TIP855：Shopify (SHOP)：代理商务的最大赢家？与 Shawn O'Malley 和 Daniel Mahncke

所以，人们开始在聊天机器人里开始他们的购物研究，而不是在谷歌上。

So, people starting their shopping research inside the chatbot instead of, for example, Google.

但完整的愿景——你给代理一个预算，然后它主动帮你买东西——现在因为Meta的Muse看起来更接近了，但还没有完全实现，我们也不知道至少在不久的将来会不会实现。

But the full vision, where you give the agent basically a budget and then it proactively buys things for you, that seems closer now due to Meta's Muse, but it's not yet fully there, and we don't know whether that will be the case in at least the near future.

要让代理做得好，能够基本上代表你做决定，它需要大量的背景信息，大量关于你生活的信息。

For an agent to do this well and be able to make decisions basically on your behalf, it needs a ton of context, a ton of information about your life.

我们昨天正好和Muse聊到这个，看起来很多人愿意注册Muse，他们只是假设把数据交给Meta，而且他们对此感到放心。

And we were just talking about that with Muse yesterday, where it seems like a lot of people are willing to sign up for Muse, and they just assume that they're handing over their data to Meta, and they're comfortable with that.

但实际上，情况并非完全如此。

And actually, it turns out not exactly to be the case.

Meta正试图建立一种屏障，或者把Muse收集的数据分流出去。

Meta is trying to create a kind of a barrier, or siphon off the data that Muse is collecting.

但重点是，要让代理能很好地代表你行事，它真的需要知道你生活中的所有相关变量，它必须成为你的私人购物助手，你的私人助理。

But the point being, for an agent to be able to act well on your behalf, it really needs to know all of the relevant variables in your life, and it has to become your personal shopper, your personal assistant.

它不只是每周帮你买杂货的机器人。

It's not just some robot that buys your groceries every week.

所以，关于怎么思考这个问题，我很大程度上借鉴了一个人，在我看来，他是卖方最好的电商分析师之一，就是MoffettNathanson的Michael Morton，Ben Thompson六月份也在Stratechery上采访过他，我们在节目笔记里放了链接。

And so I take a lot of inspiration on how to think about this from, in my opinion, one of the best e-commerce analysts on the sell side, and that's Michael Morton at MoffettNathanson, who Ben Thompson also interviewed on Stratechery in June, and we've linked to that in the show notes.

但Morton对全自动食品储藏室以及这些东西自动送到你家的想法非常怀疑。

But Morton is really skeptical of the idea of having a fully automated pantry and these things automatically being delivered to you.

而他最喜欢的证据，说来也怪，是送奶工。

And his favorite piece of evidence is, of all things, the milkman.

即使在1960年代初，送奶服务无处不在、毫无摩擦，需求也几乎可以预测到极致，但美国牛奶消费中只有30%是通过配送完成的。

So even in the early 1960s, when milk delivery was ubiquitous, frictionless, and demand was probably as predictable as it could get, only 30% of US milk consumption came via delivery.

然后从那时起，这个比例就逐渐降到了零。

And then from there, it went to zero over time.

我有点好奇我们是怎么从AI代理聊到60年代的送奶工的。

I kind of asked myself how we went from AI agents to the milkman of the '60s.

但我认为亚马逊的Subscribe & Save也是一个很好的例子，因为根据他的计算，那可能只占亚马逊总GMV的1%到3%。

But I think a good example has also been Amazon Subscribe & Save, because by his math, that captures maybe 1 to 3% of Amazon's total GMV.

这大概就说明了问题：普通家庭每年购买250到300种不同的商品，除非一个代理能解决几乎所有这些问题，否则它实际上什么也没解决。

And that kind of gives you the point of the average household buys 250 to 300 different items a year, and unless an agent solves for pretty much all of them, it hasn't really solved anything.

所以消费者其实从来就不想要东西自动送上门。

So consumers have never actually wanted stuff just showing up.

至少这是他的观点。

That's his argument, at least.

你之前跟我说过，你试过很多亚马逊的“订阅并省”服务，结果就是东西越积越多，最后都取消了，因为大多数产品的需求没法按周或按月精确预测。

And you were telling me before the call about all the different Amazon Subscribe & Save's that you've done that just led to you accumulating more stuff than you ever wanted and then canceling eventually, because for most products, your demand can't be perfectly estimated on a weekly or monthly schedule.

我经常跟你说，我基本上每天吃的东西都差不多。

And I often tell you, I basically eat pretty much the same stuff every day.

但连我也做不到那样。

And even I couldn't do that.

就算是我，每周买同样的东西，最后通常也得扔掉一些，因为已经不新鲜了。

Even for me, I totally see that when I buy the same stuff every week, usually you have to throw something away because it's not good anymore.

而AI购物真正发挥作用的地方，是在那些需要深思熟虑的产品发现上。

And where AI shopping really does work is with these high-consideration product discoveries.

比如帮你找到，怎么说呢，一款最适合冲浪的新手表，或者三周后去旅行要用的新飞钓竿。

So maybe helping you find, I don't know, the best option for a new watch for surfing, or maybe a new fly fishing rod for a trip in three weeks.

而这正是Shopify商家真正在卖的东西。

And that's exactly the kind of stuff that Shopify merchants actually sell.

所以是那种具体的、需要深思熟虑的产品。

So it's specific, high-consideration products.

嗯，我知道我们在蒙大拿州飞钓时都玩得很开心。

Well, I know we both had so much fun fly fishing in Montana.

那很好。

That's good.

这是个很好的例子。

It's a great example to use.

也许我会给自己买一根飞钓竿。

Maybe I will get myself a fly fishing rod.

但不管我们有没有意识到，我觉得Shopify早就看到了这一点，而且他们至少已经为此准备了两年。

But whether we appreciated it or not, I think Shopify has seen this for some time, and they've been building for it for at least two years.

然后在过去大约九个月里，他们推出了很多我称之为AI购物基础管道的东西。

And then over about the last nine months, they rolled out a lot of what I'll call maybe the basic plumbing for AI shopping.

所以我觉得这里面有三点值得了解，我会尽量少用专业术语来讲。

And so I think there are three pieces to this that are worth knowing, and I'll try to keep the jargon to a minimum as we go through it.

第一点是“Catalog”，大写C。基本上就是一个巨大的、实时更新的列表，包含来自数百万Shopify商家的超过十亿件商品。

And so the first is Catalog with a capital C. And so it's basically one giant, always up-to-date list of more than a billion products from millions of Shopify merchants.

每个价格、每个尺码、库存情况，所有这些信息都以AI模型可以直接读取的格式呈现。

Every price, every size, what's in stock, all of that information in a format that AI models can directly read.

换句话说，这让每个Shopify店铺都变得非常容易被AI助手理解。

So in other words, it makes every Shopify store very easy for AI assistants to understand.

Shopify的CEO对此有一句非常精彩的话。

And so Shopify CEO has a really great line about this.

他说：“其他人都得从网上抓取产品数据，”基本上就是派机器人去读网页，而Shopify本身就是数据的源头。

"Everyone else," he says, "has to scrape the internet for product data," basically sending bots to read webpages, while Shopify is the source of that data.

这一点越来越重要，因为基于Catalog的AI搜索转化为购买的比例，是依赖抓取网页的两倍。

And so that matters more and more because AI searches powered by Catalog turn into purchases twice as often as ones that rely on scraped webpages.

而且Shopify还在添加购物者真正关心的细节，对吧？

And Shopify is also adding the kind of detail that shoppers really care about in shopping, right?

这条裙子参加婚礼够正式吗？

Is this dress formal enough for a wedding?

面料透气吗？

Does the fabric breathe?

这双运动鞋适合长跑吗？

Are these sneakers good for a long run?

AI助手需要这些背景信息，Shopify正在努力满足这一点，而这些细节光靠浏览网页是得不到的。

An AI assistant needs that context, and Shopify is working to accommodate that, and it can't get that kind of insight by skimming a webpage.

而这正是Agentic Commerce需要解决的核心问题。

And that is really the problem that Agentic Commerce needs to solve.

现在你在ChatGPT里购物，其实就是在浏览Shopify的商品目录。

And when you shop inside of ChatGPT today, what you're basically doing is you're browsing Shopify's catalog.

我觉得那些问题对你来说挺有切身体会的，毕竟你最近四五周就参加了三场婚礼。

I feel like those questions were personal to you, having visited- ... three weddings in the last, what is it, four or five weeks.

不过这也是你最近几周经常提到的点，我觉得说得很好。

But that's a point that you brought up in the last couple of weeks quite often, and I think it's a good one.

想想看，每次有人购物时，AI要浏览网页需要多少计算能力，对吧？

Just think about how much computing power it takes for an AI to comb through webpages every time someone shops, right?

而目录直接提供干净的产品数据，这正好说明了AI长期来看是如何让Shopify占优势的。

Catalog hands it clean product data instead, which is a great example of AI tilting things in Shopify's favor in the long term.

接下来的术语可能有点技术性，但第二部分叫做“通用商务协议”，简称UCP。

And the jargon is maybe going to get a bit technical here, but the second piece is called the Universal Commerce Protocol, or just UCP.

怎么理解呢，Daniel，我们不是技术人员，所以我会尽力解释，我相信评论区会有人说我们没说到点子上，但我就这么讲了。

And the way to think of this, and Daniel, we're not technical people, so I'm going to do my best to explain it, and I'm sure somebody in the comments will tell me that we didn't quite hit the nail, but here it goes.

