苹果选定Gemini用于Siri AI，谷歌推出AI购物工具，Meta新晋实力玩家 | 2026年1月12日

欢迎各位收看《The Information》的TI TV。

Welcome everyone to The Information's TI TV.

我是阿卡什·帕斯特里卡。

My name is Akash Pastricha.

今天是1月12日，星期一。

It is Monday, January 12.

今天早上，《The Information》发布了一篇独家深度报道，探讨了AI驱动的行车记录仪公司Motive及其背后的AI支持团队。

First up this morning, The Information published an exclusive deep dive looking at AI powered dashcam maker Motive and the people assisting its AI behind the scenes.

我们将与揭露这一报道的记者们深入探讨这一内容。

We will dive into that with the reporters who broke the story.

接着，我们会讨论谷歌在周末发布的关于Gemini新AI购物工具的公告。

We'll then discuss Google's weekend announcement about new AI shopping tools for Gemini.

我们还将深入分析我们上周末发布的报道，探讨为什么Robinhood的股价在过去一年中表现远优于Coinbase。

We're also taking an in-depth look at the story that we published over the weekend over why Robinhood shares have done so much better than Coinbase over the past year.

随后，我们将与位于奥斯汀的Excelsius公司CEO对话，该公司今早宣布筹集了6500万美元，以扩大其数据中心冷却技术的规模。

We'll then turn to a conversation with the CEO of Austin based Excelsius, which announced this morning it raised $65,000,000 to scale its data center cooling technology.

我们将在节目最后邀请主编杰西卡·莱森，她对今天早上的两则重大新闻有评论。

We will end the show with our editor in chief Jessica Lesson, who has a reaction to two big news stories this morning.

Meta任命迪娜·鲍威尔·麦科密克为总裁兼副主席，以及苹果决定采用谷歌的Gemini来更新其Siri人工智能功能。

Meta's appointment of Dina Powell McCormick as president and vice chairman, and Apple's decision to tap Google's Gemini for its AI Siri update.

这将是一场精彩的节目，让我们马上进入正题。

It's going to be a fun show, and so let's get right on into things.

《信息报》发布了一篇独家深度报道，聚焦Motive公司，这是一家向卡车公司销售人工智能车载摄像头以监控司机的公司。

The Information published an exclusive deep dive into Motive, a company that sells AI powered dashcams to monitor drivers of trucking companies.

在该公司计划今年上市之前，我们的记者调查了其人工智能工具在多大程度上依赖人工核查。

Ahead of its planned IPO this year, our reporters looked into the extent to which its AI tools rely on people checking things.

现在加入我们的是这篇报道的记者：财经部门副局长科里·温伯格和财经记者迈克尔·罗登。

Joining me now are the reporters behind that story, Corey Weinberg, Deputy Bureau Chief of Finance, and Michael Rodden, our Finance Reporter.

欢迎你们两位。

Welcome to you both.

今天早上见到你们真好。

It's great to see you this morning.

嘿，卡什。

Hey, Kash.

嘿，最近怎么样？

Hey, how's it going?

科里，今天我们聊聊行车记录仪。

Corey, we're talking about dashcams today.

Motive的行车记录仪 supposed to do？

What is the dashcam at Motive supposed to do?

你们发现了什么？

What did you find?

给我们详细讲讲整个报道过程。

Walk us through all the reporting.

Motive是这个近年来迅速发展的行业的领军企业之一，它们向大型卡车和物流公司销售这些软件驱动的摄像头。

Motive is one of the leaders in this interesting industry that's really boomed over the last few years, selling these software enabled cameras to big trucking and logistics companies.

它们在一个你可能听说过上市公司Samsara的市场中竞争。

They compete in a world where you might know publicly traded company Samsara.

它们是另一家销售这种人工智能驱动的行车记录仪的公司。

They're another company that sells these AI powered dashcams.

这些大型卡车公司购买它们，以便监控司机的行为。

And, you know, they they are purchased by these large trucking companies so they can keep tabs on their on their drivers.

