我正在买入的人工智能股票
目录
原文标题（英文）：These are the AI stocks I’m buying today (most haven’t heard of these): 原文链接：X 原文 作者：Con（@_Con） 发布时间：2026-07-28（Asia/Shanghai）
摘要#
- 核心问题：作者试图建立一套人工智能股票投资地图，不只买市场最熟悉的芯片股，而是把受益者分成“人工智能赋能者、人工智能受益者、人工智能分发者”三类，并说明自己如何筛选和配置这些公司。
- 关键论点：① 赋能者提供芯片、存储、光通信、算力、数据中心和电力等底层基础设施；② 受益者利用人工智能提升既有产品、降低成本或增强定价权，但必须拥有专有数据、高转换成本或网络效应，避免被通用模型商品化；③ 分发者掌握设备、应用或平台入口，能够把模型能力快速触达海量用户，不一定需要拥有最强的模型；④ 作者分别用领导力、收入质量、护城河、非对称性等指标评估赋能者，用产品增强、专有数据、成本与定价、转换成本等指标评估受益者，用分发控制、参与度、生态壁垒和规模触达评估分发者；⑤ 作者计划三类等权配置，以分散单一公司或单一环节的风险。
- 关键标的：赋能者包括 $ASYS、$NVDA、$MU、$AAOI、$NBIS、$IREN、$OPTX、$BRUN、$SIVE、$SHMD，并补充 $SMR、$OKLO、$CEG；受益者包括 $NOW、$TEAM、$ZETA、$HUBS、$CRM、$SOFI、$DUOL、$RDDT、$SHOP、$DDOG；分发者包括 $META、$GOOGL、$AAPL、$AMZN、$MSFT、$SPOT、$SNAP、$DUOL。
- 风险与不确定性：全文主要是作者个人投资框架与持仓清单，没有提供估值、财务预测、仓位规模或可核验的买入价；部分小盘标的流动性、执行能力和业务映射仍需单独尽调。作者也承认有些股票当下未必是最佳买点，之后才会发布具体进出场方案。等权配置只能降低个股风险，无法消除人工智能主题估值回落、资本开支放缓或技术路线变化带来的系统性风险。
- Takeaway：最有价值的不是照抄清单，而是用“基础设施—应用受益—用户分发”三层框架检查人工智能价值最终在哪里沉淀，并把商业护城河、估值与入场时点纳入后续独立研究。
中文译文#
这里有三种不同类型的人工智能股票是我正在买入的。
但人们只在买其中一种。因此，本文将介绍我正在买入的所有人工智能股票类型。
让我们开始：
我买入的股票可以分为三组：
1）人工智能赋能者；2）人工智能受益者；3）人工智能分发者。
下面解释它们分别是什么。
人工智能赋能者，是帮助建设人工智能基础设施的公司。可以想到半导体、能源、存储、新型云服务和光子技术，典型标的包括 $NBIS、$AAOI、$MU。
人工智能受益者，是从人工智能建设中获益的公司。它们通常能让产品变得更好、更便宜，例如软件即服务公司，以及 $ZETA、$TEAM、$NOW。
人工智能分发者则有所不同，我过去也很少谈论它们。这些公司会以最高效的方式把开源人工智能模型分发给用户，例如 $AAPL、$GOOGL、$META。
这三类公司合在一起，都将从人工智能建设中受益。
而我正在买入这三个细分领域，以从中获利。
在具体讨论公司之前，先介绍我的框架，以确认所买的东西足够好：
我会使用若干指标来判断每家公司是否优秀。
对于人工智能赋能者，我使用下面这套框架（由 @mkfilko 制定）：
1）赋能属性：公司是否在建设人工智能发展的基础？ 2）领导力：创始人在这一细分领域是否经验丰富？ 3）收入质量：收入是经常性的，还是一次性的？ 4）收入增长：是否存在增长拐点的逻辑，还是只有稳定增长？ 5）护城河：这件事是否只有该公司能做，至少其他公司无法在不投入大量精力、时间和资金的情况下做到？ 6）非对称性：最坏情形与潜在回报相比如何？ 7）信念缺口：我今天能够证明的事实，与未来催化剂将证明的事情之间有多大距离？
如果这一领域的公司通过全部指标，在我看来就是可以买入的标的。
对于人工智能受益者，我使用一套简单框架，全部围绕一个问题：“人工智能会让这家公司更强，还是会取代这家公司存在的理由？”
1）人工智能改进产品，但不会变成产品本身：公司本来就应解决重要的商业问题，人工智能只是帮助它把问题解决得更好。 2）拥有专有数据：人工智能的质量取决于数据。如果 OpenAI 能访问同一数据集，护城河就较弱；如果不能，这就是强大优势。 3）人工智能降低成本或提高定价权，最好两者兼备：例如客服人员从 100 人降至 60 人，或因为人工智能创造了更多价值，月费从 200 美元提高到 280 美元。 4）高转换成本：如果客户因多年历史数据、定制工作流、员工培训等原因很难离开，人工智能就不会摧毁这家公司。 5）网络效应或规模优势：人工智能与规模相互增强——客户越多，数据越多，产品越聪明，又会吸引更多客户。这种循环很重要。 6）人工智能没有使其行业商品化：这是最容易被忽视的一点。要问自己，某人能否借助 GPT 在六个月内做出该产品 80% 的功能？如果能，这是坏事；如果不能，则是好事。
与赋能者相同，如果这一领域的公司通过全部指标，在我看来就是可以买入的标的。
最后是人工智能分发者的框架。
这个框架很简单：
它是否拥有广泛受众，可以有效地把人工智能交付给这些人？
换句话说：“如果人工智能的能力提升 100 倍，这家公司会不会成为人们默认使用人工智能的地方？”
这一标准下的重要因素是：
1）掌握分发渠道：模型可能快速变化，但分发渠道更难复制。如果已有数十亿用户在某个应用或设备上投入时间，就是好迹象。 2）人工智能提高参与度：人工智能越有用，用户返回越频繁，例如更好的搜索结果、人工智能修图、消息和推荐。更高参与度意味着更多广告、订阅与生态锁定。 3）拥有生态护城河：用户离开时，失去的不只是一个人工智能模型，还包括应用、联系人和照片。关键是人工智能成为留下来的另一个理由，而不是唯一理由。 4）人工智能提高平台价值：应寻找会因人工智能而受益的平台。 5）能够规模化分发人工智能：拥有数十亿用户，意味着每项改进都能立即到达用户。明天推出新功能时，能影响的人越多越好。 6）不依赖拥有最佳模型：这是最大的误解。分发者不必赢得模型竞赛，可以使用其他公司的模型；真正需要的是用户关系。
这就是每类股票的理想框架。
现在进入大家最期待的部分：具体股票代码。
我会先介绍持有的人工智能赋能者，然后是受益者，最后是分发者。
