科技吞噬市场：私有化周期的延长
当我们开始建立增长基金时，一个非常简单的前提是：科技市场比以往任何时候都要大，而公司保持私有化（Private）的时间也比以往任何时候都要长。目前，全球市值最高的公司中，有六家是美国的科技公司，科技已经吞噬了整个市场。这种趋势对我们来说是一把双刃剑：它让我们有机会在公司处于私有市场时进行更多投资，但我们也必须对投入资本分红率（DPI: Distributed to Paid-In Capital，衡量基金向投资者返还现金的能力）保持警惕。
而自我们启动增长基金以来，最大的变量就是 AI。它正在以我们从未见过的速度扩张市场，AI 公司的增长速度和投资规模都打破了历史记录。在接下来的十年里，这将成为我们进行新投资的巨大顺风车。
Original English Source
It was like a very simple premise when we started. It was like tech markets are bigger than ever. Companies are staying private longer than ever. Uh and as a result of that uh you know the opportunity set for us is huge. Uh I was looking at it last night and I think I mean it kind of oscillates a little bit but I think six of the most valuable I think the six most valuable companies are US-based tech companies. It's definitely five and then sometimes you know it bounces around on number six. Um, and then, you know, it bounced around a little bit, but seven or eight of the top 10 are US-based technology companies. So, um, you know, technology has kind of swallowed the whole market. Um, you know, and I think increasingly we'll take market cap, uh, over time. Um, we've got some slides showing this whole trend and, you know, I guess databricks was an appropriate way to kick off. Um, you know, talking about the trend of companies staying private longer than ever. That's obviously a double-edged sword, you know, for us. um it gives us an opportunity to invest in companies more while they're in the private markets, but we also are very mindful uh about you know generating returns and and DPI. And then you know the big thing that's changed from when we started the growth fund is just is AI. Um you know we've got some slides on it. It's massively expanding the market. The AI companies are getting bigger uh faster than anything we've ever seen. The investment amounts are are bigger than anything we've ever seen. Uh and so it's just it it it looks to be you know a huge tailwind for us over the next 10 or so years um you know as we as we look to make new investments.资本开支竞赛：巨头背负的基建重担
AI 周期与以往任何周期都不同，其基础建设的规模远超想象。如果按照最近一个季度的趋势推算，大型科技公司的年度资本支出（Capex: Capital Expenditure，用于购买、升级和维护物理资产的资金）已达到约 4000 亿美元，其中大部分流向了 AI 基础设施（AI Infrastructure）和数据中心。
这意味着市场所需的训练和推理基础设施将由这些巨头率先建成。最有利的一点是，Google、Facebook、Amazon 和 Microsoft 等公司正在承担这一建设负担。这些公司可能是人类历史上最优秀的商业机器，它们能够承受潜在的产能过剩。这种大规模的基建投入，为所有在其之上构建应用的初创公司创造了极佳的条件。
Original English Source
The groundwork is being laid um in a way that's very different than previous cycles. Um, and the the groundwork that's being laid is bigger than anything we've ever seen before. So, I I'm all over the team. I'm like, this is too conservative. These numbers are going to end up way bigger. Uh, because I think just the big tech companies in their latest quarter, if you run rate their capex from the latest quarter, um, I think it's like $400 billion of annual capex and most of that is going into AI, you know, AI infrastructure and data centers. And so um you know what that means is the infrastructure is going to get built for uh all of the training and inference needs that the market is going to need. And and this is great for all the companies that are building on top of this. Uh the best part about this is it's mostly the large tech companies that are bearing the burden of the buildout. Um and so you you've probably all seen the charts of like capex spend as a percentage of their overall sales. It turns out they're the best companies, you know, probably ever created. Um, you know, companies like Google, Facebook, um, you know, Amazon and Microsoft, and they can bear, um, you know, potential capacity overbuild and and things like that. And so if you just put it in a conservative view, which again, I think the number is going to end up way bigger than this. So, um, so the buildout's massive and this bodess very very well uh for uh for our for our portfolio companies that are building on top of it.超越摩尔定律：AI 成本的断崖式下跌
在基础设施扩张的同时，AI 的投入成本和质量正在以超越摩尔定律（Moore's Law）的速度优化。在过去两年中，访问这些模型的成本下降了 99% 以上，实现了超过 100 倍的降幅。与此同时，模型的前沿能力（Frontier Capabilities: 模型在推理、编码等复杂任务上的最先进表现）大约每 7 个月就会翻倍。
这种“成本骤降、质量飙升”的态势，预示着 AI 最终会像电力或 Wi-Fi 一样普及。当你去别人家使用电灯时，你不会想“嘿，让我分摊几分钱电费”。在未来，AI 也会演变成这种无处不在的基础设施。
Original English Source