我的理解是，它就像一种通用语言，让任何AI助手都能在任何商店购物。

The way to think of it is sort of like a common language, as I understand it, that lets any AI assistant shop at any store.

实际上，Shopify和谷歌合作，在1月份把它作为开放标准一起推出了。

And so actually, Shopify and Google worked on this and launched it together in January as an open standard.

基本上，它涵盖了整个购物流程，从找产品、加购物车、结账，甚至处理退货。

And basically, it covers the whole shopping trip, from finding a product, to putting it in the cart, to checking out, and even working on returns.

而且它兼容任何支付提供商，所以每家商店都能保留自己的结账、订阅、折扣等等，这样就能实现代理商务，而不会让ChatGPT或Claude抢走零售商和商家的客户关系。

And so it works with any payment provider, so each store keeps its own checkout, subscriptions, discounts, all of it, and basically allows Agentic Commerce to happen without ChatGPT or Claude stealing the customer relationship from retailers and merchants.

所以整个行业很快就跟进了，对吧？

And so the industry has gotten behind it quite fast, right?

亚马逊、Meta、微软、Salesforce和Stripe都加入了管理UCP这个协议的组织。

You had Amazon, Meta, Microsoft, Salesforce, and Stripe all join the group that oversees the UCP, this protocol.

然后像Target、Etsy和Wayfair这些零售商也选择了加入。

And then retailers like Target and Etsy and Wayfair have opted into it, too.

这可能是唯一一个亚马逊没有立刻反对的AI相关的东西，但我也觉得这不算什么大承诺，对吧？

That might be the only thing regarding AI that Amazon didn't immediately block, but I also assume it's less of a commitment, right?

我觉得这听起来没那么乐观，因为它是一个开放标准，对吧？

I don't think it's as bullish as it sounds because it's an open standard, right?

所以加入只是意味着同意AI代理购物时的技术规则，并不代表这些公司，比如，要为使用这个协议向Shopify或Google支付任何费用。

So signing on just means agreeing to the technical rules for how AI agents will shop, and it does not mean that those companies, for example, owe Shopify or Google any kind of toll for using that protocol.

不然的话，我敢肯定他们不会加入。

Otherwise, I'm very certain they wouldn't have joined.

当然，亚马逊希望自己的购物代理能流行起来，那就意味着这些代理需要在亚马逊之外也能工作。

And of course, Amazon wants its own shopping agents to catch on, which means those agents need to be able to work outside of Amazon.

所以，一个所有大型科技公司都认同的通用标准，不管是谁制定的，在这里是他们的主要竞争对手之一Shopify，都有助于实现这一点。

And so a common standard that all the big tech companies are aligned on, that helps with that, no matter who wrote it, in this case, one of their chief rivals, Shopify.

这听起来可能有点矛盾，因为正如你提到的，亚马逊刚刚把Muse从它的网站上屏蔽了。

And so that might sound contradictory because Amazon just blocked Muse from its site, as you're kind of alluding to there.

但亚马逊是想鱼和熊掌兼得。

But Amazon is trying to have its cake and eat it, too.

我觉得它希望自己的智能体能在其他任何地方购物，同时把其他公司的智能体挡在亚马逊之外。

I think it wants its own agents to shop everywhere else while keeping other companies' agents out of Amazon.

也许他们现在能做到，但我认为这不可持续。

And maybe they can do that for now, but I don't think it's sustainable.

但对Shopify来说，特别之处在于，现在任何开发者都可以用UCP构建，并接入Shopify的商品目录，无需批准就能接触到Shopify的所有商家。

But what is special for Shopify is that now any developer can build using the UCP and plug into Shopify's catalog, giving them access to all of Shopify's merchants with no approval needed.

所以，任何人构建的每一个新AI购物应用，基本上都会自动成为Shopify商家销售商品的一个新店面。

So every new AI shopping app anyone builds becomes basically another storefront for Shopify merchants to sell through by default.

这给Shopify带来了巨大的信誉，连亚马逊都决定不能忽视一个由Shopify和谷歌共同制定的标准。

And it gives Shopify a ton of credibility that even Amazon decided it couldn't afford to ignore a standard that Shopify and Google had both written.

亚马逊在这场讨论中真的很有意思。

Amazon is so interesting in this entire debate.

我觉得回到送牛奶的例子，日常购物的消费者可能不会通过Muse下单，因为我们看到过订阅和保存的运作方式。

I think going back to the milk delivery example, everyday buyers might not go through Muse because of what we've seen with the Subscribe and Save dynamic.

不过，亚马逊可能确实有办法阻止外部代理，而且不会造成太大影响，甚至长期来看也是这样。

Amazon might actually have a good shot, though, of blocking outside agents without a major impact, even in the long term.

至少我是这么认为的，因为你可以在亚马逊上继续像现在这样购买你的日常用品，然后通过代理去购买那些需要慎重考虑的高价值商品。

At least that's why I'm saying, because you can keep buying your monthly routine on Amazon just as you're doing today, and then shop for these high consideration assets or products through an agent.

即使那样，你最终可能还是会回到亚马逊，但更重要的是，亚马逊仍然可以为你将要购买的大部分商品展示广告。

And you might still end up on Amazon even then, but more importantly, Amazon can still show you ads for the majority of products that you are going to buy.

这样一来，它只是失去了那些高价值商品的销售机会，而这些商品，我不知道你怎么想，我通常不会在亚马逊上买那些特别的东西。

Then it's only losing them for these high consideration assets or products, which usually, I don't know about you, I don't often buy that special stuff on Amazon.

大多数情况下，这些购买发生在亚马逊生态系统之外。

Mostly, it's somewhere outside of the Amazon ecosystem.

对我来说，理解这个协议最重要的一点是，它让商家可以继续使用他们想要的任何支付处理服务。

For me, the most important thing to understand about this protocol is that it lets merchants keep using whatever payment processor they want.

顺便说一下背景，早期大家担心谷歌最终会进军Shopify的支付业务，所以Shopify联合制定了这个标准，就是为了防止这种情况发生。

And just for context, the early fear was that Google would eventually come for Shopify's payments business, and so Shopify co-wrote this standard precisely so that wouldn't happen.

而且就算他们没这么做，谷歌为什么要为了每笔交易那点支付处理的小钱操心呢？它有那么庞大、那么巨大的广告业务要保护。

And even if they hadn't done that, why should Google care about pennies on a transaction for payment processing when it has this massive, gigantic advertising business to protect?

这才是它的核心重点。

That is its core focus.

这也是我脑子里最大的疑问。

And that is also the biggest question, at least in my head.

谷歌、亚马逊这些公司的广告业务会变成什么样？

What will happen to the advertising businesses of Google, Amazon, and co., those type of companies?

我觉得Meta可能是唯一比较安全的，因为人们在找娱乐内容的时候会看到广告。

And I think Meta is probably the only one that's pretty safe because they play ads when people are looking for entertainment.

他们完全不参与购买过程。

They're not anywhere in the buying process.

不管你有没有用代理，或者直接购物，人们还是会去Instagram找娱乐。

And if you have agents, or you just shop without them, people will still go to Instagram for entertainment.

但我觉得亚马逊和谷歌其实面临着一个威胁，这个威胁大概一年前还没人能想象到，这些公司可能会失去他们业务的核心，而对他们两家来说，现在核心就是广告业务。

But Amazon and Google, I think they actually face a threat here that probably just a year ago, nobody would've imagined that it's possible that those companies might lose the core of their business, which for both of them by now is their advertising business.

我个人相信他们能想出解决办法，但这个威胁绝对不容小觑。

And I personally believe they will be able to figure it out, but it's certainly a threat that is not to be underestimated.

我也这么认为。

I think so, too.

有意思的是，Shopify声称自己是目前唯一一个让商家能直接在ChatGPT、Copilot和谷歌里销售的平台，而且所有这些都能通过Shopify在一个地方管理。

And so it's interesting that Shopify claims that it's currently the only platform that lets merchants sell directly within ChatGPT, Copilot, and Google, and then you can manage that all from one place, of course, through Shopify.

所以我觉得关键点在于，聊天机器人并没有把Shopify从结账流程中踢出去，我们之前提到过，商家解决方案业务是Shopify增长的最大引擎，而支付处理又是这个引擎中推动收入的最大部分。

So I think the key point is that the chatbots don't cut Shopify out of the checkout process, where we talked about merchant solutions business as the biggest engine of growth for Shopify, and then payment processing is the biggest part of that engine in terms of driving revenue.

所以支付处理这个角度对Shopify来说真的很重要，基本上，聊天机器人负责和购物者对话。

So the payment processing angle is really important for Shopify, and basically, the chatbot handles the conversation with the shopper.

但Shopify仍然在幕后处理一切。

But Shopify still runs everything behind the scenes.

它处理订单，通过Shopify Payments收款，然后帮忙安排发货。

It processes the order, it takes the payment through Shopify Payments, and then it helps get it shipped.

我们快速休息一下，听听今天赞助商的介绍。

Let's take a quick break and hear from today's sponsors.

不用我说你也知道，房租和房贷越来越贵了。

You don't need me to tell you that rent and mortgages are only getting more expensive.

所以Bilt推出了三款信用卡，专门为房租、房贷以及其他日常消费提供最丰厚的积分奖励。

That's why Bilt created three credit cards designed to offer the richest rewards on rent, mortgage, and everything else you spend on.

