文] 如果你认真对待超大规模（hyperscale）业务，你就必须认真对待这一点。

[原文] If you're serious about being a model company, you have to basically say, "What are the ways people can do things on top of the model so that I can have an ISV ecosystem?"

[译文] 如果你认真对待成为一家模型公司，你基本上必须说：“人们可以在模型之上做些什么，以便我能拥有一个独立软件供应商（ISV）生态系统”?

[原文] Unless I'm thinking I'll own every category, that just can't be that.

[译文] 除非我想独占每一个类别，否则就不可能。

[原文] Then you won't have an API business and that, by definition, will mean you'll never be a platform

[译文] 那么你就不会拥有应用程序编程接口（API）业务，根据定义，这意味着你永远不会成为一个

[原文] company that's successfully deployed everywhere.

[译文] 成功部署到任何地方的平台公司。

[原文] Therefore the industry structure is such that it will really force people to specialize.

[译文] 因此，行业结构会迫使人们进行专业化（specialize）。

[原文] In that specialization, a company like Microsoft should compete in each layer by its merits, but not think that this is all about the road to game set match, where I just compose vertically all these layers.

[译文] 在这种专业化中，像微软这样的公司应该凭借自身优势在每一层竞争，而不是认为这一切都是通往“大局已定”（game set match）的道路，即我只是垂直组合所有这些层。

[原文] That just doesn't happen.

[译文] 那根本不会发生。

End of Content

章节 15：暂停超大规模建设：追求弹性与客户多样性

📝 本节摘要：

访谈者提到微软在 2023 年下半年暂停了大规模数据中心租赁和建设，导致其容量预测有所下降，并询问原因。Satya 解释称，暂停建设是由于战略调整：微软需要确保其 Azure 计算集群具有弹性（fungibility），能够支持 AI 的所有阶段（训练、推理、数据生成），并避免被锁定在单一芯片代次（generation）上。更重要的是，微软不想成为单一客户（OpenAI）的大规模托管商（hoster），而是要建立一个支持多样化工作负载和客户的超大规模集群，以便通过长期尾部工作负载实现持续盈利和再投资。

Start of Content

[原文] So last year Microsoft was on path to be the largest infrastructure provider by far.

[译文] [Q]: 去年，微软有望成为迄今为止最大的基础设施提供商。

[原文] You were the earliest in 2023, so you went out there, you acquired all the resources in terms of leasing data centers, starting construction, securing power, everything.

[译文] 你们是 2023 年最早行动的，因此你们出面获取了所有资源，包括租赁数据中心、开始建设、确保电力供应，所有这些。

[原文] You guys were on pace to beat Amazon in 2026 or 2027.

[译文] 你们当时有望在 2026 年或 2027 年击败亚马逊。

[原文] Certainly by 2028 you were going to beat them.

[译文] 肯定到 2028 年你们会击败他们。

[原文] Since then, let’s call it, in the second half of last year, Microsoft did this big pause, where they let go of a bunch of leasing sites that they were going to take, which then Google, Meta, Amazon in some cases, Oracle, took these sites.

[译文] 从那时起，姑且称之为去年下半年，微软进行了一次大规模暂停（big pause），放弃了他们原本打算接手的一批租赁站点，随后 Google、Meta、Amazon 在某些情况下，还有 Oracle 接手了这些站点。

[原文] We're sitting in one of the largest data centers in the world, so obviously it's not everything,

[译文] 我们正坐落在世界上最大的数据中心之一，所以显然不是全部，

[原文] you guys are expanding like crazy.

[译文] 你们正在疯狂扩张。

[原文] But there are sites that you just stopped working on. Why did you do this?

[译文] 但有些站点你们停止了工作。你们为什么这样做?

[原文] This goes back a little bit to, what is the hyperscale business all about?

[译文] [SN]: 这又回到了超大规模业务的本质是什么的问题上。

[原文] One of the key decisions we made was that if we're going to build out Azure to be fantastic for all stages of AI—from training to mid-training to data gen to inference—we just need fungibility of the fleet.

[译文] 我们做出的一个关键决定是，如果我们要将 Azure 打造得在人工智能的所有阶段（从训练到中期训练，到数据生成，再到推理）都表现出色，我们就需要集群（fleet）的弹性或互换性（fungibility）。

[原文] So that entire thing caused us basically not to go build a whole lot of capacity with a particular set of generations.

[译文] 所以，整个考虑导致我们基本上没有使用特定的几代（generations）芯片去构建大量的容量。

[原文] Because the other thing you have to

[译文] 因为你必须认识到的另一件事是

[原文] realize is that having up to now 10x'ed every 18 months enough training capacity for the various OpenAI models, we realized that the key is to stay on that path.

[译文] 认识到直到现在，我们已经每 18 个月将各种 OpenAI 模型的训练容量增加了 10 倍，我们意识到关键是要保持这条道路。

[原文] But the more important thing is to have a balance, to not just train, but to be able to serve these models all around the world.

[译文] 但更重要的是要保持平衡，不仅仅是训练，还要能够为世界各地的这些模型提供服务（serve）。

[原文] Because at the end of the day, the rate of monetization is what will then allow us to keep funding.

[译文] 因为归根结底，变现率（monetization）才能让我们持续获得资金。

[原文] And then the infrastructure was going to need us to support multiple models.

[译文] 此外，基础设施需要我们支持多种模型。

[原文] So once we said that that's the case, we just course-corrected to the path we're on.

[译文] 所以一旦我们确定了这种情况，我们就对我们所走的道路进行了修正（course-corrected）。

[原文] If I look at the path we're on, we are doing a lot more starts now.

[译文] 如果我看我们现在走的路径，我们现在正在做更多的启动项目（starts）。

[原文] We are also buying up as much managed capacity as we can, whether it's to build, whether it's to lease, or even GPUs as a service.

[译文] 我们也在尽力收购受管容量（managed capacity），无论是通过建设、租赁，还是甚至是以服务形式提供的图形处理器（GPUs as a service）。

[原文] But we're building it for where we see the demand and the serving needs and our training needs.

[译文] 但我们是根据我们看到的，在哪里有需求、服务需求和我们的训练需求来建设的。

[原文] We didn't want to just be a hoster for one company and have just a massive book of business with one customer.

[译文] 我们不想仅仅成为一家公司的托管商（hoster），与一个客户拥有大量的业务。

[原文] That's not a business, you should be vertically integrated with that company.

[译文] 那不是一门生意，你应该与那家公司垂直整合。

[原文] Given that OpenAI was going to be a successful independent company,

[译文] 鉴于 OpenAI 将成为一家成功的独立公司，

[原文] which is fantastic. It makes sense.

[译文] 这太棒了。这是有道理的。

[原文] And even Meta may use third-party capacity, but ultimately they're all going to be first-party.

[译文] 即使是 Meta 可能会使用第三方容量，但最终他们都会是第一方（first-party）。

[原文] For anyone who has large scale, they'll be a hyperscaler on their own.

[译文] 对于任何拥有大规模资源的公司，他们将成为他们自己的超大规模服务商（hyperscaler）。

[原文] To me, it was to build out a hyperscale fleet and our own research compute.

[译文] 对我来说，（调整的目的）是建立一个超大规模集群和我们自己的研究计算能力（research compute）。

[原文] That's what the adjustment was.

[译文] 这就是我们进行的调整。

[原文] So I feel very, very good.

[译文] 所以我感觉非常非常好。

[原文] By the way, the other thing is that I didn't want to get stuck with massive scale of one generation.

[译文] 顺便说一句，另一件事是我不想被困在单一代次（one generation）的大规模资源上。

[原文] We just saw the GB200s, the GB300s are coming.

[译文] 我们刚刚看到了 GB200s，GB300s 正在到来。

[原文] By the time I get to Vera Rubin, Vera Rubin Ultra, the data center is going to look very different because the power per rack, power per row, is going to be so different.

[译文] 等到 Vera Rubin、Vera Rubin Ultra 问世时，数据中心看起来会非常不同，因为每个机架（rack）的功率、每行（row）的功率将截然不同。

[原文] The cooling requirements are going to be so different.

[译文] 冷却要求也将截然不同。

[原文] That means I don't want to just go build out a whole number of gigawatts that are only for a one-generation, one family.

[译文] 这意味着我不想只是建设大量的千兆瓦（gigawatts）容量，而这些容量只适用于单一的代次、单一的家族。

[原文] So I think the pacing matters, the fungibility and the location matters, the workload diversity matters, customer diversity matters and that's what we’re building towards.

[译文] 所以我认为节奏（pacing）很重要，弹性（fungibility）和位置很重要，工作负载多样性（workload diversity）很重要，客户多样性（customer diversity）很重要，这就是我们正在努力建设的方向。

[原文] The other thing that we've learned a lot is that every AI workload does

[译文] 我们学到的另一件重要的事情是，每种人工智能工作负载（AI workload）确实

[原文] require not only the AI accelerator, but it requires a whole lot of other things.

[译文] 不仅需要人工智能加速器（AI accelerator），它还需要大量的其他东西。

[原文] In fact, a lot of the margin structure for us will be in those other things.

[译文] 事实上，我们大部分的利润结构（margin structure）将来自那些其他的东西。

[原文] Therefore, we want to build out Azure as being fantastic for the long tail of the workloads, because that's the hyperscale business, while knowing that we've got to be super competitive starting with the bare-metal for the highest end training.

[译文] 因此，我们希望将 Azure 打造成非常适合长尾工作负载（long tail of the workloads）的服务，因为这才是超大规模业务，同时要知道我们必须从最高端训练的裸机（bare-metal）开始，就具备超强的竞争力。

[原文] But that can't crowd out the rest of the business, because we're not in the business of just doing five contracts with five customers being their bare-metal service.

[译文] 但这不能挤占我们其他业务的空间，因为我们的业务不是仅仅为五个客户提供裸机服务（bare-metal service）。

[原文] That's not a Microsoft business.

[译文] 那不是微软的业务。

[原文] That may be a business for someone else, and that's a good thing.