At the same time this is happening, the input cost and input quality is getting remarkably better like faster than Moore's law. So on the left hand side, you don't need to look at the details of this. Just trust me when I tell you the cost of the inputs um, you know, of accessing these models has declined 99% or a little more than 99% over the last two years. So you know sort of 100x declines greater than Moore's law decrease. At the same time the models have been improving in sort of frontier capabilities by a double factor every 7 months. So massive decline in the input cost um at the same time that the quality is going way up and this bodess really well for building new stuff and new capabilities uh on top of AI. Uh, you know, I think our house view now is that AI is going to end up like, you know, electricity or Wi-Fi. Like if you're if you're if you're getting if you're accessing, you know, electricity at somebody's house, you're not like, "Hey, let me chip in, you know, a few pennies for, you know, sitting in, you know, a room with light uh in your house." And I think it'll end up being the same thing in the fullness of time with with uh with uh with AI.价值分配法则：90% 的红利归于终端用户
AI 的市场机会远大于传统的软件市场。在移动互联网和云计算周期中，创造了约 10 万亿美元的新市值。而 AI 对经济的影响将更为深远：美国软件支出仅占 GDP 的 1%，而白领薪酬占 GDP 的 20%。AI 将通过增强、增效或替代，深入影响这 20% 的领域。
关于价值捕获，我的经验法则是：90% 的价值将流向终端客户，而提供服务的公司捕获剩下的 10%。即便只是这 10%，也将创造巨大的市值。以 iPhone 为例，尽管用户支付 1000 美元，但其创造的消费者剩余（Consumer Surplus: 消费者愿意支付的价格与实际支付价格之间的差额）远超此数。Google 的搜索和 Gmail 也是如此，虽然每年仅从每个用户身上变现约 200 美元，但提供的价值巨大。AI 将创造海量的新剩余价值，并由终端用户和捕获机会的公司共同分享。
Original English Source
The market opportunity for AI is so much greater than um than the software market and I think that's really exciting. Um, if you look at the previous cycle that we went through of mobile phones plus cloud computing, the big story behind that was basically creating 10 trillion or so of new market value uh across software companies, internet companies, mega cap tech companies and I think AI is going to be much larger because I think the impact on the economy is going to be much larger. Um and so you know if you look at the simple math that we have on the page uh US software spend is like 1% of GDP. US white collar payroll is like 20% of GDP. And so um you know there's a lot of areas where I think we'll see you know augmentation or or potential you know cost savings or efficiencies or or replacements um you know using technology. There's always a question when these things happen of how much uh the new companies are able to capture versus the end customers. My rule of thumb is like 90% of the value goes to the end customers. Um and you know 10% of the value goes to the companies serving them and and it turns out that that's just a massive amount of market cap, you know, if you're the 10% that you're capturing. The examples I always give are like, you know, what what does your iPhone cost? Like, I don't know, the latest like give or take a thousand bucks. Like, if you know, gun to your head, what would you pay for an iPhone? Like, you know, if you're if you're on the higher end income spectrum, like probably far greater than $1,000. Um, and you know, the difference between that and the $1,000 you pay is is the surplus. And it turns out Apple's still a really great business. Um or you know if you take the Google properties like you know search and and Gmail and now I guess increasingly AI stuff they monetize you per year. First of all you get it for free which is massive surplus but they're only monetizing you per year probably like 200 bucks or something like that. Um and there's a tremendous amount more value delivered um you know I would argue than that per user. So I think the big story is going to be massive new surplus created. A ton of it gets captured by you know end customers, end users, whether it's businesses or consumers and a massive amount of new market cap goes to companies um that are uh that are capturing that opportunity.全球分发奇点：AI 的即时扩张力
AI 的需求端表现出惊人的速度：ChatGPT 达到 3650 亿次搜索仅用了 2 年，而 Google 用了 11 年。这种 5.5 倍的加速源于 AI 是构建在互联网和云计算之上的，具备即时全球分发的能力。
与早期互联网不同，AI 不需要交付新的硬件产品。目前全球约有 15 亿至 20 亿活跃用户在使用 AI 工具，超过一半的全球互联网人口已经尝试过 AI。这种分发速度降低了增长风险。此外，AI 的商业模式也在进化，通过价格歧视（Price Discrimination: 根据不同用户的支付意愿收取不同价格的策略）实现更有效的变现。例如，OpenAI 在印度推出每月 3-4 美元的订阅，而在美国则有每月 200-300 美元的高端产品。
Original English Source