这样他们就能了解驾驶舱内和道路上实际发生的情况。

So they can know sort of what's actually happening inside the cab and and on the road.

Motive 拥有一款由人工智能驱动的摄像头。

And, you know, Motiv has this this camera that is powered by AI.

它使用计算机视觉模型来检测司机是否伸手拿手机或跟车过近，旨在向最终客户——卡车公司——发出重大违规行为甚至碰撞事故的警报。

It uses computer vision models to detect when, you know, a driver is reaching for their cell phone or following too close on the road, It's supposed to alert its ultimate customers, the trucking companies, of major violations that happen and even collisions.

所以它基本上是用来监控司机，确保他们安全驾驶。

So it's supposed to keep tabs on the drivers basically, make sure that they're driving well.

我猜如果发生事故，它会通知公司说：嘿，出问题了。

And I guess in the event that there's a crash of some kind, then I guess what, it alerts the company saying, Hey, there's an issue.

你必须对此做出回应。

You got to respond to it.

这就是它的目的吗？

Is that the idea?

是的。

Yeah.

公司希望掌握路上发生的任何不良事件。

Companies want to get a beat on anything bad that happens on the road.

他们需要知道这些情况，以便处理保险、应对当局，以及所有那些麻烦事。

They need to know for insurance reasons, for dealing with the authorities, all that nasty stuff.

好的。

Okay.

这家公司已经发展得相当大了，对吧？

And this company has grown pretty big, right?

我的意思是，你写到它今年计划上市。

I mean, you wrote that it's looking to go public this year.

是的。

Yeah.

它在去年九月就突破了五亿美元的年度经常性收入，除非市场彻底崩盘，否则它将成为今年首家重要的风险投资支持的上市公司。

It crossed $500,000,000 in annual recurring revenue back in September, and it'll be it's slated to be, unless the market's totally melt down, it's sort of the first major venture backed listing this year.

因此，这个故事的核心在于，这其中有多少是人工智能的功劳，又有多少是人为干预的结果。

So now the crux of the story is how much of this is AI and how much of this is actual people in the loop.

你在报道中发现，有哪些人正在帮助这个系统中的人工智能？

What did you find in your reporting about about the people that are helping the AI in this system?

Motiv公司的员工似乎扮演着相当重要的角色。

The humans at Motiv seem to play a pretty important role.

基本上，人工智能行车记录仪捕捉到的每一个视频片段。

Basically, every video snippet that the AI dashcams capture.

无论是潜在的碰撞事故，还是司机错过的停车标志，这些视频都会被发送到一支位于巴基斯坦的400人团队，他们的工作就是快速连续地审查这些视频。

So so whether it's like a potential collision or a stop sign that a driver missed, those videos come in and they get fed to a team of 400 staff in Pakistan whose job it is is to review these videos in really quick succession.

他们只有大约一分钟或半分钟的时间来判断，这是否是一次真实的闯红灯行为，或是否真的发生了碰撞。

So they've got, like, a minute or half a minute to say, this is a real stop sign violation or this is a crash.

一旦这些视频被核实，就会被发送给车队管理者。

And then once those videos are verified, they get sent to the fleet manager.

这些员工每月薪酬约为125美元，正是该公司声称能为车队管理者提供最真实地面情况的基础。

And so these these humans who are paid about the equivalent of a $125 a month central to the company's claim that we're gonna give fleet managers the most accurate representation of what's happening on the ground.

车队管理者不会对碰撞事件、司机闯红灯或跟车过近等情况一无所知。

You know, the the fleet managers aren't gonna be in the dark about when there's a crash or if they've driven past a stop sign or if they've tailgated someone too close.

因此，该公司在招股说明书中对此做了一些阐述。

So the the company talks about that a little bit in their prospectus.

他们称之为‘人在环中’。

They call it human in the loop.