[译文] 那可能是其他人的业务，这是件好事。

End of Content

章节 16：地理位置、延迟与数据主权的影响

📝 本节摘要：

访谈者提问：既然一些 AI 任务可能需要数小时或数天，为什么数据中心的位置（如亚特兰大）仍然重要？Satya 强调，位置对超大规模业务至关重要，因为它涉及多种因素：并非所有工作负载都是异步的，有些需要考虑延迟（latencies）；地理位置必须满足严格的数据主权法（data residency laws），例如欧盟数据边界（EU Data Boundary）要求；同时，还需要将计算、存储（如 Cosmos DB）和会话数据共同部署，以优化经济效益。

Start of Content

[原文] What we have said is that we're in the hyperscale business, which is at the end of the day a long tail business for AI workloads.

[译文] [SN]: 我们说过，我们从事的是超大规模业务，它归根结底是人工智能工作负载的长尾业务。

[原文] And in order to do that, we will have some leading bare-metal-as-a-service capabilities for a set of models, including our own.

[译文] 为了做到这一点，我们将为包括我们自己的模型在内的一系列模型提供一些领先的裸机即服务（bare-metal-as-a-service）能力。

[原文] And that, I think, is the balance you see.

[译文] 我认为，这就是你看到的平衡。

[原文] Another question that comes around this whole fungibility topic.

[译文] [Q]: 围绕整个弹性（fungibility）话题的另一个问题。

[原文] Okay, it's not where you want it, you would

[译文] 好的，它不在您想要的位置，您会

[原文] rather have it in a good population center, like Atlanta.

[译文] 宁愿将它放在一个人气良好的中心，比如亚特兰大。

[原文] We're here.

[译文] 我们就在这里。

[原文] There's also the question of, how much does that matter as the horizon of AI tasks grows?

[译文] 还有一个问题是，随着人工智能任务范围的扩大，这有多重要?

[原文] 30 seconds for a reasoning prompt, or 30 minutes for a deep research, or it's going to be hours for software agents at some point and days and so on and so forth, the time to human interaction.

[译文] 30 秒用于推理提示（reasoning prompt），或 30 分钟用于深度研究，或者在某个时候软件代理需要数小时，甚至数天，与人类交互的时间。

[原文] Why does it matter if it's location A, B, or C?

[译文] 为什么它在位置 A、B 或 C 会有关系?

[原文] It’s a great question.

[译文] [SN]: 这是一个很好的问题。

[原文] That's exactly it.

[译文] 就是这样。

[原文] In fact, that's one of the other reasons why we want to think about what an Azure region looks like and

[译文] 事实上，这也是我们想要思考 Azure 区域是什么样子，以及

[原文] what is the networking between Azure regions.

[译文] Azure 区域之间的网络连接是什么样的原因之一。

[原文] This is where I think as the model capabilities evolve and the usage of these tokens evolves, whether it's synchronously or asynchronously, you don't want to be out of position.

[译文] 在这里，我认为随着模型能力的发展和这些代币使用方式的演变，无论是同步（synchronously）还是异步（asynchronously），你都不希望处于不利地位。

[原文] Then on top of that, by the way, what are the data residency laws?

[译文] 顺便说一句，除此之外，数据驻留法（data residency laws）是什么?

[原文] There’s the entire EU thing, where we literally had to create an EU Data Boundary.

[译文] 存在整个欧盟（EU）的问题，我们实际上不得不创建一个欧盟数据边界（EU Data Boundary）。

[原文] That basically meant that you can't just roundtrip a call to wherever, even if it's asynchronous.

[译文] 这基本上意味着你不能只是将调用往返（roundtrip）到任何地方，即使它是异步的。

[原文] Therefore you need to have maybe regional

[译文] 因此，你可能需要拥有区域性的

[原文] things that are high density, and then the power costs and so on.

[译文] 具备高密度（high density）的东西，以及电力成本等等。

[原文] But you're 100% right in bringing up that the topology as we build out will have to evolve.

[译文] 但你提出我们建设的网络拓扑结构（topology）将不得不演变是 100% 正确的。

[原文] One, for tokens per dollar per watt.

[译文] 首先，是为了每瓦特每美元的代币数量。

[原文] What are the economics?

[译文] 经济效益是什么?

[原文] Overlay that with, what is the usage pattern?

[译文] 将它与使用模式（usage pattern）叠加。

[原文] Usage pattern in terms of synchronous, asynchronous.

[译文] 使用模式包括同步、异步。

[原文] But also what is the compute storage?

[译文] 但也要考虑计算存储是什么?

[原文] Because the latencies may matter for certain things.

[译文] 因为延迟（latencies）对某些事情可能很重要。

[原文] The storage better be there.

[译文] 存储最好在那里。

[原文] If I have a Cosmos DB close to this for session data or even for an autonomous thing, then that also has to be

[译文] 如果我有一个 Cosmos DB 数据库用于会话数据，甚至用于自主事物，那么它也必须

[原文] somewhere close to it, and so on.

[译文] 在它附近，等等。

[原文] All of those considerations are what will shape the hyperscale business.

[译文] 所有这些考量因素都将塑造超大规模业务。

End of Content

章节 17：竞争策略：不追逐容量，聚焦长期业务逻辑

📝 本节摘要：

访谈者指出，由于微软的暂停，Oracle 等竞争对手正在迅速扩张容量，预计到 2027 年可能在总容量上超过微软，尽管 Oracle 的投资回报率（ROIC）和毛利率较低。Satya 重申了微软的立场：不追逐纯粹的容量数字，而是专注于其独特的业务逻辑：避免成为单一模型公司的大型托管商（hoster），即使 Oracle 在此细分市场获得成功。微软的关注点是建立一个能够产生高利润的长尾客户业务，并通过持续投资确保未来 50 年的业务形态。

Start of Content

[原文] Prior to the pause, what we had forecasted for you, by 2028 you were going to be 12–13 gigawatts.

[译文] [Q]: 在暂停之前，我们对您的预测是，到 2028 年，您的容量将达到 12–13 千兆瓦（gigawatts）。

[原文] Now we're at nine and a half or so.

[译文] 现在我们大约在九点五左右。

[原文] But something that's even more relevant—and I just want you to more concretely state that this is the business you don't want to be in—is that Oracle's going from 1/5th your size to bigger than you by the end of 2027.

[译文] 但一个更相关的问题——我希望您更明确地说明这不是您想从事的业务——是 Oracle 的规模将从您的五分之一发展到 2027 年底比您更大。

[原文] While it's not a Microsoft-level quality of

[译文] 虽然这不是微软级别的

[原文] return on invested capital, they're still making 35% gross margins.

[译文] 投资资本回报率（return on invested capital）质量，但他们仍然获得了 35% 的毛利率（gross margins）。

[原文] So the question is, maybe it's not Microsoft's business to do this, but you've created a hyperscaler now by refusing this business, by giving away the right of first refusal, et cetera.

[译文] 所以问题是，也许这不应该是微软的业务，但你们现在通过拒绝这项业务，放弃了优先购买权（right of first refusal）等，创造了一个超大规模服务商（hyperscaler）。

[原文] First of all, I don't want to take away anything from the success Oracle has had in building their business and I wish them well.

[译文] [SN]: 首先，我不想否认 Oracle 在建立其业务上取得的成功，我祝愿他们一切顺利。

[原文] The thing that I think I've answered for you is that it didn't make sense for us to go be a hoster for one model company

[译文] 我认为我已经回答你的问题是，对我们来说，成为一家模型公司的托管商（hoster）是没有意义的

[原文] with limited time horizon RPO.

[译文] 并且其营收承诺期（RPO, Revenue per Purchase Order）有限。

[原文] The thing that you have to think through is not what you do in the next five years, but what you do for the next 50.

[译文] 你必须想清楚的不是你未来五年要做什么，而是你未来五十年要做什么。

[原文] We made our set of decisions.

[译文] 我们做出了自己的决定。

[原文] I feel very good about our OpenAI partnership and what we're doing.

[译文] 我对我们与 OpenAI 的合作以及我们正在做的事情感觉非常好。

[原文] We have a decent book of business.

[译文] 我们有相当规模的业务。

[原文] We wish them a lot of success.

[译文] 我们祝愿他们取得巨大成功。

[原文] In fact, we are buyers of Oracle capacity.

[译文] 事实上，我们也是 Oracle 容量的买家。

[原文] We wish them success.

[译文] 我们祝愿他们成功。

[原文] But at this point, I think the industrial logic for what we are trying to do is pretty clear, which is that it's

[译文] 但在这一点上，我认为我们正在努力的工业逻辑（industrial logic）非常清晰，那就是

[原文] not about chasing…

[译文] 这不是关于追逐……

[原文] First of all, I track, by the way, your things whether it's AWS or Google and ours, which I think is super useful.

[译文] 首先，顺便说一句，我追踪你的分析，无论是关于 AWS、Google 还是我们，我认为这非常有用。

[原文] But it doesn't mean I have to chase those.

[译文] 但这并不意味着我必须去追逐那些。

[原文] I have to chase them for not just the gross margin that they may represent in a period of time.

[译文] 我必须追逐的不仅仅是它们在一段时间内可能代表的毛利率。

[原文] What is this book of business that Microsoft uniquely can go clear, which makes sense for us to clear?

[译文] 微软能够独特地去清理的、对我们来说有意义的业务是什么?

[原文] That's what we'll do.

[译文] 这就是我们将要做的。

End of Content

章节 18：Azure Foundry 的价值：提供全套 AI 工作负载基础设施

📝 本节摘要：

访谈者提出，如果模型公司（如 OpenAI/Anthropic）成为平台，那么企业长尾客户将直接使用他们的服务，Azure 将如何为长尾客户提供直接价值？Satya 回答，一个“真实的”AI 工作负载不仅仅是一个简单的模型 API 调用，它需要完整的计算、存储、数据库（如 Cosmos DB、SQL DB）和安全基础设施。Azure Foundry 的存在就是为了让客户能够在一个地方配置和使用多种模型（OpenAI、开源、Grok 等），并配套完整的 Azure 基础设施，这与单纯向模型公司出售裸机（bare-metal）服务是不同的业务细分。

Start of Content

[原文] I have a question even stepping back from this, I take your point that it's a better business to be in, all else equal, to have a long tail

[译文] [Q]: 我有一个退后一步的问题，我接受您的观点，在其他条件相同的情况下，拥有一个长尾客户群是一个更好的业务

[原文] of customers you can have higher margin from rather than serving bare metal to a few labs.