人工智能赋能者清单：
- $ASYS：制造半导体生产设备。芯片需求越大，其工具需求也越大。
- $NVDA：制造驱动人工智能的图形处理器。每个主要人工智能模型都依赖 NVIDIA 硬件。
- $MU：供应人工智能系统所需的存储器。模型越大，所需高带宽存储器就显著增加。
- $AAOI：提供连接人工智能图形处理器集群的光网络。更快的人工智能需要更快的数据传输。
- $NBIS：通过图形处理器云出租人工智能基础设施，帮助公司在不自建数据中心的情况下训练模型。
- $IREN：拥有大规模电力和数据中心基础设施，并正利用这些资产扩展人工智能算力。
- $OPTX：开发改善高速通信的先进光学技术。人工智能数据中心需要更快、更高效的连接。
- $BRUN：建设支持不断增长电力需求的关键基础设施。没有更多电力，人工智能就无法扩张。
- $SIVE：通过改善数据中心连接和基础设施，缓解人工智能最大的瓶颈之一。网络越快，人工智能越快。
- $SHMD：提供高性能电子设备所需的专用组件。随着人工智能硬件增长，这类零件需求也应增长。
此外，$SMR、$OKLO、$CEG 等能源公司在这里也很不错，这是我的个人看法。
这里有一些很少被听说的名字，但我认为它们会表现得很好。
人工智能受益者清单：
- $NOW：人工智能让其工作流软件更快、更有价值，客户更不愿离开。
- $TEAM：人工智能帮助团队在 Jira 和 Confluence 内更高效地工作，改进了本已不可或缺的产品。
- $ZETA：人工智能让营销平台更聪明、更有效；更好的结果提高平台价值。
- $HUBS：人工智能帮助企业自动化更多销售和营销工作，让 HubSpot 更容易证明其价值。
- $CRM：人工智能把 Salesforce 变成每位销售人员更强大的助手；生产率改善提高客户价值。
- $SOFI：人工智能可降低成本、改善客服并优化贷款决策，长期应有助于提高利润率。
- $DUOL：人工智能以更低成本提供个性化课程，改善学习体验和盈利能力。
- $RDDT：人工智能改善搜索、推荐和广告；Reddit 还拥有互联网最有价值的数据集之一。
- $SHOP：人工智能帮助商户更高效地建店、营销和服务客户；更好的工具吸引更多企业。
- $DDOG：人工智能会创造更多软件，也意味着更多系统需要监控。随着基础设施变复杂，Datadog 将受益。
在我看来，这些股票都被严重超卖了。
人工智能分发者清单：
- $META：已有数十亿人使用其应用，每项人工智能改进都能立即交付给用户。
- $GOOGL：人工智能让搜索、YouTube、Android 和 Workspace 更有用；Google 拥有全球最大的分发网络之一。
- $AAPL：人工智能让 iPhone 和 Apple 生态更有价值。当人工智能内置于设备时，用户不需要新的应用。
- $AMZN：人工智能改善购物、Alexa 和 AWS；Amazon 能以巨大规模向消费者和企业交付人工智能。
- $MSFT：人工智能直接集成到 Windows、Office、Teams 和 GitHub；微软已掌握数百万人每天使用的软件。
- $SPOT：人工智能改善音乐发现、推荐和个性化播放列表，更高参与度让用户持续回来。
- $SNAP：人工智能增强消息、镜头和内容创作，让用户有更多理由留在平台。
- $DUOL：人工智能正成为学习体验的核心。产品越好，Duolingo 能触达和留住的学习者越多。
这就是我投资的完整人工智能股票清单。
它基本覆盖所有感兴趣的领域，也符合我的全部标准。
此外，最坏情况下，一只股票表现不佳，但其他股票表现很好；最佳情况下，全部表现出色。
所以，我在这里等权投资三类股票。
总的来说，我在买入一些别人没有关注的名字，也有着大多数人没有采用的理由。
不过，其中一些现在可能不是最佳买入机会。
我很快会发布这些股票适当的进场和退出位置。
不要错过。
原文（英文）#
原文标题（英文）：These are the AI stocks I’m buying today (most haven’t heard of these):
There are 3 different types of AI stocks I’m buying here.
Yet, people are only buying 1. So, in this article, I’m going to go over all the types of AI stocks I’m buying.
Let’s dive in:
The stocks I’m buying can be classified into 3 groups.
- AI enablers, 2) AI beneficiaries, and 3) AI distributors.
But, let’s go over what these actually are.
AI enablers are companies that aid the AI buildout. Think semis, energy, memory, neoclouds, photonics, basically $NBIS, $AAOI, $MU.
AI beneficiaries are companies that benefit off of the AI buildout. They usually can make their product better and cheaper. Think SaaS companies, and such, or $ZETA, $TEAM, $NOW.
AI distributors are different, though. And I haven’t talked about them much. These are companies that will distribute open-source AI models to people in the most efficient manner. Think $AAPL, $GOOGL, $META.
These 3 combined, are the stocks that will benefit off of the AI buildout.
And, these are the 3 niches I’m buying in to profit from it.