无论你是租房还是自有住房，每次支付住房相关费用都能获得高达1.25倍的积分，而且没有交易手续费，不管你在哪里住。

Whether you rent or own, you can earn up to 1.25x points on every housing payment without the transaction fee, no matter where you live.

积分可以兑换机票、酒店、Lyft打车服务、未来的住房付款，还有更多。

Points can be redeemed for flights, hotels, Lyft rides, a future housing payment, and much more.

把你生活中最不可避免的一笔开销，变成最划算的一笔。

Turn one of the most unavoidable payments in your life into the most rewarding.

找到适合你的卡，去joinbilt.com/tip申请吧。

Find the card that's right for you and apply at joinbilt.com/tip.

网址是J-O-I-N-B-I-L-T.com/tip。

That's J-O-I-N-B-I-L-T.com/tip.

用我们的链接，这样他们就知道是你推荐来的。

Use our link so they know we sent you.

适用条款。

Terms apply.

需经批准并符合资格要求。

Subject to approval and eligibility.

Bilt卡由Column发行，Column是FDIC成员，并依据万事达卡国际公司的许可发行。

Bilt cards are issued by Column in a member FDIC pursuant to license from Mastercard International Incorporated.

对在线交易感兴趣，但还没迈出第一步？

Curious about online trading but haven't taken the first step yet?

你不是一个人。

You're not alone.

而Plus500期货是一个很好的起点。

And Plus500 Futures is a great place to start.

期货市场瞬息万变，通过Plus500，你可以探索石油、黄金、标普500、比特币等热门资产。

The futures markets are moving fast, and with Plus500 you can explore popular assets like oil, gold, S&P 500, Bitcoin, and more.

从加密货币到大宗商品，总有什么在发生。

From crypto to commodities, there's always something happening.

这个平台超级好用，你可以直接在手机上随时交易。

The platform is super easy to use, so you can trade on the go right from your phone.

你只需要100美元就能开始，马上参与进来。

You can get started with just $100 and jump into the action.

看到感兴趣的东西了吗？

See something interesting?

账户开通后，只需点击几下就能交易。

Once your account is open, you can trade it in just a couple of clicks.

如果你还没准备好，可以用免费模拟账户练习一下。

And if you're not quite ready yet, you can practice with a free demo account.

没有风险，也没有压力。

No risk, no pressure.

凭借20年的经验，Plus500让交易比以往任何时候都更容易上手。

With 20 years of experience, Plus500 makes trading more accessible than ever.

去plus500.com看看吧。

Check it out at plus500.com.

期货交易涉及亏损风险，并非适合所有人。

Trading in futures involves risks of loss and is not suitable for everyone.

并非所有申请者都能通过审核。

Not all applicants will qualify.

Plus500，让交易更上一层楼。

Plus500, it's trading with a plus.

每周我们都会在节目中分析一些公司，并问同一个问题：管理层真的在高效运营这家公司吗？

Every week, we analyze companies on this show and ask the same question: Is management actually running this business efficiently?

对于目前大多数公司来说，诚实的答案是这样的。

Here's the honest answer for most companies right now.

那些领先的公司，正是那些把人工智能应用到日常事务中的公司。

The ones pulling ahead are the ones putting AI to work on the everyday stuff.

这正是NetSuite Next的用武之地。

That's exactly what NetSuite Next was built for.

NetSuite Next 会在你最需要的时候给你提供关键信息。

NetSuite Next surfaces insights exactly when you need them.

它能让智能代理处理日常任务和问题，你还可以用日常语言问复杂问题，就像跟一个懂行的同事聊天一样。

It puts agents to work on routine tasks and problems, and it lets you ask complex questions in plain language, like talking to a knowledgeable colleague.

但技术只是其中的一部分。

But technology is only part of the story.

AI 也在改变人们和团队协作的方式，这种变化已经开始了。

AI is also reshaping how people and teams work together, and that shift is already underway.

今天你就能拿到专家编写的商业指南。

Expert-created business guide in your hands today.

这本指南叫《为智能体时代做好准备：AI 如何改变日常工作》，现在你可以免费获取。

It's called "Aligning for the Agentic Era: How AI is Changing Everyday Work," and right now it's yours absolutely free.

里面会告诉你 AI 如何重塑整个组织的工作方式，以及如何让你的企业从中受益。

Inside, you'll discover how AI could reshape work across your entire organization, and how to position your business to prosper from it.

我自己在用，你也应该试试。

I use this, and you should too.

如果你的收入至少达到七位数，就去 netsuite.com/tip 免费获取我们的商业指南《为智能体时代做好准备：AI 如何改变日常工作》。

If your revenues are at least in the seven figures, get our free business guide, "Aligning for the Agentic Era: How AI is Changing Everyday Work" at netsuite.com/tip.

这份指南在 netsuite.com/tip 上免费提供。

The guide is free to you at netsuite.com/tip.

netsuite.com/tip。

netsuite.com/tip.

好了，回到节目。

All right, back to the show.

我真没想到你会对支付业务这么兴奋。

I would've never thought that I see you as excited about a payments business.

但不得不说，这让我想起了美客多，电商加支付的飞轮效应非常强大。

But I got to say, it kind of reminds me of Mercado Libre, where the flywheel of having the e-commerce plus the payments is incredibly powerful.

而且显然，对 Shopify 来说最重要的是，你保住了抽成率。

And obviously, and that's the most important thing to Shopify, you're keeping the take rate.

所以在这种情况下，比如 ChatGPT 只是门面，可以这么说。

So in this case, for example, ChatGPT is only the shop window, so to say.

我觉得没错。

I think that's right.

所以只要通过ChatGPT卖出的每一美元，和通过商家自己网站卖出的每一美元，给Shopify带来的收入是一样的，那Shopify在智能体商业时代就会过得很好，甚至可能更兴旺，因为我们的智能体可能会推动整体消费增长。

And so long as a dollar sold through ChatGPT pays Shopify the same as a dollar that goes through the merchant's own website, then Shopify is going to do just fine in the age of agentic commerce and maybe even thrive, as perhaps our agents will drive more consumption, generally.

随便花20美元买点东西会变得非常顺畅，几乎没有摩擦。

Become so much more frictionless to just spend $20 on something arbitrary.

而且你看看Shopify的数据，智能体已经变得越来越重要了。

And when you look at especially Shopify's data, it looks like agents are becoming already more and more important.

具体数字来看，从2025年1月起，归因于AI搜索的Shopify店铺订单增长了大约15倍。

So putting some numbers on it, orders to Shopify stores attributed to AI search are roughly 15X since January of 2025.

仅第二季度，AI带来的流量和归因于AI的订单都同比增长了三倍，而且AI引荐的访问中有一半直接落在产品页面上。

And in Q2 alone, both AI-driven traffic and AI-attributed orders tripled year over year, and half of AI referred sessions land directly on a product page.

这基本上是传统搜索转化率的两倍半。

So that's basically where you have two and a half times the rate of traditional search.

所以很多人跳过了大量浏览环节。

So people are skipping a lot of the browsing.

然后在一次高盛会议上，我记得是九月份，CFO说通过那些AI渠道来的顾客，转化率大概能高出80%。

And then at a Goldman Sachs conference in, I think it was September, the CFO said that shoppers convert roughly 80% better if they come through those AI rails, basically.

而且通过AI渠道来的新买家，比例几乎是其他渠道的两倍。

And then new buyers arrive from AI channels at nearly twice the rate of other channels.

所以某种程度上，我觉得这些数字挺多的，听起来好像进展比实际情况要快一些。

So to some extent, I feel like those are a lot of numbers and they make it sound like it's maybe a bit further than it actually is.

但如果你看整体数字，比如有75%的AI归因购买来自前100大产品类别之外，这真的很疯狂，你想想看。

But then if you look at the total numbers, for example, you have this crazy 75% of AI-attributed purchases coming from outside of the top 100 product categories, and that's insane if you think about it.

嗯，这个数据确实很重要，你能再解释一下这意味着什么吗？这样我们都能更好地理解。

Well, it's a really important stat, so maybe you can just explain more of what that means so we can all appreciate it.

好的，当然。

Yeah, sure.

一般来说，传统搜索是按热度排名的，对吧？

Generally, traditional search ranks things by popularity, right?

所以这对那些有大笔SEO预算的大品牌有利，他们能确保自己的页面在谷歌的任何相关搜索中都能排到前面。

So that favors big brands with big SEO budgets who can ensure that their pages land near the top for any relevant search on Google.

因为你知道我也知道，在谷歌上找东西，你绝对不会翻到第三页。

Because as you and I know, you would never go to page three if you want to find anything on Google.

要么就在第一页靠前的位置，要么你根本看不到它。

Either it's on page one and pretty up the top, or you just never see it.

但AI代理是根据具体的意图来匹配的。

But AI agents, they match on specific intent.

比如在财报电话会议上，Shopify的管理层讲了一个故事：有位家长想给10岁的孩子找一款无屏幕手机，结果被引导到了一个叫Tin Can Phone的小型Shopify商家，那正是他们想要的。

So on an earnings call, for example, Shopify's management told, there was a story about a parent searching for a screenless phone for their 10-year-old kid, and they were being led to a small Shopify merchant called Tin Can Phone, and it's pretty much exactly what they wanted.