[译文] 您可以从中获得更高的利润，而不是向少数实验室提供裸机服务。

[原文] But then there's a question of, which way is the industry evolving?

[译文] 但问题是，行业正在朝着哪个方向发展?

[原文] If we believe we're on the path to smarter and smarter AIs, then why isn't the shape of the industry that the OpenAIs and Anthropics and DeepMinds are the platform on which the long tail of enterprises are actually doing business?

[译文] 如果我们相信我们正朝着越来越智能的人工智能发展，那么为什么行业的形态不是 OpenAI、Anthropic 和 DeepMind 成为企业长尾客户实际开展业务的平台?

[原文] They need bare metal, but they are the platform.

[译文] 他们需要裸机，但他们是平台。

[原文] What is the long tail that is directly using Azure?

[译文] 直接使用 Azure 的长尾客户是什么?

[原文] Because you want to use the general cognitive core.

[译文] [SN]: 因为你想要使用通用认知核心（general cognitive core）。

[原文] But those models are all going to be available on Azure, so any workload that says, "Hey, I want to use some open source model and an OpenAI model," if you go to Azure Foundry today, you have all these models that you can provision, buy PTUs, get a Cosmos DB, get a SQL DB, get some storage, get some compute.

[译文] 但这些模型都将在 Azure 上可用，因此任何说“嘿，我想使用一些开源模型和一个 OpenAI 模型”的工作负载，如果你今天去 Azure Foundry，你就可以配置（provision）所有这些模型，购买预留训练单元（PTUs），获得 Cosmos DB，获得 SQL DB，获得一些存储，获得一些计算资源。

[原文] That's what a real workload looks like.

[译文] 真实的（real）工作负载就是这样。

[原文] A real workload is not just an API call to a model.

[译文] 一个真实的工作负载不仅仅是对模型的应用程序编程接口（API）调用。

[原文] A real workload needs all of these things to go build an app or instantiate an application.

[译文] 一个真实的工作负载需要所有这些东西来构建一个应用程序（app）或实例化（instantiate）一个应用程序。

[原文] In fact, the model companies need that to build anything.

[译文] 事实上，模型公司构建任何东西都需要这些。

[原文] It's not just like, "I have a token factory."

[译文] 它不仅仅是：“我有一个代币工厂”。

[原文] I have to have all of these things.

[译文] 我必须拥有所有这些东西。

[原文] That's the hyperscale business.

[译文] 这就是超大规模业务。

[原文] And it's not on any one model, but all of these models.

[译文] 而且它不依赖于任何一个模型，而是所有这些模型。

[原文] So if you want Grok plus, say, OpenAI plus an open source model, come to Azure Foundry, provision them, build your application.

[译文] 所以如果你想要 Grok 加上，比如说，OpenAI 加上一个开源模型，来到 Azure Foundry，配置它们，构建你的应用程序。

[原文] Here is a database.

[译文] 这是一个数据库。

[原文] That's kind of what the business is.

[译文] 这就是业务的样子。

[原文] There is a separate business called just selling raw bare-metal services to model companies.

[译文] 存在一个单独的业务，叫做只向模型公司出售原始的裸机服务（raw bare-metal services）。

[原文] And that's the argument about how much of that business you want to be in and not be in and what that is.

[译文] 这就是关于你想在多大程度上参与该业务、不想参与以及该业务是什么的争论。

[原文] It's a very different segment of the business, which we are in, and we also have limits to how much of it is going to crowd out the rest of it.

[译文] 它是业务中一个非常不同的细分市场，我们参与其中，但我们也限制了它将挤占（crowd out）我们其余业务的空间。

[原文] But that's kind of at least the way I look at it.

[译文] 但至少这就是我的看法。

End of Content

章节 19：重新审视暂停：地理位置、摩尔定律与执行速度

📝 本节摘要：

访谈者再次追问暂停建设的原因，指出微软本可以自己使用那 3.5GW 的容量。Satya 澄清，暂停不是因为不想建设，而是因为需要以不同的方式建设，必须考虑工作负载类型、地理位置（geo-type）和时间安排。由于数据主权和监管需求，微软必须在全球（阿联酋、印度、欧洲）建设容量。最关键的学习是：需要乘着摩尔定律（Moore's law）的浪潮前进，避免在单一芯片代次上投入过多资本并面临四年或五年的折旧。微软现在强调光速执行（speed-of-light execution），将建设周期压缩到 90 天，从而能够在每一代芯片迭代中进行扩展，而不是造成巨大的产能不平衡。

Start of Content

[原文] There are sort of two questions here.

[译文] [Q]: 这里有两类问题。

[原文] One is, why couldn't you just do both?

[译文] 一是，为什么不能两者兼顾?

[原文] The other one is, given our estimates on what your capacity is in 2028, it's three and a half gigawatts lower.

[译文] 另一个是，根据我们对您 2028 年容量的估计，它低了 3.5 千兆瓦。

[原文] Sure, you could have dedicated that to OpenAI training and inference capacity, but you could have also dedicated that to actually just running Azure, running Microsoft 365, running GitHub Copilot.

[译文] 当然，您可以将这些容量专用于 OpenAI 的训练和推理，但您也可以将其专门用于运行 Azure、Microsoft 365、GitHub Copilot。

[原文] I could have just built it and not given it to OpenAI.

[译文] [SN]: 我可以建造它而不提供给 OpenAI。

[原文] Or I may want to build it in a different location.

[译文] 或者我可能想在不同的位置建造它。

[原文] I may want to build it in the UAE, I may want to build it in India, I may want to build it in Europe.

[译文] 我可能想在阿联酋（UAE）建造它，我可能想在印度建造它，我可能想在欧洲建造它。

[原文] One of the things is, as I said, where we have real capacity constraints right now, given the regulatory needs and the data sovereignty needs, we've got to build all over the world.

[译文] 我说过，其中一件事是，考虑到监管需求和数据主权（data sovereignty）需求，我们目前在某些地方存在真正的容量限制，我们必须在世界各地进行建设。

[原文] First of all, stateside capacity is super important, and we want to build everything.

[译文] 首先，美国本土的容量超级重要，我们想建造所有东西。

[原文] But when I look out to 2030, I have a global view of what is Microsoft's shape of business by first-party and third-party.

[译文] 但当我展望 2030 年时，我对微软的第一方和第三方业务形态有一个全球性的看法。

[原文] Third-party segmented by the frontier labs and how much they want versus the inference capacity we want to build for multiple models, and our own research compute needs.

[译文] 第三方业务按前沿实验室及其需求量，与我们希望为多个模型构建的推理容量，以及我们自己的研究计算需求来划分。

[原文] That's all going into my calculus.

[译文] 所有这些都纳入了我的考量。

[原文] You're rightfully pointing out the pause, but the pause was not done because we said,

[译文] 你指出暂停是正确的，但暂停不是因为我们说，

[原文] "Oh my God, we don't want to build that."

[译文] “天哪，我们不想建造那个”。

[原文] We realized that we want to build what we want to build slightly differently by both workload type as well as geo-type and timing as well.

[译文] 我们意识到我们想要建造的东西，在工作负载类型、地理类型和时间安排上，需要略有不同。

[原文] We'll keep ramping up our gigawatts, and the question is at what pace and in what location.

[译文] 我们会继续增加我们的千兆瓦（gigawatts）容量，问题是以什么速度和在什么位置。

[原文] And how do I ride Moore's law on it, which is, do I really want to overbuild three and a half in 2027 or do I want to spread that in 2027-28 knowing even…

[译文] 以及我如何顺应摩尔定律（Moore's law），也就是说，我真的想在 2027 年过度建设 3.5GW，还是想在 2027-28 年分散建设，甚至考虑到……

[原文] One of the biggest learnings we had even with Nvidia is that their pace increased in terms of their migrations.

[译文] 即使对于英伟达（Nvidia），我们最大的经验教训之一是他们在迁移方面的速度加快了。

[原文] That was a big factor.

[译文] 这是一个重要因素。

[原文] I didn't want to go get stuck for four or five years of depreciation on one generation.

[译文] 我不想在一代芯片上陷入四五年的折旧（depreciation）。

[原文] In fact, Jensen's advice to me was two things.

[译文] 事实上，黄仁勋（Jensen）给我的建议有两点。

[原文] One is, get on the speed-of-light execution.

[译文] 一是，实现光速执行（speed-of-light execution）。

[原文] That's why the execution in this Atlanta data center....

[译文] 这就是为什么亚特兰大这个数据中心的执行……

[原文] I mean, it's like 90 days between when we get it and to hand off to a real workload.

[译文] 我的意思是，从我们获得它到交付给一个真实工作负载，只有大约 90 天。

[原文] That's real speed-of-light execution on that front.

[译文] 在这方面，那是真正的光速执行。

[原文] I wanted to get good on that.

[译文] 我想在这方面做得好。

[原文] And then that way I'm building each generation in scaling.

[译文] 然后通过这种方式，我在扩展时建设每一代芯片。

[原文] And then every five years,

[译文] 然后每五年，

[原文] you have something much more balanced.

[译文] 你会拥有一个更加平衡的东西。

[原文] So it becomes literally like a flow for a large-scale industrial operation like this where you're suddenly not lopsided, where you've built up a lot in one time and then you take a massive hiatus because you're stuck with all this, to your point, in one location which may be great for training, or it may not be great for inference because I can't serve, even if it's all asynchronous, because Europe won't let me round-trip to Texas.