但我们试图揭开这层神秘面纱，展示真实情况：这400人的团队正承受着越来越大的压力，需要提高准确性，确保不遗漏任何碰撞事件，并尽可能缩短向车队管理者发送警报的时间。

But we tried to really pull back the the curtains on this and show exactly what's happening, which is that this team of 400 people are under increasing pressure to improve their accuracy, make sure they're not missing any collisions, and really drill down the time that they're getting these alerts to fleet managers.

这就引发了一个问题：这个每天点击上千个视频的庞大人力团队，对公司有多重要？而与此同时，公司还在大力宣传其AI技术及其准确性。

So it raises this question about, you know, how important is this huge workforce of of people clicking on a thousand videos a day to the company and while the company is also making these big claims about its AI technology and the accuracy of it.

它向潜在投资者宣称，自己拥有市场上最精准的解决方案。

It tells potential investors that it has the most accurate offering on the market.

它能捕捉到15种安全事件中的99%，这使其领先于竞争对手。

It captures 99% of 15 safety events, and that puts it ahead of its competitors.

但我们说的是，如果你看看它实际的工作方式，就会发现有一个庞大的团队，大约400名数据标注员，每天要点击观看五十万段视频之类的。

But we're saying it's just like, well, if you look at how it actually works, you know, there is this really big team of, you know, 400 data annotators making sure that they're clicking on, you know, half a million videos a day or something like that.

科里，我只是想弄清楚，为什么从人工转向人工智能，以及这两种方式中哪一种实际上更准确。

Corey, I mean, I just want to get behind sort of why the transition from human to AI and which of these ways of doing this job is actually more accurate.

因为你的报道中让我惊讶的一点是，这两种系统都会出错，而且出错的程度都很高。

Because one of the things that surprised me in your story was you both reported the extent to which I mean, both of these systems can make mistakes.

哪一种最终会更好呢？

Which one ends up being better?

我的意思是，两边各自漏掉了多少次碰撞？

I mean, many missed collisions have there been on either side?

是的，这确实是一个非常有趣的实际问题，某种程度上也是哲学性的问题：通过人类来深入挖掘那些AI难以捕捉的边缘案例。

Yeah, that's one of the really interesting practical and somewhat philosophical questions, I think, all of this is is by trying to use humans to really drill down and nail the the real edge cases that are difficult for AI to capture.

也就是AI研究者常说的追求99.99999%准确率时的‘长尾’问题——是使用大量人工操作中心更有效，还是让AI模型相互检查会更好？

The really long tail that sort of AI researchers would talk about in sort of the pursuit of 100% or 99.99999% accuracy, is it better to use sort of these operation centers worth of humans, or will the the AI models checking AI models be a better way to do it?

而在这一问题上，整个行业内部存在分歧。

And that's where there's some disagreement across the industry.

我们与某些竞争对手如Samsara以及其他采用不同于Motive方式的公司进行了交流。

We talked to certain competitors like Samsara and and some others that do it differently than Motive.

你知道，Motive当然没有放弃自己的做法。

You know, Motive is certainly not running away from its approach.

由于目前正处于IPO静默期，他们没有为我们这个故事接受采访。

It's currently in the quiet period for its IPO, so they didn't talk to us for this story.

但他们的首席执行官谢蕾亚·马卡尼过去曾表示，这种方法对他们和他们的客户都有效，有助于实现尽可能高的准确率。

But their CEO, Shreya Makani, has said in the past that this is an approach that's working for them and it's working for their customers to get to as high of accuracy as possible.

那么，迈克尔，我想知道这个故事背后更大的图景是什么？

And so, Michael, I wonder what is the bigger picture here around this story?

你之前稍微提到了这一点。

You touched on it a little bit.

我的意思是，当人类参与其中时，公司却将AI宣传为AI。

I mean, idea of companies marketing AI as AI when there are humans involved.

你认为这是人工智能领域更广泛的问题吗？我的意思是，我们在机器人领域也讨论过这个问题。

Do you see this as a broader issue that the AI sector I mean, we've talked about this with the robotics sector.

我不确定我们之前在节目中是否深入讨论过计算机视觉。

I don't know if we've talked about it as much with computer vision on the show.