But let’s go over my framework to make sure what I’m buying is good (before diving into actual companies):
I use a few metrics to see if each company is good.
For AI enablers, I use this framework (made by @mkfilko):
- Enabler: Does this company build the foundation of the AI buildout?
- Leadership: Is the founder experienced within this niche?
- Revenue Quality: Is it recurring revenue, or a onetime sum.
- Revenue Growth: Is there a thesis for an inflection point, or just steady growth.
- Moat: Is it something only this company can do (at least without a lot of effort + time + money)?
- Asymmetry: What’s the worst-case scenario vs the possible reward?
- Conviction Gap: How big is the space between what I can prove today and what the next catalysts will prove.
If a company within this niche passes all of these metrics, that is a buy in my book.
For AI beneficiaries, I use a simple framework, all based on one question: “Does AI make this company stronger, or does AI replace the reason this company exists?”
- AI improves the product; it doesn’t become the product: The company should already solve an important business problem. AI simply helps them solve it better.
- They own proprietary data: AI is only as good as the data. If OpenAI can access this dataset then it has less of a moat. If OpenAI can’t, that’s powerful.
- AI lowers costs OR increases pricing power (ideally both): Instead of needing 100 support reps, now they only need 60, or instead of charging $200/month, they now charge $280/month because AI creates significantly more value.
- High switching costs: If it’s painful for a customer to leave this company’s service (due to years of historical data, custom workflows, employee training, etc.), AI don’t destroy them.
- Network effects or scale advantages: AI compounds with scale: The more customers -> the more data -> the smarter the product -> the more customers. It is a cyclical cycle. And it is important.
- AI isn’t commoditizing their industry: This is the most overlooked one. Ask yourself: could someone using GPT build 80% of this product in six months? If yes… that’s bad. If no… That’s good.
Like AI enablers, if a company within this niche passes all of these metrics, that is a buy in my book.
And lastly, the framework for AI distributors:
This one is simple.
Does it have a wide audience of people it can effectively ship AI to?
Basically, “If AI gets 100x better, does this company become the default place people use it?”
The important things are under this criterion:
- They own the distribution: Models can change fast, but distribution is much harder to replicate. If billions of users already spending their time on an app (or device), that’s already a good sign.