但如果他们用谷歌搜索，根本找不到这个商家。

And yet, if they would've searched through Google, they would've never found it.

所以从某种程度上说，AI发现机制更有利于那些小众的、专门化的长尾商家。

So in some ways, AI discovery favors these long tail of small, specialized merchants.

你猜怎么着？

And guess what?

那正是Shopify的客户群体。

That's exactly what Shopify's customer base is.

完全正确。

That's exactly right.

而且关于这一点，MoffettNathanson发布了一些数据，衡量AI平台给网店带来多少流量，并与谷歌搜索做了对比。

And to that point, MoffettNathanson has released some data measuring how much traffic AI platforms send to online stores and compared that with Google Search.

为了说明背景，他们提到的只是那种泛泛的品牌。

And for context, they refer to just kind of the generic brand.

Brand.com指的是一个公司及其自己的网站。

Brand.com is a company and their own website.

所以听到这个的时候，你可以想象任何在Shopify上运营的独立店铺。

So think of any independent store running on Shopify when you hear that.

但在常规的谷歌搜索中，这些brand.com、这些类似Shopify的网站的流量增长速度，大约是亚马逊、Wayfair、Chewy和Etsy这些大型市场平台的两倍。

But in regular Google search, traffic to those brand.com, those Shopify-type sites grew about two times faster than traffic to big marketplaces like Amazon, Wayfair, Chewy, and Etsy.

这又是一个听起来很惊人的数据。

Which is another one of those stats that sound incredible.

但公平地说，这也是意料之中的，对吧？

But to be completely fair, it is what you would expect, right?

最好的Shopify品牌应该比亚马逊增长得更快。

The best Shopify brands should be outgrowing Amazon.

嗯，大多数情况下，这只是规模的问题。

Well, for the most part, it's just a matter of scale.

但是当他们汇总来自AI平台的流量时，这些品牌官网的流量增长速度基本上是传统市场平台的四到八倍。

But when they aggregated the traffic coming from AI platforms, this brand.com was basically outpacing the legacy marketplaces by four to eight times.

所以莫顿说，他在这个漏斗里看到的品牌官网比以往任何时候都多，现在这些也反映在了实际数据中。

And so Morton has said he's seeing more brand.com sites in this funnel than ever before, and now it's showing up in the measured data.

正是这些数据让他今年坚定了上调股票评级的信心。

And that data is what drove his conviction to upgrade the stock this year.

他还说，引用原话：“GMV的增长会比大多数人预期的更高、更持久。”

And he said, quote, "GMV is going to grow higher for longer than most people think."

确实，肯定比我们去年预期的要持久。

Well, definitely for longer than we thought last year.

我觉得换句话说，简单来说，AI聊天机器人正在为小型在线商店带来更多流量，而不是更少。

I think in other words, if you just want to put it more simply, AI chatbots are driving more traffic to small online stores and not less.

这挺有意思的，因为我觉得在90年代，很多人以为互联网会让零售商遍地开花，上架产品的门槛会被消除，这样一来，大零售商的权力就会被削弱。

Which is kind of interesting, because I think in the '90s, many folks thought the internet would lead to a proliferation in retailers, and then the barriers to listing products would be erased, and that would basically take power away from the largest retailers.

但后来我们都知道实际发生了什么。

But then we all know what really happened.

就是亚马逊进来了，利用互联网把成千上万的第三方零售商整合到自己的平台上。

It was Amazon coming in, using the internet to consolidate thousands of third-party retailers on their platform.

但你引用的数据表明，也许现在，AI聊天机器人可能不是在给沃尔玛、亚马逊这些巨头更多权力，而是在帮助独立品牌，借用托比的话说，就是在给反抗军武装。

But the data you cited suggests that maybe now, rather than handing even more power to Walmart and Amazon or those sort of huge retailers, AI chatbots may actually be helping independent brands, sort of arming the rebels, to borrow Tobi's line here.

而且，如果我没记错的话，这里还有一个重要的二阶效应，比如我们看看Etsy。

And there's an important second-order effect here too, if I'm not mistaken, where if we look, for example, at Etsy.

这家公司超过100%的营业利润都来自我们之前提到的站内广告，同时向卖家收取大约25%的综合费率。

The company earns over 100% of its operating profit from these on-site ads that we mentioned earlier, while charging sellers an all-in take rate of around 25%.

所以，就像我们说的，在代理商务的世界里，这种商业模式是撑不住的，因为代理是替人购物的，它们不会像你我那样被广告影响，尽管我们可能觉得自己不会受影响。

So, as we talked about, in a world of agentic commerce, that business model won't last because agents are doing the shopping on behalf of people, and they're not swayed by advertising in the same way that maybe you and I are, even though we like to think that we probably aren't.

而且这一点越被证实，一个成功的Etsy卖家可能就会开始问："那我为什么要付25%的费用来被找到呢？我自己开个Shopify店铺，现在被谷歌AI搜索或ChatGPT推荐的机会比以往任何时候都大，而且成本只是那的一小部分。"

And the more that proves to be true, a successful Etsy seller might start asking, "Well, why am I paying 25% to be found when a Shopify store of my own now has a better chance than ever being recommended by Google's AI search or ChatGPT at basically a fraction of the cost?"

不过反过来想，就像我们之前说的，要抓取整个互联网需要大量的计算资源。

The counterpoint would be that it takes significantly more compute to scrape the internet, as we talked about earlier.

所以理论上，智能代理应该更倾向于像亚马逊这样索引完善的平台，这样它们能更轻松地找到想要的东西。

So agents should, in theory, favor well-indexed platforms like Amazon, where they can find what they're looking for with much less effort.

所以，品牌官网在AI推荐上表现优于亚马逊，这确实证明了Shopify在Shopify Catalog上所做工作的重要性。

So the fact that brand.com is outperforming Amazon on AI referrals is really a testament to the importance of what Shopify has built in Shopify Catalog.

说到底，我觉得像亚马逊这样能实际发货的市场平台，对商家还是会有很大的控制力，因为AI可能会改变我们购物的方式，但它改变不了的事实是，你需要庞大的物流网络才能真正把东西送到你家门口。

Ultimately, I think marketplaces that physically fulfill, like Amazon, for example, will maintain lots of leverage over their merchants because AI might revolutionize how we shop, but it can't really change the reality that you need a massive logistics footprint to actually bring things to your door.

而真正被挤压的是那些纯粹做发现和导购的中间商，就像我们刚才提到的Etsy的例子。

Whereas pure discovery middlemen are the ones actually getting squeezed, going back to the Etsy example.

而且，从Etsy流失的商家，最终很可能会流向Shopify。

And the idea is that merchants defecting from Etsy will most likely eventually find their way to Shopify.

如果我们再往后退一步看，实际上只有大约20%的零售交易是在线上完成的。

And if we really step all the way back, only about 20% of retail commerce happens online.

如果你在90年代告诉别人，到2026年，所有零售购物中有20%会在线上进行，我想大家都会非常震惊。

And if you told somebody in the '90s that that would be the case in 2026, that 20% of all retail shopping would be online, I think people would've been really shocked by that.

而且不管怎样，Shopify的管理层一直认为，消除购买过程中的摩擦，给每个人一个相当于口袋里的私人购物助手，会推动更多零售支出转移到线上，这个观点我认同。

And anyways, Shopify management has argued to this point, and I buy it, that stripping out the friction from buying and giving everyone what amounts to a personal shopper in their pocket will push more retail spending online.

所以我们将会看到，线上零售在整体零售中的占比会更高。

So we'll see a higher percentage of all retail commerce occurring online.

如果这个观点哪怕只有部分正确，那么整个蛋糕对亚马逊和Shopify来说都会变得更大。

And if that's even partly right, then the whole pie gets bigger for both Amazon and Shopify.

所以我们又一次看到了这个模式：上市公司的CEO声称AI实际上对他们的业务有利。

So once again, we've come across the pattern here where a CEO of a publicly traded company claims that AI is actually good for their business.

看起来每个人都是这么说的，但在这个案例中，我确实觉得Shopify的逻辑相当扎实。

It seems like everyone does that, but actually in this case, I do think the logic for Shopify is pretty solid.

对，我觉得这里说得更准确。

Yeah, I think it's more accurate here.

所以也许回到我们最初的那期节目，我认为我们最大的结构性担忧是亚马逊的Buy with Prime计划，它针对第三方网站，有点像是一个警告信号——卖家不再需要选择Shopify生态系统，而是可以通过亚马逊为顾客提供配送服务，即使他们不在亚马逊市场上销售。

So maybe just to go back to our original episode, I think our biggest structural worry was Amazon and their Buy with Prime initiative for third-party websites sort of being a warning shot where instead of opting into the Shopify ecosystem, sellers could now offer their customers deliveries via Amazon, even when they're not selling on the Amazon Marketplace.

那对Shopify来说是个相当大的打击，而且这其实是我们当时怎么说的。

And that was a pretty big stab onto Shopify's toes, and this is actually how we put it back then.

“让我担心的是，亚马逊可能还有几张王牌，如果情况真的恶化，他们可能会拿出来进一步打击Shopify。”

"What scares me is that Amazon might have some aces up its sleeve that it could pull out to damage Shopify further if push really came to shove."

那么现在，一年半过去了，既然亚马逊已经签署了Shopify基本上参与制定的协议，这个担忧还成立吗？还是你的看法有所改变？

So now, a year and a half later, now that Amazon has signed on the protocol Shopify basically co-wrote, does that fear still hold up or did your opinion change on that?