但从宏观角度看，你在报道中是否对此进行了更深入的调查？

But big picture, is this something that you are looking into a little more closely in your reporting?

是的。

Yeah.

现在AI确实是个大热门。

It's definitely like, you know, AI is such a big thing at the moment.

但我认为这个故事揭示了AI当前的局限性，以及它的丰富性和强大之处。

But I I think this story shows the current limitations of AI and how rich and powerful it is.

你知道的。

You know?

在很多情况下，你仍然需要人类在幕后工作。

In a lot of cases, you still need humans working behind the scenes.

在Motive公司，他们已经暗示了

In Motive, they've hinted

在

at

他们的招股说明书中，人类似乎在支撑其AI主张方面起着核心作用。

that in their prospectus, and the humans seem pretty central to stacking up their AI claims.

但我们已经看到很多案例，一些公司试图对客户耍花招。

But we've seen a lot of cases where companies have tried to pull a bit of a hoodwink over their customers.

比如去年，我认为有一家叫Builder AI的公司倒闭了，他们声称拥有非常强大的AI技术。

So it's like there was a Builder AI that collapsed last year, I think, who said that they had this, like, really powerful AI technology.

但实际上，背后却是成千上万的人，主要来自发展中国家，正在不停地敲键盘工作。

But in reality, it was just, you know, thousands of sort of humans behind the scenes in, you know, developing countries tapping away.

还有亚马逊的无人超市，号称完全由AI驱动，但其实背后有一群人跟踪顾客把什么东西放进了购物车。

You know, there there was the Amazon supermarket where people could walk through that was supposedly all AI, but there was a bunch of people behind the scenes tracking what the customers put in their shelves.

就连Waymo，我本人很欣赏它，

You know, even Waymo, which which I, you know, I take and I I really like.

但偶尔也需要人类通过远程机器人进行干预。

You know, every now and then, a human will have to intervene through a a telerobet robot.

他们必须打电话并亲自驾驶汽车。

They have to phone in and drive the car themselves.

所以我认为，投资者应该对公司的主张保持谨慎，深入探究那些声称拥有强大AI工具的公司究竟是如何运作的。

So I think that it's a good sign for investors to just be really careful about what claims companies are making and to really try and dig into how the companies that say they have these really powerful AI tools, how they actually work.

科里，最后一个问题是给你。

Corey, last question for you.

由于静默期，你没有机会与该公司交谈。

You didn't have a chance to speak with the company because of the quiet period.

我想知道，如果你能见到他们的首席执行官并面对面交流，你会想向他们问哪些问题？

I wonder if you did have the CEO, and you had him in front of you, what questions would you want to know the answers to from them?

我想了解他们计划如何降低AI所捕获内容的人工审核率的具体细节。

I want to know more specific detail on their plan to lower the rate of human review on what the AIs capture.

他们公开表示，希望随着时间推移减少这一比例。

They've said publicly that they they aim to reduce it over time.

在我们的报道中，我们无法确定他们是否已经制定计划，或者具体计划是什么。

In our reporting, we weren't able to kind of determine whether they have or or what their exact plans are.

我对他们必须在支付人力以确保100%准确性与推动构建无缝运行的技术之间做出的权衡很感兴趣。

And I'm interested in some of this trade off that they have to make around, sort of paying for humans to make sure that there's 100% accuracy versus the drive to actually build technology that works seamlessly.

我认为这是硅谷所有首席执行官都在面临的难题。

I think that's something that CEOs across Silicon Valley are grappling with.

很好。

Great.

好了，Corey和Michael，感谢你们的到来。

Well, Corey and Michael, I want to thank you for coming on.

这位是来自《The Information》新闻部的Corey Weinberg和Michael Rodden。

That is Corey Weinberg and Michael Rodden from our newsroom here at The Information.

我们很快再见。

See you guys very soon.

好的。

Okay.

谷歌正在为Gemini添加AI购物工具。

Google is adding AI shopping tools for Gemini.