[译文] 因此，对于像这样的大型工业运营来说，它实际上变成了一种流（flow），你不会突然失衡（lopsided），比如你一次性建造了大量产能，然后不得不进行大规模停顿，因为你像你所说的那样，被困在某个可能非常适合训练的位置，但可能不适合推理（inference），因为我无法提供服务，即使都是异步的，因为欧洲不会允许我往返于德克萨斯州。

[原文] So that's all of the things.

[译文] 所有这些都是考量因素。

End of Content

章节 20：租用外部容量：新云服务商的整合与 Azure 生态的互补

📝 本节摘要：

访谈者指出微软近期宣布了与 Iris Energy、Nebius 和 Lambda Labs 等“新云服务商”（neoclouds）合作，租用外部容量。Satya 认为这是应对市场需求的灵活策略：在微软自身容量不足时，可以采取租赁或 GPU 即服务（GPUs-as-a-service）的方式。他甚至欢迎这些新云服务商将容量带入 Azure Marketplace，因为客户使用他们的容量时，仍会同时消费 Azure 的核心服务（计算、存储、数据库），从而实现共赢。

Start of Content

[原文] How do I rationalize this statement with what you've done over the last few weeks?

[译文] [Q]: 我如何将您的说法与您过去几周所做的事情合理化?

[原文] You've announced deals with Iris Energy, with Nebius, and Lambda Labs, and there's a few more coming as well.

[译文] 你们宣布了与 Iris Energy、Nebius 和 Lambda Labs 的交易，并且还有一些即将到来。

[原文] You're going out there and securing capacity that you're renting from the neoclouds rather than having built it yourself.

[译文] 你们正在从新云服务商（neoclouds）那里租用容量，而不是自己建设。

[原文] It's fine for us because now when you have line of sight to demand, which can be served where people are building, it's great.

[译文] [SN]: 对我们来说没问题，因为现在当你能够预见需求，并且这些需求可以在其他人正在建设的地方得到服务时，这很棒。

[原文] In fact we will take leases, we will take build-to-suit, we'll even take GPUs-as-a-service where we don't have capacity

[译文] 事实上，我们将进行租赁（leases），我们将进行定制建造（build-to-suit），甚至会接受我们没有容量时的图形处理器即服务（GPUs-as-a-service）。

[原文] but we need capacity and someone else has that.

[译文] 但我们需要容量，而其他人拥有它。

[原文] And by the way, I would even sort of welcome every neocloud to just be part of our marketplace.

[译文] 顺便说一句，我甚至会欢迎每一个新云服务商（neocloud）成为我们市场（marketplace）的一部分。

[原文] Because guess what?

[译文] 因为你猜怎么着?

[原文] If they go bring their capacity into our marketplace, that customer who comes through Azure will use the neocloud, which is a great win for them, and will use compute, storage, databases, all the rest from Azure.

[译文] 如果他们将容量带到我们的市场中，通过 Azure 来的客户将使用新云服务商的服务，这对他们来说是一个巨大的胜利，并且客户将使用来自 Azure 的计算、存储、数据库以及所有其他服务。

[原文] So I'm not at all thinking of this as, "Hey, I should just go gobble up all of that myself."

[译文] 所以我根本没有想过：“嘿，我应该自己把所有这些都吞掉”。

End of Content

章节 21：内部芯片策略与 OpenAI IP 的全访问权限

📝 本节摘要：

访谈者质疑微软的内部芯片（internal chips，如 Maia）订单量远低于 Google 和 Amazon，询问其内部芯片计划进展。Satya 认为，新加速器最大的竞争者是英伟达的前一代芯片。微软的策略是通过多种供应商（Intel、AMD、Cobalt/Maia）保持计算集群的平衡。他强调，内部芯片成功的“与生俱来的权利”（birthright）在于与 MAI 模型紧密耦合的设计。最重要的是，微软拥有对 OpenAI IP 的全部访问权限（All of it），除了消费者硬件。这种独家权利使微软无需进行重复性工作，可以专注于与 OpenAI 系统的设计、扩展和创新，同时自身致力于成为英伟达的“光速执行合作伙伴”。

Start of Content

[原文] You mentioned how this depreciating asset, in five or six years, is 75% of the TCO of a data center.

[译文] [Q]: 您提到这种折旧资产，在五到六年内，占数据中心总拥有成本（TCO）的 75%。

[原文] And Jensen is taking a 75% margin on that.

[译文] 而黄仁勋（Jensen）从中获得了 75% 的利润。

[原文] So what all the hyperscalers are trying to do is develop their own accelerator so that they can reduce this overwhelming cost for equipment, to increase their margins.

[译文] 因此，所有超大规模服务商都在努力开发自己的加速器（accelerator），以便减少这种压倒性的设备成本，从而提高利润。

[原文] And when you look at where they are, Google's way ahead of everyone else.

[译文] 当你看看他们的进展时，Google 遥遥领先于其他所有人。

[原文] They've been doing it for the longest.

[译文] 他们做这件事的时间最长。

[原文] They're going to make something like five to seven million chips of their own TPUs.

[译文] 他们将生产大约五到七百万颗自己的张量处理器（TPUs）芯片。

[原文] You look at Amazon and they're trying to make three to five million [Lifetime shipment units].

[译文] 你看看亚马逊，他们正试图生产三到五百万颗（终身出货量）。

[原文] But when we look at what Microsoft is ordering of their own chips, it's way below that number.

[译文] 但是当我们看微软自己芯片的订单量时，这个数字远低于此。

[原文] You've had a program for just as long.

[译文] 你们的项目运行时间也一样长。

[原文] What's going on with your internal chips?

[译文] 你们的内部芯片有什么进展?

[原文] It’s a good question.

[译文] [SN]: 这是一个很好的问题。

[原文] A couple of things.

[译文] 有几件事。

[原文] One is that the thing that is the biggest competitor for any new accelerator is kind of even the previous generation of Nvidia.

[译文] 一是任何新的加速器最大的竞争对手，甚至可以说是上一代英伟达芯片。

[原文] In a fleet, what I'm going to look at is the overall TCO.

[译文] 在一个集群（fleet）中，我要看的是整体总拥有成本（TCO）。

[原文] The bar I have, even for our own…

[译文] 即使对于我们自己的芯片，我设定的门槛……

[原文] By the way, I was just looking at the data for Maia 200 which looks great, except that one of the things that we learned even on the

[译文] 顺便说一句，我刚刚看了 Maia 200 的数据，看起来很棒，除了我们学到的一个教训，即使在

[原文] compute side…

[译文] 计算方面……

[原文] We had a lot of Intel, then we introduced AMD, and then we introduced Cobalt.

[译文] 我们曾大量使用英特尔（Intel），然后我们引入了 AMD，接着我们引入了 Cobalt。

[原文] That's how we scaled it.

[译文] 这就是我们扩展它的方式。

[原文] We have good existence proof of, at least in core compute, how to build your own silicon and then manage a fleet where all three are at play in some balance.

[译文] 我们拥有很好的证明，至少在核心计算方面，如何构建自己的芯片（silicon），然后管理一个让这三者在某种平衡中发挥作用的集群。

[原文] Because by the way, even Google's buying Nvidia, and so is Amazon.

[译文] 顺便说一句，即使是 Google 也在购买英伟达的芯片，亚马逊也是如此。

[原文] It makes sense because Nvidia is innovating and it's the general-purpose thing.

[译文] 这是有道理的，因为英伟达正在创新，而且它是通用（general-purpose）的东西。

[原文] All models run on it and customer demand is there.

[译文] 所有模型都在上面运行，并且有客户需求。

[原文] Because if you build your own vertical thing,

[译文] 因为如果你构建自己的垂直（vertical）产品，

[原文] you better have your own model, which is either going to use it for training or inference, and you have to generate your own demand for it or subsidize the demand for it.

[译文] 你最好拥有自己的模型，要么将其用于训练或推理，要么你必须为它创造自己的需求，或者补贴（subsidize）它的需求。

[原文] So therefore you want to make

[译文] 因此，你需要确保

[原文] sure you scale it appropriately.

[译文] 你适当地扩展它。

[原文] The way we are going to do it is to have a close loop between our own MAI models and our silicon, because I feel like that's what gives you the birthright to do your own silicon, where you literally have designed the microarchitecture with what you're doing,

[译文] 我们将要做的方式是在我们自己的 MAI 模型和我们的芯片之间建立一个闭环（close loop），因为我认为这才是你做自己芯片的与生俱来的权利（birthright），在这里，你确实是根据你正在做的事情来设计微架构（microarchitecture）的，

[原文] and then you keep pace with your own models.

[译文] 然后你跟上你自己的模型。

[原文] In our case, the good news here is that OpenAI has a program which we have access to.

[译文] 在我们这个案例中，好消息是 OpenAI 有一个我们可以访问的项目。

[原文] So therefore to think that Microsoft is not going to have something that's—

[译文] 因此，认为微软不会拥有某种——

[原文] What level of access do you have to that?

[译文] [Q]: 你们对那个项目的访问级别是什么?

[原文] All of it.

[译文] [SN]: 全部。

[原文] You just get the IP for all of that?

[译文] [Q]: 你们获得了所有这些的知识产权（IP）?

[原文] So the only IP you don't have is consumer hardware?

[译文] 那么你们唯一没有的知识产权是消费者硬件（consumer hardware）?

[原文] That's it.

[译文] [SN]: 就是这样。

[原文] Oh, okay. Interesting.

[译文] [Q]: 哦，好的。有趣。

[原文] By the way, we gave them a bunch of IP as well to bootstrap them.

[译文] [SN]: 顺便说一句，我们也给了他们一堆知识产权（IP）来引导（bootstrap）他们。

[原文] This is one of the reasons why they…

[译文] 这就是他们……的原因之一

[原文] Because we built all these supercomputers together.

[译文] 因为我们一起构建了所有这些超级计算机。

[原文] We built it for them and they benefited from it, rightfully so.

[译文] 我们为他们建造了它，他们从中受益，这是理所当然的。

[原文] And now as they innovate, even at the system level, we get access to all of it.

[译文] 现在，当他们创新时，即使是在系统层面，我们也能获得所有这些的访问权。

[原文] And we first want to instantiate what they build, for them, but then we'll extend it.