我觉得，尽管Buy with Prime的推动对Shopify来说看起来相当不利，尤其是在智能体商务时代到来之前，但似乎AI时代已经打乱了旧的战线。

I feel like for as bearish as this Buy with Prime push seemed to be for Shopify, especially in a pre-agentic commerce world, it does seem like the AI era has just scrambled the old battle lines.

所以，在一个购物从Claude或Muse开始，而不是直接在amazon.com或亚马逊应用上开始的世界里，即使是亚马逊也有兴趣让更开放的轨道存在。

And so in a world where shopping starts inside of Claude or Muse rather than directly on amazon.com or in the Amazon app, even Amazon has an interest in more open rails existing.

而大家最终都落脚的轨道，还是Shopify和谷歌一起搭建的那些。

And so the rails everyone is ending up on are, again, the ones built by Shopify alongside Google.

如果你要站在看空Shopify的人的立场上，你会怎么看？

And if you would have to step into the shoes of a Shopify bear, then how would you look at it?

所以，如果我要持怀疑态度，我会说，标准组织加入成本很低，而且最终，亚马逊仍然拥有地球上最大的封闭商务生态系统。

So maybe if I were going to be skeptical, I would say that the standards bodies are cheap to join, and ultimately, Amazon still owns the largest closed commerce ecosystem on Earth.

所以他们还是会遵循那些激励。

So they're going to follow those incentives.

而且我们也看到了，亚马逊封杀了Muse，今年早些时候还向OpenAI投了500亿美元。

And as we've seen, Amazon blocked Muse, and they completed a $50 billion investment in OpenAI earlier this year.

巧合的是——我完全是在猜测——就在那个消息公布几周后，OpenAI就搁置了那个即时结账实验，而Shopify的商家从2025年秋天起就一直在接入这个功能。

And coincidentally, and I'm completely speculating here, but within a few weeks of that announcement, we did see OpenAI shelve its instant checkout experiment that Shopify merchants had been plugging into since the fall of 2025.

所以亚马逊铁了心要尽可能维持现状，但谁知道他们能撑多久呢？

And so Amazon is hellbent on maintaining the status quo for as long as possible and Who's to say how long they'll be able to do that for?

另外，这一切变化实在太快了。

And then the other thing is this is all just changing so fast.

确实。

Absolutely.

而且我其实希望ChatGPT的结账功能不是这笔投资的唯一原因，因为花500亿就为了这个，我觉得……这代价也太高了。

And I actually hope that the checkout process of ChatGPT was not the only reason for the investment, because 50 billion for that seems like- ... quite a hefty price tag, I would say.

但如果现在要我问你，你最终怎么看？

But if I would have to ask you now, where do you net out?

你会怎么总结？

Where would that be?

就在大约一年前，亚马逊和Shopify之间似乎还是一种对立关系，对吧？

In just a year or so, it seemed like there was a dichotomy between Amazon and Shopify, right?

从长远来看，似乎总有一方要赢，另一方就得输。

Long term, one, it seemed like, would have to win at the expense of the other.

但现在这种感觉越来越不准确了。

But that really feels like a less accurate depiction of reality.

随着AI的进步和越来越深入地融入我们的生活，我觉得现在更准确的说法是，电商这块蛋糕会变大，而且可能会大很多，这对两家都有好处。

As AI improves and becomes more embedded in our lives, I think the more accurate way to think about it now is the e-commerce pie will grow, and probably grow by a lot, and that's good for both of them.

不过目前我们还不知道大语言模型和智能体会如何具体改变我们的购物方式，所以这还有待观察。

And yet we don't know at this point how LLMs and agents will change the way we shop exactly, so that remains to be seen.

至于这会不会偏向某一方，也是有可能的。

And whether that will favor one or the other is possible.

有可能两家都赢，也有可能两家都被某个现在还不存在的竞争对手抢走地盘。

It's possible that both will win, or it's possible that both will lose ground to some competitor that doesn't even exist today.

我觉得对我来说，这里更大的问题之一是，AI最终会成为导流渠道还是分销商。

I think to me, one of the bigger questions here is whether AI will end up being a referral channel or a distributor.

所以，分销商基本上就是插在你这个在线商家和你的客户之间，可能会把几十年的客户关系拱手让给别人。

So a distributor basically wedges itself between you as an online merchant and your customer, potentially handing decades-old customer relationships to someone else.

但如果它最终只是一个推荐渠道，那基本上就是谷歌搜索的升级版，让商家保留客户关系、结账流程和数据，因此也保留了大部分主动权。

But then if it actually ends up being a referral channel, on the other hand, it's basically just a better version of Google Search that allows the merchant to keep the relationship and the checkout and the data, and therefore, also most of the takeaway leverage.

那你觉得呢？

Well, you tell me.

就我所见，AI聊天机器人看起来更像是一个新的推荐渠道，而不是分销商，对吧？

From what I've seen, AI chatbots look more like a new referral channel than a distributor, right?

要让Anthropic、OpenAI或Meta成为拥有客户的中间商，他们真的得掌控整个购物体验，而且我们都得习惯让智能体替我们买所有东西。

For Anthropic, OpenAI, or Meta to become the middleman that owns the customer, they would really have to control the whole shopping experience, and we'd all have to get comfortable letting agents buy everything for us.

我们之前也聊过，这确实是智能体电商爱好者们要面对的大挑战。

And we've talked about that being really the big challenge for Agentic Commerce bros to solve.

这总是很有趣，因为你比我买的东西多得多。

Well, it's always funny because you order so much more stuff than I do.

你跟我说你每天从亚马逊收五个包裹，而我可能一年才在亚马逊上买五次东西。

You're telling me about five packages a day from Amazon, and I order maybe five times a year on Amazon.

不过，我完全能想象让一个智能代理帮我买东西，而你可能是我们俩中更持怀疑态度的人。

Yet, I can absolutely imagine an agent ordering stuff for me, and you're the more skeptical person, I think, from the two of us.

我也不知道，也许我应该听你的，毕竟你网购经验更丰富，但我不确定。

And I don't know, perhaps I should listen to you as the more experienced online shopper, but I don't know.

就像你说的，我觉得OpenAI搁置了即时结账功能，现在把顾客引导到商家自己的网站去完成购买。

Like you said, I think OpenAI shelved instant checkout and now sends shoppers to the merchant's own site to finish the purchase.

所以也许可以说，OpenAI和Anthropic对市场的那个部分并不真正感兴趣。

So maybe you could say that OpenAI and also Anthropic, they're not really interested in that part of the market.

他们也在进军企业领域。

They're also moving into the enterprise field.

所以你可能读到过这样的说法：ChatGPT更需要亚马逊，而不是亚马逊更需要ChatGPT。

So you might have read that as ChatGPT needs Amazon more than Amazon needs ChatGPT.

而且，谷歌和Meta多年前就发现，如果你只是全球最大的推荐引擎，那能有多赚钱。

And anyways, Google and Meta figured out years ago how wildly lucrative it can be if you're just the world's biggest referral engine.

对他们两家来说，这确实效果很好。

It worked pretty well for the two of them.

不过，现在大家谈论Muse的方式，而且很遗憾，我还没能测试它，因为它还没在欧洲上线。

That said, the way people currently talk about Muse, and again, unfortunately, I couldn't yet test it because it's not yet out in Europe.

在我看来，它更像是一个分销渠道，而不是推荐渠道。

To me, it looks more like the distributor and not a referral channel.

所以我不太确定。

So I don't know.

也许这就是为什么Shopify的股价在Muse最初发布后从大约155美元跌到大约130美元，然后亚马逊又阻止了Muse在其网站上购物。

Perhaps that's why Shopify's shares slid from around 155 bucks to about 130 bucks after Muse was initially announced, and then Amazon blocked Muse from shopping on its site.

而就在第二天，Shopify基本上宣布了，我引用一下，“与Muse深度合作，在所有Shopify商店中通过Shop Pay实现代理结账。”

While just the next day, Shopify basically announced, and I quote this, "Partnering deeply with Muse to enable agentic checkout with Shop Pay on all Shopify stores."

然后市场对此反应相当积极。

And then the stock took that as a pretty good sign.

我想接下来两天股价上涨了14%。

I think it rose 14% over the next two days.

所以我不太确定。

So I don't know.

我觉得市场好像也不太清楚该怎么应对这个，就像我们俩一样。

It seems to me like the market doesn't exactly know what to do with this either, kind of like the two of us.

我想说，这期节目之后，我们48小时内都不提“代理”或“代理式”这些词了。

I'm going to say that we take a 48-hour hiatus from saying the words agents or agentic after this episode.

不过，玩笑归玩笑，其实，跟听众们说一声，昨天我和丹尼尔通了个电话，我手机上下载了Muse，也稍微玩了一下。

But yeah, jokes aside, I actually, for the listeners, Daniel and I were on a call yesterday, and I do have Muse downloaded on my phone and have played around with it a bit.

我基本上是把我的iPhone投屏到电脑上，然后跟丹尼尔共享屏幕，这样他就能看到我在Muse应用里的操作。

And I basically mirrored my iPhone onto my computer for the call and then screen shared my screen with Daniel, and so he could see me going through the Muse app.

所以他对这个也有点了解。

So he's got a little bit of exposure to it.