该公司上周末宣布了一系列AI购物功能，包括通过Gemini购物的功能。

The company announced a slate of AI commerce features over the weekend, including the ability to buy stuff with Gemini.

AI与购物正日益成为竞争激烈的战场，因此我想邀请我们的高级编辑梅雷迪思·马齐利来为大家详细分析。

AI and shopping is, of course, becoming an increasingly competitive battleground, and so I want to bring on our senior editor Meredith Mazzili to break it all down.

梅雷迪思，欢迎再次做客我们的节目。

Meredith, welcome back to the show.

很高兴你来到这里。

It's great to have you here.

嘿，科什。

Hey, Kosh.

谢谢邀请我。

Thanks for having me.

这是与梅雷迪思一起的周五周一特别节目。

It's the Friday Monday special with Meredith.

总是关于购物。

Always with shopping.

总是和购物有关。

Always with shopping.

好吧，我想听听你对这些工具的初步看法，因为我们上周五刚聊过OpenAI如何进入这种代理式电商领域。

Well, okay, so I want to get your initial reactions here to the tools because we were just talking on Friday about how OpenAI has dived into this agentic commerce space, I guess.

我们之前谈到了亚马逊，现在谷歌也加入了。

We were talking about Amazon, and here we have Google coming up.

你觉得怎么样？

And, you know, what did you think?

是的。

Yeah.

我的意思是，这反映了谷歌如何应对AI对其核心搜索业务的威胁，可能和OpenAI在聊天中推出结账功能的时间点关系不大。

So, I mean, I think this says a lot about how Google is responding to the threat of AI to their core search business and maybe has a little less to do with, you know, the timing of OpenAI's own in chat checkout rollout and things like that.

我会详细分析这些，但首先，我认为有必要深入了解一下谷歌周日具体发布了什么。

So I will unpack all of that, but let's first, I think, would be helpful to dig into what exactly it is that Google announced, on Sunday.

第一，是在Gemini和Google AI模式中内置的AI结账功能，正如你所说，这让你可以直接在谷歌内购买商品，而无需跳转到其他网站。

So number one, AI checkout inside Gemini, inside Google AI mode, which basically, like you said, lets you just buy stuff there, without leaving Google, without going to other sites.

你开始结账了。

You're starting to check out there.

第二点，他们还宣布了另一种非常有趣的内容，即会出现在该AI界面中的广告类型。

And number two, another thing they announced that was super interesting was types of ads that would go inside that AI interface.

这可能表现为零售商付费投放的促销活动或特别优惠。

So that might look like promotions or special offers that the retailers basically pay to place inside that AI interface.

这对谷歌来说是一个巨大的转变。

So that's a big shift for Google.

它不再是从谷歌将你引导至其他网站、靠点击付费，而对之后发生的事情漠不关心。

So it's moving from Google sending you out to other sites, basically getting paid for those clicks to send to other sites and not really caring what happens after that.

现在，他们试图将更多购物行为纳入AI内部，并掌控整个流程的更多环节。

Now what they're trying to do is bring more of that shopping inside AI, and controlling more of that process.

在我看来，这清楚地表明，核心的AI搜索业务，或者说核心的传统搜索业务，正面临压力。

And, you know, to me that reads really just as, okay, the core AI search business or the core regular search business rather is under pressure.

没错。

Right.

对。

Right.

你知道吗，我们在九月份看到OpenAI也宣布了类似的功能，即在ChatGPT内完成结账。

You know, we did see OpenAI in September announced something very similar checking out inside of ChatGPT.

但我觉得，对于OpenAI和谷歌来说，他们只是各自独立地走到了同一个地方。

But, I think, you know, for OpenAI and Google, they're just kind of arriving at the same place independently.

OpenAI正在从零开始构建盈利模式，而谷歌则试图保护其现有的盈利方式，或寻找新的赚钱途径。

OpenAI, they are trying to build monetization from scratch, whereas Google is trying to defend their existing monetization or come up with new ways to make money.