[译文] 我们首先想要为他们实例化（instantiate）他们构建的东西，但随后我们会进行扩展。

[原文] So if anything, the way I think about your question is, Microsoft wants to be a fantastic, I'll call it, speed-of-light execution partner for Nvidia.

[译文] 因此，如果非要说的话，我对你的问题的看法是，微软希望成为英伟达出色的、我称之为“光速执行”（speed-of-light execution）的合作伙伴。

[原文] Because quite frankly that fleet is life itself.

[译文] 因为坦白地说，那个集群（fleet）就是生命本身。

[原文] Obviously Jensen's doing super well with his margins, but the TCO has many dimensions to it and I want to be great at that TCO.

[译文] 显然黄仁勋在利润方面做得非常好，但总拥有成本（TCO）有很多方面，我想在总拥有成本方面做到最好。

End of Content

章节 22：OpenAI 独家协议的范围：PaaS API 的排他性

📝 本节摘要：

访谈者要求澄清微软与 OpenAI 新协议中关于“无状态 API 调用”（stateless API calls）的独家权利。Satya 澄清了战略决定：OpenAI 的平台即服务（PaaS）API 业务对 Azure 具有排他性（exclusive）。这意味着任何需要使用 OpenAI 无状态 API 的合作伙伴（如 Salesforce）都需要通过 Azure 运行。然而，OpenAI 的软件即服务（SaaS）业务（如 ChatGPT）则可以在任何地方运行，并且他们可以自由选择其他合作伙伴来构建 SaaS 产品。

Start of Content

[原文] On top of that, I want to be able to really work with the OpenAI lineage and the MAI lineage and the system design, knowing that we have the IP rights on both ends.

[译文] [SN]: 最重要的是，我希望能够真正与 OpenAI 谱系和 MAI 谱系以及系统设计合作，因为我们知道我们在两端都拥有知识产权（IP rights）。

[原文] Speaking of rights, you had an interview a couple days ago where you said that in the new agreement you made with OpenAI you have rights, the exclusivity, to the stateless API calls that OpenAI makes.

[译文] [Q]: 谈到权利，您几天前接受采访时说，您与 OpenAI 签订的新协议中，您拥有对 OpenAI 进行的无状态应用程序编程接口（stateless API）调用的独家权利（exclusivity）。

[原文] We were sort of confused about if there's any state whatsoever.

[译文] 我们有点困惑，如果存在任何状态（state）会怎么样。

[原文] You were just mentioning a second ago that all these complicated workloads that are coming up are going to require memory and

[译文] 您刚才提到，所有这些即将出现的复杂工作负载都需要内存和

[原文] databases and storage and so forth.

[译文] 数据库和存储等等。

[原文] Is that now not stateless if ChatGPT is storing stuff on sessions?

[译文] 如果 ChatGPT 正在存储会话（sessions）中的内容，那现在是不是就不是无状态的了?

[原文] That's the reason why.

[译文] [SN]: 这就是原因。

[原文] The strategic decision we made, and also accommodating for the flexibility OpenAI needed in order to be able to procure compute for…

[译文] 我们做出的战略决策，同时也顾及了 OpenAI 为了采购计算资源而需要的灵活性……

[原文] Essentially think of OpenAI having a PaaS business and a SaaS business.

[译文] 本质上，你可以把 OpenAI 看作是拥有一个平台即服务（PaaS）业务和一个软件即服务（SaaS）业务。

[原文] The SaaS business is ChatGPT.

[译文] 软件即服务（SaaS）业务是 ChatGPT。

[原文] Their PaaS business is their API.

[译文] 他们的平台即服务（PaaS）业务是他们的应用程序编程接口（API）。

[原文] That API is Azure-exclusive.

[译文] 那个应用程序编程接口（API）是 Azure 独家（Azure-exclusive）的。

[原文] The SaaS business, they can run it anywhere.

[译文] 软件即服务（SaaS）业务，他们可以在任何地方运行。

[原文] And they can partner with anyone they

[译文] [Q]: 他们可以与任何他们

[原文] want to to build SaaS products?

[译文] 想要的人合作来构建 SaaS 产品吗?

[原文] If they want a partner and that partner wants to use a stateless API, then Azure is the place where they can get the stateless API.

[译文] [SN]: 如果他们想要一个合作伙伴，而那个合作伙伴想要使用一个无状态应用程序编程接口（stateless API），那么 Azure 就是他们可以获得无状态 API 的地方。

[原文] It seems like there's a way for them to build the product together and it's a stateful thing…

[译文] [Q]: 看起来他们有一种可以一起构建产品的方式，而且它是一个有状态（stateful）的东西……

[原文] No, for even that they'll have to come to Azure.

[译文] [SN]: 不，即使对于那个，他们也必须来到 Azure。

[原文] Again, this is done in the spirit of "what is it that we value as part of our partnership."

[译文] 重申一下，这是本着“作为我们合作关系的一部分，我们重视什么”的精神完成的。

[原文] And we made sure that, at the same time, we were good partners to OpenAI given all the flexibility they needed.

[译文] 同时，我们确保在给予 OpenAI 所需灵活性的情况下，我们也是他们的好伙伴。

[原文] So for example, Salesforce wants to integrate OpenAI.

[译文] [Q]: 那么例如，Salesforce 想要集成 OpenAI。

[原文] It's not through an API.

[译文] 它不是通过一个应用程序编程接口（API）。

[原文] They actually work together, train a model together and deploy it on, let's say, Amazon now.

[译文] 他们实际上是共同合作，一起训练一个模型，然后部署到，比如说，亚马逊（Amazon）上。

[原文] Is that allowed or do they have to use your…

[译文] 这是被允许的吗，还是他们必须使用您的……

[原文] For any custom agreement like that, they will have to come run it…

[译文] [SN]: 对于任何像那样的定制协议（custom agreement），他们必须在我们的平台上运行……

[原文] There are some few exceptions, the US government and so on, that we made, but other than that, they'd have to come to Azure.

[译文] 我们做了一些少数例外，比如美国政府等，但除此之外，他们必须使用 Azure。

End of Content

章节 23：从软件公司到知识密集型工业企业的转变

📝 本节摘要：

访谈者指出微软的资本支出（capex）在两年内翻了三倍，标志着公司正在从软件业务向资本密集型工业业务转型。Satya 同意这种结构性变化是巨大的，并将微软描述为“资本密集型和知识密集型”的业务。他强调，微软必须利用其知识（即软件优化能力）来提高投资资本回报率（ROIC），通过软件改进实现 5 倍、10 倍甚至 40 倍的代币吞吐量提升。这种软件工程能力是超大规模服务商与传统托管商的根本区别，也是微软能持续发展的基础。

Start of Content

[原文] Stepping back, when we were walking back and forth through the factory, one of the things you were talking about is that Microsoft,

[译文] [Q]: 退一步说，当我们来回走过工厂时，您谈到的一件事是，微软

[原文] you can think of it as a software business, but now it's really becoming an industrial business.

[译文] 你可以把它看作是一个软件业务，但现在它真的正在变成一个工业业务（industrial business）。

[原文] There's all this capex, there's all this construction.

[译文] 存在所有这些资本支出（capex），存在所有这些建设。

[原文] If you just look over the last two years, your capex has sort of tripled.

[译文] 如果你看看过去两年，你的资本支出（capex）几乎翻了三倍。

[原文] Maybe you extrapolate that forward, it actually just becomes this huge industrial explosion.

[译文] 也许你推断下去，它实际上就变成了这种巨大的工业爆炸。

[原文] Other hyperscalers are taking loans.

[译文] 其他超大规模服务商正在贷款。

[原文] Meta has done a $20 billion loan at Louisiana.

[译文] Meta 在路易斯安那州进行了一笔 200 亿美元的贷款。

[原文] They've done a corporate loan.

[译文] 他们做了一笔公司贷款。

[原文] It seems clear everyone's free cash flow is going to zero,

[译文] 似乎很明显，所有人的自由现金流（free cash flow）都将归零，

[原文] which I'm sure Amy is going to beat you up if you even try to do that, but what's happening?

[译文] 我确信如果您尝试那样做，艾米（Amy，指 CFO）会痛批您，但到底发生了什么?

[原文] I think the structural change is what you're referencing, which is massive.

[译文] [SN]: 我认为您指的是结构性变化，这是巨大的。

[原文] I describe it as we are now a capital-intensive business and a knowledge-intensive business.

[译文] 我将其描述为我们现在是一个资本密集型（capital-intensive）业务和知识密集型（knowledge-intensive）业务。

[原文] In fact, we have to use our knowledge to increase the ROIC on the capital spend.

[译文] 事实上，我们必须利用我们的知识来提高资本支出（capital spend）上的投资资本回报率（ROIC）。

[原文] The hardware guys have done a great job of marketing Moore's Law, which I think is unbelievable and it's great.

[译文] 硬件公司在推广摩尔定律（Moore's Law）方面做得非常出色，我认为这令人难以置信，非常棒。

[原文] But if you even look at some of the stats I even did in my earnings call, for a given GPT family,

[译文] 但如果你看看我在财报电话会议上提到的一些数据，对于一个特定的 GPT 家族，

[原文] the software improvements of really throughput in terms of tokens-per-dollar-per-watt that we're able to get quarter-over-quarter, year-over-year, it’s massive.

[译文] 我们能够获得的、季度环比和年度环比的软件改进，在每瓦特每美元的代币吞吐量（tokens-per-dollar-per-watt）方面，是巨大的。

[原文] It's 5x, 10x, maybe 40x in some of these cases, just because of how you can optimize.

[译文] 在某些情况下，它可以是 5 倍、10 倍，甚至可能是 40 倍，仅仅是因为你可以如何优化。

[原文] That's knowledge intensity coming to bring out capital efficiency.

[译文] 那就是知识密集度（knowledge intensity）带来资本效率（capital efficiency）。

[原文] That, at some level, is what we have to master.