- AI increases engagement: The more useful AI becomes, the more often users return. Some examples could be better search results, AI photo editing, AI messaging, AI recommendations, etc.. All because higher engagement means: more ads, more subscriptions, more ecosystem lock-in.
- They have an ecosystem moat: If users leave, they won’t just hurt themselves due to an AI model, but apps, contacts, photos. The key is AI becomes another reason to stay. Not the only reason.
- AI makes the platform more valuable: You’re looking for platforms in which AI will benefit.
- They can distribute AI at scale: Having billions of users means every AI improvement reaches users immediately. If they launch a new AI feature tomorrow, how many people are going to feel it (the more the better).
- They aren’t dependent on having the best model: This is the biggest misconception. They don’t need to win the model race (they’re just using others, what they need is user relationship).
That’s the ideal framework for each stock.
Now, let’s get into the part I know all of you want: some actual tickers…
I’m going to go over the AI enabler stocks I’m in, then the AI beneficiaries, followed by AI distributors last.
For AI enablers, here’s my list:
$ASYS: Builds the equipment used to manufacture semiconductors. If more chips are needed, more of its tools are needed too. $NVDA: Makes the GPUs powering AI. Every major AI model depends on NVIDIA’s hardware. $MU: Supplies the memory AI systems need. Bigger AI models require significantly more high-bandwidth memory. $AAOI: Provides the optical networking that connects AI GPU clusters. Faster AI requires faster data movement. $NBIS: Rents AI infrastructure through its GPU cloud. It helps companies train AI without building their own data centers. $IREN: Owns massive power and data center infrastructure. It’s using those assets to expand into AI compute. $OPTX: Builds advanced optical technologies that improve high-speed communication. AI data centers need faster, more efficient connections. $BRUN: Builds critical infrastructure that supports growing power demand. AI can’t scale without more electricity. $SIVE: Helps solve one of AI’s biggest bottlenecks by improving data center connectivity and infrastructure. Faster networks mean faster AI. $SHMD: Provides specialized components used in high-performance electronics. As AI hardware grows, demand for these parts should grow too.
*Some energy names such as $SMR, $OKLO, $CEG (and the such are great here too imo).
You have some underheard of names here.
But I think those will perform very well.
For AI beneficiaries, here’s my list:
$NOW: AI makes its workflow software faster and more valuable. Customers become even less likely to leave. $TEAM: AI helps teams work more efficiently inside Jira and Confluence. It improves an already essential product. $ZETA: AI makes its marketing platform smarter and more effective. Better results make its platform more valuable. $HUBS: AI helps businesses automate more of their sales and marketing. That makes HubSpot even easier to justify. $CRM: AI turns Salesforce into a more powerful assistant for every salesperson. Better productivity leads to higher customer value. $SOFI: AI can lower costs, improve customer support, and make lending decisions smarter. That should help margins over time. $DUOL: AI creates more personalized lessons at a lower cost. It makes learning better while improving profitability. $RDDT: AI improves search, recommendations, and advertising. Reddit also owns one of the most valuable datasets on the internet. $SHOP: AI helps merchants build stores, market products, and serve customers more efficiently. Better tools attract more businesses. $DDOG: AI creates more software, which means more systems to monitor. Datadog benefits as infrastructure becomes more complex.
These are drastically oversold imo.
And lastly, for AI distributors, here’s my list:
$META: Billions of people already use its apps. Every AI improvement can be shipped to users instantly. $GOOGL: AI makes Search, YouTube, Android, and Workspace more useful. Google has one of the largest distribution networks in the world. $AAPL: AI makes the iPhone and Apple’s ecosystem even more valuable. People don’t need a new app when AI is built into their devices. $AMZN: AI improves shopping, Alexa, and AWS. Amazon can deliver AI to both consumers and businesses at enormous scale. $MSFT: AI is built directly into Windows, Office, Teams, and GitHub. It already owns the software millions use every day. $SPOT: AI improves music discovery, recommendations, and personalized playlists. Better engagement keeps users coming back. $SNAP: AI enhances messaging, lenses, and content creation. It gives users more reasons to spend time on the platform. $DUOL: AI is becoming a core part of the learning experience. As the product improves, Duolingo can reach and retain even more learners.
This is the full list of all the AI stocks I’m invested in here.
It covers really all areas of interest, and they check all my boxes.
Plus, worst case scenario, one does bad, but the other do really great.
Best case scenario, everything does great.
So, I’m investing into all 3 of them equally here.
Overall, I’m buying into some names people aren’t.
And, with reasoning most don’t have.
But some might not be the best buying opportunity right now.
Soon, I’ll be posting the proper entries/exits of these stocks.
Don’t miss out.
这里还没有任何文章可以列出。