回到Shopify的话题，看起来代理式销售的经济效益跟Shopify商家任何其他在线交易完全一样，因为每次在Shopify商店通过Muse购买都会走Shop Pay。

And so just to go back to Shopify, apparently, it seems agentic sales will carry the exact same economics of really any other online store transaction from a Shopify merchant, because every Muse purchase on a Shopify store is going to run through Shop Pay.

所以九月份的时候，你找不到比Shopify和亚马逊之间更鲜明的对比了。

So in September, you couldn't have asked for a clearer contrast between Shopify and Amazon.

确实，这是第一次有热门AI代理找上亚马逊的门，结果被拒之门外，而Shopify却全力押注，把旗下所有商店都跟Muse紧密绑定。

Really, the first time a popular AI agent showed up at Amazon's door, Amazon turned them away, while Shopify has just doubled down on it and opened every one of its stores to really be tied in with Muse.

就像你说的，我们现在真的还太早，事情变化太快了，几乎很难说未来会怎么发展。

And as you said, we're basically just so early, and things are changing so fast that it almost seems impossible to say how things will play out.

而且说实话，我们连亚马逊和Shopify自己是怎么想的都不知道。

And actually, we don't even know how Amazon and Shopify actually think about it.

如果他们心里有数，知道五年后应该是什么样子，我完全可以想象，比如亚马逊会慢慢改变对agent的看法或者处理方式。

If they have a vision for how this is supposed to look in five years' time, I could very well imagine that, for example, Amazon is changing how they think or approach agents over time.

可能也取决于他们自己的agent能有多成功，尤其是在亚马逊生态系统之外。

Maybe also dependent on how successful their own agents will become, especially outside of the Amazon ecosystem.

但就在Shopify交易两天后，马克·扎克伯格说Meta计划通过从Muse完成的交易中抽取一小笔费用来赚钱。

But just two days after the Shopify deal, Mark Zuckerberg said that Meta plans to make money from Muse by taking a small fee on the transactions that it completes.

他没说费用会是多少，也没说谁来付，Shopify声称自己的分成未来不会变，但那个费用总得有人出。

So he didn't say how big the fee will be or who pays it, and Shopify claims its own cut will not change in the future, but that fee has to come from somewhere.

这点我们很清楚。

We know that.

而Meta是第一个公开表示想对交易本身收费的主要AI平台，这再次说明，他们更倾向于我们之前讨论的那种分销商模式，而不是仅仅做一个推荐渠道。

And Meta is the first major AI platform to say out loud that it wants to basically take a toll on the transaction itself, which, again, would be a big step toward that distributor framework we talked about earlier, as opposed to just being a referral channel.

为了继续推进我们的讨论，也许我们该再深入探讨一个看跌的观点。

So to keep us moving along here, maybe there's one other bear case point we should linger on.

你经常从科技投资者那里听到这样的说法：在AI编写软件的世界里，商家会自己用AI生成代码来搭建类似Shopify的平台，然后不再向Shopify这个中间商支付费用。

And you hear it a lot from folks that invest in tech, and it's something along the lines of, in a world where AI writes the software, merchants will just vibe code their own Shopify and then stop paying tolls to Shopify as a middleman.

我们在节目中报道过的许多软件公司都曾面临这种情况，所谓的SaaS末日论，即担心AI会取代它们，导致Adobe、CoStar和Intuit等公司的股票被抛售。

And so We've heard that so many times about software companies we've covered on this show that have been caught up in the so-called SaaS apocalypse, the sell-off from these stocks on fears that AI will replace them with companies like Adobe, CoStar, and Intuit.

但我认为市场对这个想法有点过度反应了。

But I just think that the market has taken things one step too far with that idea.

对，放下你的显示器和Excel表格，去外面走走，接触一下现实，和真正的商家聊聊，你会发现没有人真的想用AI生成代码来替代Shopify提供的核心服务。

Yeah, just step out from behind the monitors and the Excel sheets, go touch some grass, maybe go talk to some real merchants, and I think you'll find that no one is actually trying to vibe code essential functions like the services that Shopify provides.

相比这样做的风险和挑战，节省的成本简直微不足道。

The savings are just so negligible compared to the risks and challenges of doing so.

我不知道。

I don't know.

也许这个观点以后会被证明是错的，但这是我最为怀疑的一点，而且我认为单干真的省不了钱，因为你失去了Shopify的规模优势。

Maybe that won't age well, but that's the thing I'm most skeptical about, and I just don't think going it alone would even really save money because you're missing out on Shopify's scale.

比如Shopify，它跟Stripe谈的费率比任何单个商家自己谈的都要好。

So Shopify, for example, has a better fee structure with Stripe than any individual merchant could ever negotiate.

所以你要是自己单干，去随便写个代码，就享受不到这个优势了。

And so you lose that when you go off on your own to vibe code something.

而且Shopify的商家可以一键接入Meta的广告追踪工具，这个整合现在能存在，是因为Shopify的CEO和马克·扎克伯格亲自谈成的合作。

And Shopify merchants get a one-click connection to Meta's ad tracking tools, and that's an integration that exists now because literally the CEO of Shopify and Mark Zuckerberg personally worked on that deal together.

我完全能理解，Shopify提供的价值很大程度上在软件之外，所以不太可能被随便写个代码就替代掉。

I could definitely see how Shopify offers enough value, basically outside of the software to not be vibe coded away by anyone.

就像你说的，物流、广告，还有那些幕后的工作，都得靠这样的平台来做。

As you said, it's between logistics, the advertising, and all that basically behind the curtain work that has to be done by such a platform.

几乎不可能想象有人能靠随便写代码搞定这些，虽然长远来看可能会限制他们的定价能力，但比起Shopify，我更担心的是其他公司。

It's almost impossible to imagine that anyone could vibe code any of this, and it might limit their pricing power way down the line, but they are definitely businesses I'm a lot more concerned about than Shopify.

说实话，每次讨论投资时，如果不用再纠结AI的问题，我反而挺高兴的。

And honestly, I'm kind of glad whenever I don't have to make a call about AI anymore when we talk about an investment.

我觉得现在有个被低估的能力，就是承认有些事情现在无法预知，所以它们不是我们通常该找的那种低难度目标，至少我们在投资时应该尽量找这样的目标。

And I think it has become an underrated skill to also accept that some things are unknowable today and therefore not the one-foot hurdles that we usually should look for, at least we try to look for in those investments.

而且就像你之前提到的，我们也不该忘记，算力和大语言模型在现实世界中是有限制的。

And as you mentioned before, we also shouldn't forget that there are real-world constraints on compute and LLMs.

这也不是什么神奇的技术，虽然有时候感觉像。

And this isn't some magic technology, even though it sometimes feels like that.

它不会一夜之间就让所有软件平台都过时。

It just renders every software platform obsolete overnight.

我们一直在聊迈克尔·莫顿，还引用他的话，因为我觉得他对Shopify的看法非常清晰。

We keep on talking about Michael Morton and quoting him because I think he's very clear about Shopify.

基本上，他认为即使有无限的资源、一支工程师大军，还有Anthropic给的空白支票，你也没法重建一个比直接付费使用Shopify更便宜的东西，还能享受随之而来的所有好处，比如拿到Stripe的批量折扣。

And so basically, he has argued that even with unlimited resources, an army of engineers, and a blank check from Anthropic, you could not rebuild it to be cheaper than just paying to use Shopify and enjoying all the benefits that come with that, to the point of getting volume discounts with Stripe.

我觉得这在更广泛的软件讨论中也是成立的。

And this is true, I think, in the broader software debate, too.

现实很无聊，东西就是会坏。

The boring reality is that stuff just breaks.

如果你是一个在TikTok、Etsy、亚马逊还有自己网站上卖货的商家，然后你用了vibe coding搞了一套替代Shopify的方案，结果在黑色星期五出了故障，那你就只能靠自己了，生意就毁了，你为了省几毛钱把自己坑了。

If you're a merchant who's selling on TikTok and Etsy and Amazon, and also your own website, and then you've vibe coded this alternative stack to Shopify, and you run into a glitch on Black Friday, well, you're on your own and your business is on fire, and you shot yourself in the foot just to save a few pennies.

我都有点不好意思了，但最后我还是得再聊一次AI。

I almost feel bad about it, but I got to get back to AI one last time.

那你怎么看AI故事的另一半，就是AI来运营商家业务这个想法？

So what's your take on the other half of the AI story, where the idea is basically that AI runs the merchant's business?

这应该就是Shopify的Sidekick产品，如果你愿意这么叫的话，它就像是内置在Shopify里的一个AI业务伙伴。

So this is Shopify's, I think, Sidekick product, if you want to call it that way, which is meant to be like an AI business partner built right into Shopify.

它了解你的产品、你的客户、你所有的交易记录，而且还能在你的店铺里实际操作。

So it knows your products, it knows your customers, your entire transaction history, and it can also actually build within your store.

它能编辑你的产品图片，分析你的数据，还能创建定制应用。

So it edits your product photos, it analyzes your data, and it creates custom apps.

我觉得听起来挺不错的，而且使用量似乎也很高。

I think it sounds pretty good, and usage seems to be pretty strong.

我觉得每天活跃使用Sidekick的商家数量同比增长了3.5倍。

I think daily active merchants using Sidekick is up three and a half times year over year.