[译文] 在某种程度上，这就是我们必须掌握的。

[原文] Some people ask me, what is the difference between a classic old-time hoster and a hyperscaler?

[译文] 有些人问我，一个经典的旧式托管商（hoster）和一个超大规模服务商（hyperscaler）有什么区别?

[原文] Software.

[译文] 软件。

[原文] Yes, it is capital intensive,

[译文] 是的，它是资本密集型的，

[原文] but as long as you have systems know-how, software capability to optimize by workload, by fleet...

[译文] 但只要你拥有系统专有技术（systems know-how），以及根据工作负载、根据集群（fleet）进行优化的软件能力……

[原文] That's why when we say fungibility, there's so much software in it.

[译文] 这就是为什么当我们说弹性（fungibility）时，其中包含了如此多的软件。

[原文] It's not just about the fleet.

[译文] 这不仅仅关乎集群。

[原文] It's the ability to evict a workload and then schedule another workload.

[译文] 它是驱逐（evict）一个工作负载，然后调度（schedule）另一个工作负载的能力。

[原文] Can I manage that algorithm of scheduling around?

[译文] 我能管理那个调度的算法吗?

[原文] That is the type of stuff that we have to be world-class at.

[译文] 这就是我们必须做到世界一流的方面。

[原文] So yes, I think we'll still remain a software company, but yes, this is a different business and we're going to manage.

[译文] 所以，是的，我认为我们仍将是一家软件公司，但是，是的，这是一个不同的业务，我们将进行管理。

[原文] At the end of the day,

[译文] 归根结底，

[原文] the cash flow that Microsoft has allows us to have both these arms firing well.

[译文] 微软拥有的现金流（cash flow）使我们能够让这两个方面都良好运作。

End of Content

章节 24：研发计算的资本支出与市场预测

📝 本节摘要：

鉴于 AGI 的长期前景，Satya 建议将超智能研究计算（research compute）视为研发（R&D）费用来核算。除了研发投入外，其余资本支出都应由需求驱动，并需制定合理的预测计划。他承认，领先的实验室（如 OpenAI 和 Anthropic）为了筹集计算资源而做出的市场收入预测是合理的。微软作为领先的研发公司，必须在人才和计算资源上投入大量资金，确保高研究员/GPU 比率，并具有在尚未成为传统智慧之前进行扩张的资产负债表能力。

Start of Content

[原文] It seems like in the short term you have more credence on things taking a while, being more jagged.

[译文] [Q]: 看起来在短期内，您更相信事情需要一些时间，发展会更不均衡（jagged）。

[原文] But maybe in the long term you think the people who talk about AGI and ASI are correct.

[译文] 但也许从长远来看，您认为谈论通用人工智能（AGI）和超级人工智能（ASI）的人是正确的。

[原文] Sam will be right, eventually.

[译文] 萨姆（Sam）最终会是对的。

[原文] I have a broader question about what makes sense for a hyperscaler to do, given that you have to invest massively in this thing which depreciates over five years.

[译文] 我有一个更宏大的问题：鉴于您必须对这个在五年内折旧的东西进行大规模投资，超大规模服务商（hyperscaler）做什么才有意义?

[原文] So if you have 2040 timelines to the kind of

[译文] 因此，如果您对未来的时间表是 2040 年

[原文] thing that somebody like Sam anticipates in three years, what is a reasonable thing for you to do in that world?

[译文] 而像萨姆（Sam）这样的人预期在三年内会发生，那么在这个世界中，您做什么才是合理的?

[原文] There needs to be an allocation to, I'll call it, research compute.

[译文] [SN]: 需要为我称之为研究计算（research compute）的东西分配资源。

[原文] That needs to be done like you did R&D.

[译文] 这需要像你进行研发（R&D）一样来做。

[原文] That's the best way to even account for it, quite frankly.

[译文] 坦白地说，这是最好的核算方式。

[原文] We should think of it as just R&D expense and you should say, "What's the research compute and how do you want to scale it?"

[译文] 我们应该将它视为研发（R&D）费用，并且您应该说：“研究计算是什么，您想如何扩展它”?

[原文] Let's even say it's an order of magnitude scale in some period.

[译文] 让我们假设在某个时期内，它的规模是量级上的扩展。

[原文] Pick your thing, is it two years?

[译文] 选择你所做的，是两年吗?

[原文] Is it 16 months? What have you.

[译文] 是 16 个月吗？等等。

[原文] That's sort of one piece, which is table stakes, that's R&D expenses.

[译文] 这是一部分，是基本要求（table stakes），是研发（R&D）费用。

[原文] The rest is all demand driven.

[译文] 其余的都是需求驱动的。

[原文] Ultimately, you're allowed to build ahead of demand, but you better have a demand plan that doesn't go completely off kilter.

[译文] 最终，你被允许超前于需求进行建设，但你最好有一个不会完全失衡（off kilter）的需求计划。

[原文] Do you buy…

[译文] [Q]: 您会相信……

[原文] These labs are now projecting revenues of $100 billion in 2027–28 and they're projecting revenue to keep growing at this rate of 3x, 2x a year…

[译文] 这些实验室现在预测 2027-28 年的收入将达到 1000 亿美元，并且他们预测收入将保持每年 3 倍、2 倍的速度增长……

[原文] In the marketplace there's all kinds of incentives right now, and rightfully so.

[译文] [SN]: 现在市场上有各种各样的激励因素，这是理所当然的。

[原文] What do you expect an independent lab that is sort of trying to raise money to do?

[译文] 你期望一个试图筹集资金的独立实验室做什么?

[原文] They have to put some numbers out there such that they can actually go raise money so that they can pay their bills for compute and what have you.

[译文] 他们必须拿出一些数字，以便他们能够实际去筹集资金，从而支付他们的计算费用等等。

[原文] And it's a good thing.

[译文] 这是件好事。

[原文] Someone's going to take some risk and put it in there, and they've shown traction.

[译文] 总有人会承担一些风险并投入其中，而且他们已经展示了吸引力（traction）。

[原文] It's not like it's all risk without seeing the fact that they've been performing, whether it's OpenAI, or whether it's Anthropic.

[译文] 这并不是说没有看到他们一直在表现的事实就全是风险，无论是 OpenAI 还是 Anthropic。

[原文] So I feel great about what they've done,

[译文] 所以我对他们所做的一切感觉非常好，

[原文] and we have a massive book of business with these chaps.

[译文] 我们与这些伙伴有着大量的业务往来。

[原文] So therefore that's all good.

[译文] 因此，一切都很好。

[原文] But overall ultimately, there's two simple things.

[译文] 但总的来说，最终只有两件简单的事情。

[原文] One is you have to allocate for R&D.

[译文] 一是你必须为研发（R&D）分配资金。

[原文] You brought up talent.

[译文] 你提到了人才。

[原文] The talent for AI is at a premium.

[译文] 人工智能人才处于溢价（at a premium）。

[原文] You have to spend there.

[译文] 你必须在那方面花钱。

[原文] You've got to spend on compute.

[译文] 你必须在计算上花钱。

[原文] So in some sense researcher-to-GPU ratios have to be high.

[译文] 所以从某种意义上说，研究员与图形处理器（researcher-to-GPU）的比率必须很高。

[原文] That is sort of what it takes to be a leading R&D company in this world.

[译文] 这就是在这个世界上成为一家领先的研发公司所需要的。

[原文] And that's something that needs to scale, and you have to have a balance sheet that allows you to scale that long before it's conventional wisdom and so on.

[译文] 这是一件需要扩展的事情，你必须拥有一张允许你在它成为传统智慧（conventional wisdom）之前很久就进行扩展的资产负债表（balance sheet）。

End of Content

章节 25：地缘政治转向与美国科技堆栈的全球信任建设

📝 本节摘要：

访谈进入地缘政治环节，指出全球正从 90 年代的单极世界转向双极/多极世界，各国都在追求“主权 AI”（Sovereign AI）。Satya 认为，美国科技行业和政府的最高优先事项是确保全球对美国技术堆栈的信任。他指出美国科技的全球市场份额建立在信任之上。他呼吁美国政府为美国公司在世界各地的外国直接投资（FDI）提供支持和认可，因为 AI 工厂的建设是美国在全球营销的最佳方式。微软则通过政策和技术（如欧盟数据边界、Azure 上的主权服务、机密计算）来满足各国的数据主权和隐私担忧，以此构建信任。

Start of Content

[原文] That's kind of one thing. But the other is all about knowing how to forecast.

[译文] [SN]: 这是一件事。但另一件事是关于知道如何预测。

[原文] As we look across the world, America has dominated many tech stacks.

[译文] [Q]: 当我们放眼全球时，美国主导了许多技术堆栈（tech stacks）。

[原文] The US owns Windows through Microsoft, which is deployed even in China, that's the main operating system.

[译文] 美国通过微软拥有 Windows，它甚至在中国也有部署，那是主要的操作系统。

[原文] Of course, there's Linux, which is open source, but Windows is deployed everywhere in China on personal computers.

[译文] 当然，有 Linux，它是开源的，但 Windows 在中国的个人电脑上随处可见。

[原文] You look at Word, it's deployed everywhere.

[译文] 你看 Word，它随处可见。

[原文] You look at all these various technologies, it's deployed everywhere.

[译文] 你看所有这些各种技术，它们随处可见。

[原文] And Microsoft and other

[译文] 微软和其他

[原文] companies have grown elsewhere.

[译文] 公司在其他地方也实现了增长。

[原文] They're building data centers in Europe and in India and in all these other places, in Southeast Asia and LatAm and Africa.

[译文] 他们正在欧洲、印度以及所有其他地方，在东南亚、拉丁美洲和非洲建设数据中心。

[原文] In all of these different places, you're building capacity.

[译文] 在所有这些不同的地方，你们都在建设容量。

[原文] But this seems quite different.

[译文] 但这看起来相当不同。

[原文] Today, the political aspect of technology, of compute…

[译文] 今天，技术和计算的政治方面……

[原文] The US administration didn't care about the dot-com bubble.

[译文] 美国政府不关心互联网泡沫。

[原文] It seems like the US administration, as well as every other administration around the world, cares a lot about AI.