The demand side is the bigger, more interesting thing. Uh, so I read a stat yesterday um that the time to get to 365 billion searches on chat GPT was 2 years. The time for Google to get to 365 billion searches was 11 years. So it's five and a half times longer. So the big story on the demand side for me this time around is AI is built on the back of the internet and cloud computing. And because of that it sort of allows for immediate global distribution. Like if you look at the way you know Google and Facebook started for example like it started you know much much they started much more you know like small build both had like a network effect dynamic which just takes longer um and you know we didn't have full sort of internet proliferation across five you know five and a half billion people in the world and you know smartphones in the hands of everybody able to access the internet and so what that means is because of the nature of this technology you don't have to deliver a new hardware product and because we have global you know internet buildout and because we have cloud computing the whole world can access this and so if you just take chat GPT again they got to the scale that they're at five and a half times faster than Google which is staggering um but you know there's probably I don't know a billion you know I think the latest they said there's more than a billion monthly active users there's probably another billion or so people who have tried it so if you add up all the different plat platforms and a bunch of people have probably tried, you know, Google products and Facebook products. It's probably well over half of the global internet population has used AI tools already. Um, and you know, we know that there's probably in some shape or form somewhere between one and a half and two billion user active users of these products. So just the speed at which they got to distribution is unlike anything we've ever seen before. And so that that is heartening to me that the supply buildout will be utilized um maybe in a more predictable way uh than you know broadband uh in the early internet buildout days uh just because it you know it's it's it's built on the back of the previous infrastructure stuff.能源瓶颈与护城河：从算力到核能
随着算力瓶颈的突破，能源正成为下一个核心限制因素。我们对核能（Nuclear Power）持乐观态度，并已在该领域进行投资。大型科技公司正在核电站附近建设数据中心，甚至重启三哩岛核电站。除了能源，冷却技术（Cooling）也将迎来创新浪潮，以防止芯片和环境过热。
在评估 AI 应用的质量时，我们最看重总留存率（Gross Retention Rate）和获客难易度。在医疗工作流、客户支持和高端财务分析等领域，AI 的集成度越高、企业特定规则越深，其粘性（Stickiness）就越强。虽然目前许多 AI 应用的毛利率受到模型成本的挤压，但我们相信随着模型层竞争的加剧，输入成本将持续下降，从而提升应用层的盈利能力。
Original English Source
Um, I mean, yeah, like as of right now through our current means of energy production, yeah, I mean, energy is a bottleneck and so, you know, we've made investments on the nuclear side. I'm quite optimistic that you know we now have um you know like an embrace I would say of um you know nuclear power. I think three mile island's going to get powered back up. The big tech companies are building data centers near you know nuclear power plants. you know, we have figured out there there's a lot of natural gas in you know, places like uh West Texas that um you know, can be you can you can you can build training large training clusters like very near to them um and and pretty efficiently uh power those data centers. Um but yeah, we're going to find we're going to need, you know, different uh different sources, I think, is the short answer. And, you know, we're probably most optimistic about nuclear. ... The big component that I I think most folks have not yet uh realized or zone in on is is the cooling piece. Uh and so you'll see a whole wave of innovation around that part as well once you figure out how to generate all this energy, how to actually cool all this stuff down without boiling our oceans and and uh making the world melt down. ... If you had to pick two topline stats to look at uh to assess the business model Um it would be um gross retention rate. ... And then ease of customer acquisition. And so the way you see that is like organic customer demand, you know, high value of dollars that they're willing to pay relative to how much it costs to acquire them either via marketing or sales.📌 文中提及的人物和组织
人物: David George, Elon Musk, Sam Altman
公司/组织: OpenAI, Google, Meta, Microsoft, Amazon, xAI, Anthropic, Databricks, Salesforce