每日会话量增长了近5倍，而且仅今年第二季度就有近3400万次对话。

Daily sessions is up almost five times, and we have nearly 34 million conversations just in Q2 of this year.

而且仅仅一年内，商家就通过它创建了36000个定制应用。

And then you have 36,000 custom apps built by merchants through it in just a single year.

这基本上是上一季度的三倍。

So that's basically triple to the prior quarter.

所以，数字看起来确实惊人，但你觉得长期来看怎么样？

So again, the numbers look insane, but what's your take on it in the long term?

这简直太疯狂了。

It's just nuts.

对于那些四处观望、想知道ChatGPT发布四年后AI到底会如何影响整个经济生产力的人来说，我觉得这就是一个绝佳的例子。

For anyone looking around wondering how AI will actually impact productivity across the economy four years after the release of ChatGPT, I think that's a heck of an example.

我还读到，在Sidekick的帮助下，由Shopify支持的新店铺中，有近10%在注册当天就完成了首笔销售。

And I've also read that with the help of Sidekick, nearly 10% of new shops powered by Shopify now make their first sale within a single day of signing up.

而这个比例在短短三年内就增长了43%。

And so that share is up 43% in just three years.

而且教皇当选时，教皇T恤在一个小时内就上架开卖了。

And when the Pope was elected, Da Pope T-shirts were on sale within an hour.

所以Shopify把这叫做文化速度下的商业。

And so Shopify calls it commerce at the speed of culture.

而且每次创业变得更容易，Shopify的商家漏斗就越来越宽，然后我们就有了这些按需供应的Da Pope T恤。

And every time it gets easier to start, Shopify's funnel of merchants just gets wider and wider, and then we get these Da Pope T-shirts available on demand.

作为一个密切关注这家公司的人，我希望你买了一件那样的T恤。

As someone who follows the company quite closely, I hope you got one of those T-shirts.

我觉得你没买，不然我肯定看到了，但说真的，挺可惜的。

I don't think you did, because otherwise I would have seen it, but it's a shame, I got to say that.

我可能比你感受更复杂一些，因为Shopify的总裁称他们可能是全球最拥抱AI的公司，意思是他们内部全面采用了AI。

And I probably feel a bit more mixed about it than you, but just because Shopify's president has called them probably the most AI-pilled company in the world, meaning they've gone all in using AI internally.

每当这样做，尤其是在这种时候，还拿出这么亮眼的数据，在我看来，即使他们运作得不错，也可能稍微有点过度宣传。

And whenever you do that, especially at these times, and you come out with these incredible numbers, to me, it seems like even if they're working quite well, they might oversell it just slightly.

不过，得承认，他们的员工人数两年基本没变，而那时收入增长超过60%。

To their credit, though, headcount has been roughly flat for two years, while revenue was growing at 60-plus percent then.

我不能说这全靠AI，但正是这种模式让运营费用从2024年第二季度的42%降到了上季度的34%。

I can't say whether that's all thanks to AI, but that formula is how operating expenses have been able to fall from 42% in Q2 of 2024 to 34% last quarter.

所以这影响非常大。

So that's a major, major impact.

确实如此，而且我觉得正因为这样，说AI已经在收入和利润率上产生了巨大影响是有道理的，可能比我们看过的任何其他公司都要明显。

And that's the case, and I think it's fair because of that, that you can make a case that AI is already showing up in a big way for revenues and margins, probably more than with any other company we've yet looked at.

也许他们确实是全球最痴迷AI的公司。

Maybe they are the most AI-pilled company in the world.

有可能。

They might be.

我明白你的意思，Airbnb的布莱恩·切斯基也经常谈到这个观点，就是AI会在内部帮助提升利润率。

I do see the point, and Brian Chesky at Airbnb has talked about this a lot, of the idea that AI will help margins internally.

但显然，这不是没有代价的。

But obviously, that doesn't come for free.

我敢肯定Shopify在AI计算上的开销简直高得离谱。

I'm sure Shopify's AI compute bills are just completely crazy.

之前我们听说Uber的时候，他们对公司在AI计算上的花费感到非常震惊，因为他们把这些工具交给了员工使用。

We heard about with Uber a while back that they had been pretty shocked by how much the company was spending On AI compute by handing over these kind of tools to their employees.

所以Shopify的CEO其实也承认了，由于Sidekick的大量使用，AI计算成本上升，给毛利率带来了一些压力。

And so Shopify's CEO has actually acknowledged that rising AI compute costs from all that Sidekick usage are putting some pressure on gross margins.

这也让我想起Adobe管理层因为免费提供计算密集型的AI创意工具而惹恼了投资者，他们的希望是用户会先对这些工具上瘾，然后随着时间推移，再逐步将这些AI工具变现。

And that also reminds me of how Adobe's management has gotten in trouble with investors for offering AI creative tools for free that are compute-intensive, and the hope there at least is that users would get hooked on the tools and then over time, they can turn to monetizing those AI tools more.

所以现在这算是一种亏本引流策略，用来获取客户。

So sort of a loss leader right now and where they're functioning for customer acquisition.

我指的是Adobe，但我觉得这个逻辑可能也适用于Shopify。

And so I'm referring to Adobe, but I think that logic probably applies to Shopify as well.

有可能，不过我得说，如果你有Shopify那样的增长，我猜市场给你的信任度会比Adobe过去几年得到的要多一些。

It might be, although I've got to say, if you have the growth of Shopify, I assume that the market will likely give you a bit more credit than Adobe has gotten in the last couple of years.

我觉得我们今天聊了很多内容，要不我们进入估值部分吧，基本上就是18个月前你第一次看这家公司时，我们决定不投资的原因，而它可能是我们在节目里看过的最优质的公司之一。

And I think we covered a lot of ground today and how about we get to the valuation section, basically the part that 18 months ago, when you first looked at the company, was the reason why we did not invest in one of the probably highest quality companies that we looked at here for the show.

那你觉得今天Shopify这家公司值多少钱？

So what do you think today the company Shopify is worth?

是的，你可以想象，我得对我的模型做一些更新。

So yeah, as you can imagine, I had some updates to make to my model.

我们会在下面的节目笔记里附上这个模型的链接，想看的人可以去看看。

And we'll have that model linked to in the show notes below for anyone who wants to see it.

除了上调未来几年的营收预期——这部分我之前确实低估了——最大的变化是把营业利润率从今年的14%提高到未来五年接近20%。

But besides revising higher their revenue over the next few years, where I was really underestimating things, the biggest change was setting their operating margins to increase to nearly 20% in the next five years, up from 14% this year.

利润率的提升能让营业利润的增长速度快于营收，这对股东来说可是件大好事。

So that jump in margins enables operating profits to actually grow faster than revenues, and that would be a wonderful thing for shareholders.

我还调整了模型，以反映Shopify——我们之前聊过——已经从大规模稀释现有股东，转变为通过回购保持股本数量不变。

And I also adjusted my model to reflect that Shopify, we talked about this earlier, has gone from being a massive net diluter of existing shareholders to, thanks to buybacks, being able to keep their share count flat.

当我们考虑双引擎，也就是盈利增长和市盈率时，你现在给这家公司定的市盈率是多少？

And when we think about the twin engine, so earnings growth and the multiple, what's the multiple that you're now putting on the business?

问得好，这次我决定请教一下本·格雷厄姆。

It's a good question, and I decided this time to invoke the knowledge of none other than Ben Graham.

他有一个市盈率矩阵，根据AAA级公司债券的利率——也就是最高质量的公司债券——以及你对公司未来几年盈利增长的预估，得出一个你愿意为这只股票支付的市盈率。

And basically, he has a price-to-earnings matrix where you look at the interest rate on AAA-rated corporate bonds, so the highest quality corporate bonds out there, and then your estimated growth rate for the company's earnings over the next couple of years, and then it spits out a multiple that you should be willing to pay for the stock.

这是个很有用的参考方法。

And it's a helpful heuristic.

如果我们遵循这个非常简单的框架，那么像Shopify这样增长的公司，支付大约35倍的营业利润可能是一个合理的价格，不过我也知道，与市场历史上对Shopify的估值相比，这算是相当保守的了。

And if we follow this very simple framework, then something like 35 times operating earnings is probably a fair price to pay for a company growing like Shopify is, though I recognize that this is pretty conservative compared to how the market has historically valued Shopify.

所以，为了讨论方便，也许我们可以更慷慨一些，因为正如我们今天所发现的，Shopify是一家非常特别的公司。

So for the sake of argument, maybe we'd be more generous because Shopify is such a special company, as we've found out today.

即使按每股营业利润的45倍来计算，我得到的估值数字看起来也不怎么样。

And even at 45 times operating profits per share, though, the valuation number that I get just doesn't look great.

不幸的是，我们遇到了和去年一样的问题，而且在我们第一次看这只股票的时候，你猜怎么着，它的交易价格是营业利润的112倍。

Unfortunately, we have the same problem that we did last year, and at the time we first looked at the stock, it was trading at, guess what, 112 times operating profits.

这个数字简直离谱，说出来都让人觉得不可思议。

It's just an obscene number to even say.

这简直违背了价值投资者的直觉。

It violates the senses of value investors.

而今天，它的交易价格恰好就是营业利润的112倍。

And well, today, it trades at exactly 112 times operating profits.

所以就算我想编，也编不出这样的数字来。

So I couldn't have made that up myself even if I wanted to.

而且就像他们说的，Shopify根本没有通过增长来消化它的高估值。

And as they say, Shopify has not grown into its multiple at all.