[译文] 似乎美国政府，以及世界上其他所有政府，都非常关心人工智能。

[原文] The question is, we're sort of in a bipolar world, at least with the US and China, but Europe and

[译文] 问题是，我们有点处于一个两极世界（bipolar world），至少在美国和中国之间，但欧洲和

[原文] India and all these other countries are saying, "No, we're going to have sovereign AI as well."

[译文] 印度以及所有其他国家都在说：“不，我们也要拥有主权人工智能（sovereign AI）”。

[原文] How does Microsoft navigate the difference to the 90s—where there's one country in the world that matters, it's America, and our companies sell everywhere and therefore Microsoft benefits massively—to a world where it is bipolar?

[译文] 微软如何应对与 90 年代（当时世界上只有一个国家重要，那就是美国，我们的公司到处销售产品，因此微软受益匪浅）的不同，过渡到一个两极世界?

[原文] Where Microsoft can't just necessarily have the right to win all of Europe or India or Singapore.

[译文] 在这个世界中，微软不一定拥有赢得整个欧洲、印度或新加坡的权利。

[原文] There are actually sovereign AI efforts.

[译文] 实际上存在主权人工智能（sovereign AI）的努力。

[原文] What is your thought process here and how do you think about this?

[译文] 您在这里的思考过程是什么，您如何看待这个问题?

[原文] It's a super critical piece.

[译文] [SN]: 这是一个超级关键的部分。

[原文] I think that the key, key priority for the US tech sector and the US government is to ensure that we not only do leading innovative work, but that we also collectively build trust around the world on our tech stack.

[译文] 我认为，美国科技行业和美国政府的关键、关键优先事项是确保我们不仅要进行领先的创新工作，还要共同在全球范围内建立对我们技术堆栈（tech stack）的信任。

[原文] Because I always say the United States is just an unbelievable place.

[译文] 因为我总是说美国简直是一个不可思议的地方。

[原文] It's just unique in history.

[译文] 它在历史上是独一无二的。

[原文] It's 4% of the world's population, 25% of the GDP, and 50% of the market cap.

[译文] 它占世界人口的 4%，国内生产总值（GDP）的 25%，以及市值的 50%。

[原文] I think you should think about those ratios and reflect on it.

[译文] 我认为你应该思考一下这些比率并进行反思。

[原文] That 50% happens because quite frankly the trust the world has in the United States,

[译文] 这 50% 的市值坦率地说是因为世界对美国的信任，

[原文] whether it's its capital markets or whether it's its technology and its stewardship of what matters at any given time in terms of leading sector.

[译文] 无论是对其资本市场（capital markets）的信任，还是对其技术以及在任何特定时间对主导行业（leading sector）管理（stewardship）的信任。

[原文] If that is broken, then that's not a good day for the United States.

[译文] 如果这种信任被打破，那么对美国来说就不是一个好日子。

[原文] We start with that, which I think President Trump gets, the White House, David Sacks, everyone really, I think, gets it.

[译文] 我们从这一点开始，我认为特朗普总统、白宫、大卫·萨克斯（David Sacks），我认为所有人都真正理解这一点。

[原文] So therefore I applaud anything that the United States government and the tech sector jointly does to, for example, put our own capital at risk, collectively as an industry, in every part of the world.

[译文] 因此，我赞赏美国政府和科技行业共同做出的任何努力，例如，作为一个行业，在世界各地共同承担我们自己的资本风险。

[原文] I would like the USG to take credit for foreign direct investment by American companies all over the world.

[译文] 我希望美国政府（USG）认可美国公司在世界各地的外国直接投资（foreign direct investment）的功劳。

[原文] It's the least talked about, but the best marketing that the United States should be doing is that it's not just about all the foreign direct investment coming into the United States, but the most leading sector, which is these AI factories, are all being created all over the world.

[译文] 这是最少被谈论的，但却是美国应该做的最好的营销：不仅是所有外国直接投资流入美国，而且是最领先的行业，即这些人工智能工厂，正在世界各地被创造出来。

[原文] By whom? By America and American companies.

[译文] 由谁创造？由美国和美国公司。

[原文] And so you start there, and then you even build other agreements around it, which are around

[译文] 所以你从那里开始，然后你甚至围绕它建立其他协议，这些协议围绕着

[原文] their continuity, their legitimate sovereignty concerns, around whether it's data residency, for them to have real agency and guarantees on privacy, and so on.

[译文] 它们的连续性，它们对合法主权（sovereignty）的担忧，无论是数据驻留（data residency），还是它们拥有真正的自主权（agency）和隐私保证等等。

[原文] In fact, our European commitments are worth reading.

[译文] 事实上，我们对欧洲的承诺值得一读。

[原文] We made a series of commitments to Europe on how we will govern our hyperscale investment there such that the European Union and the European countries have sovereignty.

[译文] 我们对欧洲做出了一系列承诺，关于我们将如何管理我们在那里的超大规模投资，以确保欧盟和欧洲国家拥有主权。

[原文] We're also building sovereign clouds in France and in Germany.

[译文] 我们还在法国和德国建设主权云（sovereign clouds）。

[原文] We have something called Sovereign Services on Azure, which literally gives people key management

[译文] 我们在 Azure 上有一个名为“主权服务”（Sovereign Services）的产品，它确实为人们提供了密钥管理（key management）

[原文] services along with confidential computing, including confidential computing in GPUs, which we've done great innovative work with Nvidia.

[译文] 服务以及机密计算（confidential computing），包括图形处理器（GPUs）中的机密计算，我们在与英伟达的合作中完成了伟大的创新工作。

[原文] So I feel very, very good about being able to build, both technically and through policy, this trust in the American tech stack.

[译文] 因此，我对能够通过技术和政策，在美国技术堆栈中建立这种信任感到非常非常好。

End of Content

章节 26：主权 AI 与开放竞争：来自半导体的教训

📝 本节摘要：

访谈者将 AI 与半导体行业进行比较，提出如果少数模型（如 TSMC 在芯片领域）占据绝对优势，主权 AI 的努力是否会失败。Satya 认为，各国追求 AI 主权的主要驱动力是利用 AI 创造经济价值，而不是追求技术本身的领先性。各国将通过要求连续性、去中心化（即避免集中风险）和利用开源模型来实现主权。他承认全球化无法立即逆转，但各国（包括美国）正在吸取大流行等教训，将韧性（resilience）纳入其关键供应链政策，从而推动像芯片制造和 AI 基础设施建设在地理上的分散化。微软的目标是尊重这些政策需求，将主权视为核心的商业要求。

Start of Content

[原文] How do you see this shaking out as you have this network effect with continual learning and things on the model level?

[译文] [Q]: 鉴于模型层面存在持续学习和网络效应（network effect），您认为这会如何演变?

[原文] Maybe you have equivalent things at the hyperscaler level as well.

[译文] 也许在超大规模服务商层面，您也有类似的东西。

[原文] Do you expect that the countries will say, "Look, it's clear one model or a couple models are the best, and so we're going to use them, but we're going to have some laws around the weights having to be hosted in our country"?

[译文] 您是否期望各国会说：“看，很明显一两个模型是最好的，所以我们将使用它们，但我们将制定一些法律，要求模型的权重必须托管在我们的国家”?

[原文] Or do you expect that there will be this push so that it has to be a model trained in our country?

[译文] 或者您是否期望会有这种推动力，使得模型必须在本国进行训练?

[原文] Maybe an analogy here is that semiconductors are very important to the economy, and people would like to have their sovereign semiconductors, but TSMC is just better.

[译文] 这里的类比可能是：半导体对经济非常重要，人们希望拥有自己的主权半导体，但台积电（TSMC）就是更好。

[原文] And semiconductors are so important to the economy that you will just go to

[译文] 而且半导体对经济太重要了，你只能去

[原文] Taiwan and buy the semiconductors.

[译文] 台湾购买半导体。

[原文] You have to.

[译文] 你不得不。

[原文] Will it be like that with AI?

[译文] 人工智能也会是这样吗?

[原文] Ultimately, what matters is the use of AI in their economy to create economic value.

[译文] [SN]: 最终，重要的是在他们的经济中使用人工智能来创造经济价值。

[原文] That's the diffusion theory, which ultimately, it's not the leading sector, but it's the ability to use the leading technology to create your own comparative advantage.

[译文] 那就是扩散理论（diffusion theory）：最终重要的不是领先行业本身，而是使用领先技术来创造自己的比较优势的能力。

[原文] So I think that will fundamentally be the core driver.

[译文] 所以我认为那将从根本上成为核心驱动力。

[原文] But that said, they will want continuity of that.

[译文] 话虽如此，他们会希望这种连续性。

[原文] So in some sense, that's one of the reasons why, I believe, there's always going to be a check to "Hey, can this one model have all

[译文] 所以从某种意义上说，我相信，这也是为什么总是会有一个制衡因素，来防止“嘿，这个单一模型是否能拥有所有

[原文] the runaway deployment?"

[译文] 失控的部署”?

[原文] That's why open source is always going to be there.

[译文] 这就是为什么开源（open source）将永远存在。

[原文] There will be, by definition, multiple models.

[译文] 根据定义，将会有多个模型。

[原文] That'll be one way.

[译文] 那将是一种方式。

[原文] That's one way for people to sort of demand continuity and not have concentration risk, that’s another way to say it.

[译文] 这是人们要求连续性、避免集中风险（concentration risk）的一种方式，这是另一种说法。

[原文] And so you say, "Hey, I want multiple models, and then I want an open source."

[译文] 所以你会说：“嘿，我想要多个模型，然后我想要一个开源模型”。

[原文] I feel that as long as that's there, every country will feel like, "Okay, I don't have to worry about deploying the best model and broadly diffusing because I can always take what is my data and my liquidity and move it

[译文] 我觉得只要有这种情况存在，每个国家都会觉得：“好吧，我不必担心部署最好的模型然后广泛扩散，因为我总能拿走我的数据和我的流动性，然后移动它

[原文] to another model, whether it's open source or from another country or what have you."