股价翻倍是因为盈利差不多翻倍了，而市场从头到尾一直愿意支付同样的溢价。

The stock doubled because the earnings roughly doubled, and the market just kept paying the same premium the whole way.

所以实际上，今天买股票的人跟我们当初放弃的是同一个赌注，只是起点更高了，而且对公司未来的要求同样是完美无缺。

So really, anyone buying shares today is making the same bet that we passed on, just from a higher starting point, and the same need for perfection from the company going forward.

Shopify过去一年确实表现不错，但112倍的经营利润意味着你不仅期望它延续去年的势头，还期望它越来越好。

Shopify definitely delivered over the last year, but 112 times operating earnings means that you do not only expect it to continue the way it has in the last year, you actually expect it to get better and better.

我最近看了Palantir，那也是一家长期以离谱倍数交易的公司，但我至少能算清楚它怎么随着时间消化这个倍数。

I looked at Palantir just recently, which is another company that traded at absurd multiples for a long time, but I could actually do the math and see how it grows into the multiple over time.

但在这里，我觉得我们还得再等几年，等市场终于说："好吧，咱们得谨慎一点了"，然后真正回落到比如35倍到45倍的盈利水平。

Here, I think it will take us another couple of years and a market that finally says, "Okay, well, let's be a bit more cautious here" and actually settle down to, let's say, 35, 45 times earnings.

而且我觉得，如果本·格雷厄姆会建议35倍盈利，我们取45倍，那已经是我们能接受的上限了。

And I think if Ben Graham would suggest 35 times earnings and we go with 45, I think that's as far as we can go.

我不知道本·格雷厄姆会不会为任何公司支付35倍的盈利，但即使Shopify是个很棒的企业，让我付112倍的经营利润，我还是觉得不太舒服。

I think, I don't know if Ben Graham would have paid 35 times earnings for any company, but I don't feel too good, even though Shopify is a fantastic business, about paying 112 times operating earnings.

是的，我当然不想押注Shopify会一直完美运营下去。

Yeah, I certainly wouldn't want to bet on Shopify basically continuing to operate perfectly.

所以我也说不准。

And so I don't know.

也许我们终究没学到什么，因为结论还是老样子。

Maybe we haven't learned anything after all because we're at the same conclusion.

而且说到底，也许我们就是太抠门的价值投资者，不愿意为这家公司出高价，尽管我们现在几乎花了两整期播客来主要讨论它有多好。

And maybe at the end of the day, we're just too stingy of value investors to pay up for this company, despite us now basically devoting two full podcasts to primarily talking about how great it is.

这次我要怪价值投资之父，对，就是本·格雷厄姆。

And this time, I'm going to blame the father of value investing, yeah, Ben Graham.

他就是我的借口。

He's my cop-out.

我无法接受支付他推荐市盈率的近两倍，甚至三倍。

I can't justify paying nearly twice the multiple he would recommend, or really triple the multiple he would recommend.

所以这就是我晚上能睡得着觉的原因——不持有这家公司。

So that's how I'm going to sleep at night by not owning this company.

但这确实是我们需要密切关注的一个公司。

But it really is one that we need to be watching closely.

所以我说这家公司可能大概值每股100美元，肯定会显得我很没眼光。

And so I'm sure I'll look terrible by saying that the company is probably about fairly valued at maybe $100 a share.

他们过去一年内在内在价值上大幅提升，但录制时股价已经到每股160美元了。

So they have increased their intrinsic value dramatically in the last year, but shares are trading at $160 per share at the time of recording.

我希望某个时候能出现足够大的抛售，这样我们就能大量买入股票。

At some point, I'm hoping we'll get a big enough sell-off so we can gobble up a bunch of shares.

如果股价接近100美元，我肯定会考虑建立一个跟踪仓位。

And if it gets close to $100, I would definitely be open to opening a tracking position.

嗯，我觉得我们花了很长时间来反思我们错得有多离谱，我们低估了Shopify，这听起来像是在重蹈覆辙，但说实话，我并不是完全同意我们当初就错了这个说法。

Well, I think we spend such a long time working out how wrong we've been, and that we underestimated Shopify, and this sounds like repeating the same mistake, but if I'm being completely honest, I don't fully agree with the notion that we were wrong anyway in the first place.

这倒不是因为我不愿意承认过去的投资决策是错误。

And it's not because I don't like considering Past investment decisions are a mistake.

我们确实承认过错误。

We definitely did that.

我们有一整期节目专门讲过去一年半里我们犯下的错误。

We had an entire episode devoted simply to the mistakes that we committed in the last year and a half.

但我们不必一直追逐那些已经把最乐观情况计入价格的股票。

But we don't have to chase stocks that price in these best-case scenarios continuously.

我们知道市场是怎么运作的，而且我很有信心，总有一天我们会有机会以对我们俩来说更有吸引力的价格买入Shopify，尤其是在整个市场都在琢磨AI和智能体对电商整体以及长期会有什么影响的时候。

We know how the markets work, and at some point, I'm pretty confident we will get a chance to buy Shopify at a price that seems more attractive to the two of us, and especially at times where the whole market tries to figure out what impact AI and agents will have on e-commerce in general, and in the long run.

而且，我也不确定。

And, I don't know.

这个月Meta和Shopify还是AI的大输家，下个月它们就成了AI的大赢家。

One month Meta and Shopify are the big AI losers, and then the next month, they are the big winners of AI.

所以我很有信心，将来我们会有机会以明显更好的价格买入它们。

So I'm pretty confident we will have the chance to buy them at significantly better prices in the future.

好的。

All right.

那我们就收个尾吧。

Well, let's bring it home.

如果今天你只记住一件事，我希望是：投资中最难的问题不是找到好公司。

If you take nothing else from today, I hope you'll remember this, listeners, that the hardest problem in investing is not finding great companies.

Shopify 从来都不难被认定为一家好公司。

Shopify was never hard to identify as great.

难的是市场通常也知道这一点，而且你往往需要付出一个不容许犯错的价格。

The hard part is that the market usually knows it, too, and you usually have to pay a price that leaves no room for error.

所以在一段时间内，错误可能不会出现，价格持续上涨。

And so over some periods, the error never comes, and the price keeps working.

但我们的工作就是持续在场，保持诚实，并准备好迎接市场先生终于对公司某些行为做出负面反应的那一天。

But our job is just to keep showing up and staying honest and be ready for those days when Mr. Market finally has a bad reaction to something that the company has done.

最后，今天分享的名言来自伟大的霍华德·马克斯，他写道：“领先时代太多，和犯错是无法区分的。”

And to close things out, today's quote comes from the great Howard Marks, who wrote, "Being too far ahead of your time is indistinguishable from being wrong."

所以，在过去一年半里，我们对 Shopify 的谨慎态度可以说和犯错无法区分，而且这种情况可能还会持续，因为这家公司的执行力不亚于世界上任何一家企业。

And so for a year and a half, our caution on Shopify has been indistinguishable from being wrong, I would say, and it may stay that way because this company is executing as well as any business on the planet.

但马克斯也在提醒所有以当前价格买入的人：即使你对公司的判断是对的，但如果太早或价格不合适，那种感觉可能和犯错一模一样，而且会持续很长时间。

But Marks is also warning everyone paying today's price that being right about a company too early or at any price can feel exactly like being wrong for a long, long time.

所以我们宁愿等待，等到我们观察名单上的好公司价格更接近合理水平，也不愿意为完美付出过高的代价。

And so we would rather wait for what we think is something closer to a fair price for a great company on our watchlist than pay for perfection.

好了，我们下次再见。

And with that, we'll see you all again next time.

感谢收听TIP。

Thanks for listening to TIP.

请在您最喜欢的播客应用上关注《投资者播客》，并访问theinvestorspodcast.com获取节目笔记和教育资源。

Follow "The Investor's Podcast" on your favorite podcast app and visit theinvestorspodcast.com for show notes and educational resources.

本播客仅供信息和娱乐之用，不提供财务、投资、税务或法律建议。

This podcast is for informational and entertainment purposes only and does not provide financial, investment, tax, or legal advice.

内容不针对个人情况，未考虑您的目标、财务状况或需求。

The content is impersonal and does not consider your objectives, financial situation, or needs.

投资有风险，包括可能损失本金，过往业绩不代表未来表现。

Investing involves risk, including possible loss of principal, and past performance is not a guarantee of future results.

听众在做出任何财务决策前，应自行研究并咨询合格的专业人士。

Listeners should do their own research and consult a qualified professional before making any financial decisions.

本节目中的任何内容都不构成对任何证券或其他金融产品的买卖推荐或邀约。

Nothing on this show is a recommendation or solicitation to buy or sell any security or other financial product.

主持人、嘉宾以及The Investor's Podcast Network可能持有节目中讨论的证券头寸，并可能随时更改这些头寸，恕不另行通知。

Hosts, guests, and The Investor's Podcast Network may hold positions in securities discussed and may change those positions at any time without notice.

对任何第三方产品、服务或广告商的提及不构成背书，The Investor's Podcast Network对其所作的任何声明概不负责。

References to any third-party products, services, or advertisers do not constitute endorsements, and The Investor's Podcast Network is not responsible for any claims made by them.

版权所有：The Investor's Podcast Network。

Copyright by The Investor's Podcast Network.

保留所有权利。

All rights reserved.