[译文] 到另一个模型，无论是开源的，还是来自另一个国家的，等等”。

[原文] Concentration risk and sovereignty, which is really agency, those are the two things that will drive the market structure.

[译文] 集中风险和主权（sovereignty）（实际上是自主权，agency），是驱动市场结构的两件事。

[原文] The thing about this is that this doesn't exist for semiconductors.

[译文] [Q]: 关于这一点是，这在半导体领域并不存在。

[原文] All refrigerators, cars have chips made in Taiwan.

[译文] 所有的冰箱、汽车都有台湾制造的芯片。

[原文] It didn't exist until now.

[译文] 这种情况直到现在才出现。

[原文] Even then, if Taiwan is cut off, there are no more cars or no more refrigerators.

[译文] 即使如此，如果台湾被切断，就不会再有汽车，也不会再有冰箱。

[原文] TSMC Arizona is not replacing any real fraction of the production.

[译文] 亚利桑那州的台积电（TSMC Arizona）并没有取代任何真正的生产份额。

[原文] The sovereignty is a bit of a scam, if you will.

[译文] 如果你愿意，可以说主权（sovereignty）有点像个骗局。

[原文] It's worthwhile having it, it's important to have it, but it's not real sovereignty.

[译文] 拥有它是有价值的，拥有它很重要，但它不是真正的主权。

[原文] We're a global economy.

[译文] 我们是一个全球经济体。

[原文] I think it’s kind of like saying, "Hey, at this point, we've not learned anything about what resilience means and what one needs to do."

[译文] [SN]: 我认为这有点像在说：“嘿，在这一点上，我们还没有学到任何关于韧性（resilience）意味着什么以及一个人需要做什么的教训”。

[原文] Any nation state, including the United States, at this point will do what it takes to be more self-sufficient on some of these critical supply chains.

[译文] 在这一点上，任何民族国家，包括美国，都会采取必要措施，在其中一些关键供应链上变得更加自给自足（self-sufficient）。

[原文] So I, as a multinational company, have to think about that as a first-class requirement.

[译文] 因此，我作为一家跨国公司，必须将此视为一流的要求（first-class requirement）来考虑。

[原文] If I don't, then I'm not respecting what is in the policy interests of that country long-term.

[译文] 如果我不这样做，那么我就是不尊重该国家长期的政策利益。

[原文] I'm not saying they won't make practical decisions in the short term.

[译文] 我不是说他们在短期内不会做出务实的决定。

[原文] Absolutely, globalization can't just be rewound.

[译文] 绝对地，全球化不能只是被倒带（rewound）。

[原文] All these capital investments cannot be made in a way, at the pace at which…

[译文] 所有这些资本投资不能以某种方式，以某种速度进行……

[原文] But at the same time, think about it, if somebody showed up in Washington and said, "Hey, we're not going to build any semiconductor plants," they're going to be kicked out of the United States.

[译文] 但与此同时，想想看，如果有人出现在华盛顿，说：“嘿，我们不会建造任何半导体工厂”，他们就会被赶出美国。

[原文] The same thing is going to be true in every other country, too.

[译文] 同样的事情在其他每个国家也会发生。

[原文] So therefore we have to, as companies,

[译文] 因此，作为公司，我们必须

[原文] respect what the lessons learned are, whether it's that the pandemic woke us up or whatever.

[译文] 尊重我们学到的教训，无论是大流行（pandemic）唤醒了我们还是其他什么。

[原文] But nevertheless people are saying, "Look, globalization was fantastic.

[译文] 但尽管如此，人们仍在说：“看，全球化非常棒。

[原文] It helped supply chains be globalized and be super efficient.

[译文] 它帮助供应链全球化并实现超高效率。

[原文] But there's such a thing called resilience, and we want resilience."

[译文] 但存在一种叫做韧性（resilience）的东西，我们想要韧性”。

[原文] So therefore that feature will get built.

[译文] 因此，该功能将得到构建。

[原文] At what pace, I think, is the point you are making.

[译文] 我认为，你强调的观点是速度。

[原文] You can't snap your fingers and say all the TSMC plants now are all in Arizona with all their capability.

[译文] 你不能弹指一挥，说台积电的所有工厂现在都带着它们的所有能力在亚利桑那州了。

[原文] They're not going to be.

[译文] 它们不会是。

[原文] But is there a plan?

[译文] 但有没有一个计划?

[原文] There will be a plan.

[译文] 会有一个计划。

[原文] And should we respect that?

[译文] 我们应该尊重它吗?

[原文] Absolutely.

[译文] 绝对应该。

[原文] So I feel that that’s the world.

[译文] 所以我觉得这就是世界。

[原文] I want to meet the world where it is and on what it wants to do going forward, as opposed to saying, "Hey, we have a point of view that doesn't respect your view."

[译文] 我希望迎接世界的现状以及它未来想做的事情，而不是说：“嘿，我们有一个不尊重你观点的观点”。

[原文] Just to make sure I understand, the idea here is that each country will want some kind of data residency, privacy, et cetera.

[译文] [Q]: 只是为了确保我理解，这里的想法是每个国家都会想要某种数据驻留（data residency）、隐私等。

[原文] And Microsoft is especially privileged here because you have relationships with these

[译文] 而微软在这里特别享有优势（privileged），因为您与这些

[原文] countries, you have expertise in setting up these kinds of sovereign data centers.

[译文] 国家有关系，您在建立这类主权数据中心（sovereign data centers）方面拥有专业知识（expertise）。

[原文] Therefore Microsoft is uniquely fit for a world with more sovereignty requirements.

[译文] 因此，微软非常适合一个具有更多主权要求（sovereignty requirements）的世界。

[原文] I don't want to sort of describe it as somehow we're uniquely privileged.

[译文] [SN]: 我不想将其描述为我们在某种程度上独享优势。

[原文] I would just say I think of that as a business requirement that we have been doing all the hard work all these decades, and we plan to.

[译文] 我只想说，我认为这是一种业务要求，我们在过去几十年里一直为此付出艰辛努力，并且我们计划继续这样做。

[原文] So my answer to Dylan's previous question was that I take—whether it's in the United States, or when the White House and the USG says,

[译文] 所以我对我之前对迪伦（Dylan）问题的回答是，我会认真对待——无论是在美国，还是当白宫和美国政府（USG）说：

[原文] "We want you to allocate more of your wafer starts to fabs in the US"—we take that seriously.

[译文] “我们希望你将更多的晶圆启动量（wafer starts）分配给美国的晶圆厂（fabs）”——我们都会认真对待。

[原文] Or whether it is data centers and the EU boundary, we take that seriously.

[译文] 或者无论是数据中心和欧盟边界（EU boundary），我们都会认真对待。

[原文] So to me, respecting what are legitimate reasons why countries care about sovereignty, building for it as a software and a physical plant, is what we'll do.

[译文] 因此对我来说，尊重各国关注主权的合法原因，并在软件和物理工厂方面为此进行构建，就是我们将要做的。

[原文] As we go to the bipolar world—US, China—it's not just you versus Amazon, or you versus Anthropic, or you versus Google.

[译文] [Q]: 当我们走向两极世界——美国、中国——这不仅仅是您对抗亚马逊，或者您对抗 Anthropic，或者您对抗 Google。

[原文] There is a whole host of competition.

[译文] 存在着一整套竞争。

[原文] How does America rebuild the trust?

[译文] 美国如何重建信任?

[原文] What do you do to rebuild the trust?

[译文] 您做了什么来重建信任?

[原文] To say, "Actually, no, American companies will be the main provider for you."

[译文] 来传达：“事实上，不，美国公司将是您的主要提供商”?

[原文] And how do you think about competition with up and coming Chinese companies, whether it be ByteDance and Alibaba or Deepseek and Moonshot?

[译文] 您如何看待与新兴的中国公司（无论是字节跳动（ByteDance）和阿里巴巴（Alibaba），还是 Deepseek 和 Moonshot）的竞争?

[原文] To add to that question, one concern is how we're talking about how AI is becoming this industrial capex race where you're rapidly having to build quickly across all loads of supply chain.

[译文] 补充这个问题，一个担忧是，我们正在谈论人工智能如何成为一场工业资本支出（capex）竞赛，您必须在所有供应链负荷上快速建设。

[原文] When you hear that, at least up until now, you just think about China.

[译文] 当您听到这些时，至少到目前为止，您只会想到中国。

[原文] This is their comparative advantage.

[译文] 这是他们的比较优势。

[原文] And especially if we're not

[译文] 特别是如果我们没有

[原文] going to moonshot to ASI next year, but it's going to be decades of buildouts and infrastructure, how do you deal with Chinese competition?

[译文] 在明年实现超级人工智能（ASI）的登月计划，而是需要数十年的建设和基础设施，您如何应对中国的竞争?

[原文] Are they privileged in that world?

[译文] 他们在那个世界是否享有优势?

[原文] It’s a great question.

[译文] [SN]: 这是一个很好的问题。

[原文] In fact, you just made the point of why trust in American tech is probably the most important feature.

[译文] 事实上，你刚才指出了为什么对美国科技的信任可能是最重要的特性。

[原文] It's not even the model capability, maybe.

[译文] 也许甚至不是模型能力。

[原文] It is, "can I trust you, the company, can I trust you, your country, and its institutions to be a long-term supplier?"

[译文] 而是：“我能信任你这家公司吗？我能信任你所在的国家及其机构成为一个长期的供应商吗”?

[原文] That may be the thing that wins the world.

[译文] 这可能就是赢得世界的关键。

[原文] That's a good note to end on.

[译文] [Q]: 这是一个很好的收尾。

[原文] Satya, thank you for doing this.

[译文] 萨提亚，感谢您接受采访。

[原文] Thank you so much.

[译文] [SN]: 非常感谢。

[原文] Thank you.

[译文] 谢谢。

[原文] Thank you.

[译文] [Q]: 谢谢。

[原文] It's awesome. You two guys are quite the team.

[译文] [SN]: 太棒了。你们两个人真是一个出色的团队。

End of Content