所以我想问你，你认为谷歌在这里相对于OpenAI有多大优势？因为这其中一个关键部分是，他们将利用谷歌支付，而这一系统已经存在一段时间了。

So I want to ask you about how much of an advantage you think Google has here over OpenAI, because one of the key parts of this is that they're going be leveraging Google Pay, which is something that's been in place for a while.

我们曾邀请我们的电子商务记者安·吉恩上节目，她谈到过支付系统有多难搞，不仅在AI领域如此，在任何领域都是如此。

We've had Anne Guine, our e commerce reporter on the show here talking about how difficult payments is as a thing to figure out, not just with AI, but at all.

因此，谷歌似乎已经解决了支付问题，也许他们已经构建好了这套系统。

And so Google seems to have payments figured out and maybe they've built that.

这里另一个难点在于，从零售商那里获取数据很困难。

The other part here is it's tough to get data from retailers.

所以也许他们在这方面并没有优势。

So maybe they don't have an advantage there.

我的意思是，你觉得谷歌在这里有优势吗？

Mean, do you think Google has a leg up here?

所以我认为，谷歌拥有自己的谷歌支付和数字钱包，这绝对是他们的优势。

So I think the fact that Google does have its own Google Pay and its own digital wallets is definitely advantage for them.

正如我们之前所写的，支付和产品数据存在许多非常棘手的问题，这些因素使得基础层面的AI购物变得极其困难。

As we've written about before, there are just lots of really tricky payments and product data, things that just make AI shopping at a very basic level, very difficult.

因为他们之前曾涉足不同的购物尝试，并且与商家在传统广告方面合作密切，因此他们在这一领域拥有更多经验。

Because they've dabbled in different shopping efforts before they work a lot with merchants on traditional ads, you know, they do have more experience in that space.

我认为，所有这些AI工具最终都将取决于商家的采用程度。

I think what it's gonna keep coming down to with all of these AI tools is merchant adoption.

我认为这目前是最大的未知问题。

And I think that's the big open question right now.

商家是否愿意让谷歌在购买过程中接管更多控制权，脱离商家自身？

Will merchants want Google to be taking more control over the whole purchase process away from the merchants.

因此，越来越多地将商家自己的网站和品牌排除在外。

So kind of cutting out the merchant's own website, their own branding more and more out of the equation.

这就是悬而未决的问题。

And that's the open question.

对于所有这些AI购物工具来说，情况都是一样的。

And that's the same for all of these AI shopping tools.

我认为，对于那些产品差异化较小的商家来说，他们会更倾向于采用这种方式。

I think with merchants that maybe have less differentiated of products, you're going to see more adoption on that side.

我认为这解释了为什么一些大型零售商正在积极加入。

I think that explains why you see some big box retailers really jumping in.

但当涉及到夫妻店或那些直接面向消费者、希望讲述品牌故事的小品牌时，他们的考量就有所不同了。

But when it comes to kind of mom and pop shops, small brands that maybe go direct to consumer and really want to tell that brand story, I think the calculus is a little different for them.

所以，你不认为这在跟上时代方面对所有零售商来说都是显而易见的吧？

So you don't think this is I mean, this is not really a slam dunk for all retailers insofar as keeping up with the times.

你的意思是，对于大型零售商来说，当然如此，因为他们必须采取全渠道策略，无处不在。

You're saying for the big retailers, sure, because they have to have this sort of omnichannel approach and be everything everywhere.

但对于小型零售商来说，这相当于背离了几年前许多零售商热衷于建立的DTC关系。

But for the small retailers, I mean, it's sort of a walk away from that DTC relationship that we saw so many retailers getting excited about building a couple of years ago.

没错，正是如此。

Yeah, exactly.

我认为这个方程式中的每个人都试图弄清楚，这有点像是先有鸡还是先有蛋的问题。

And I think everybody in this equation is trying to figure out it's a bit of a chicken and egg situation.

消费者真的愿意这样做吗？

Will shoppers actually want to do this?

商家是否会以足够多的方式参与进来，让 shoppers 有足够多的东西可以选购？

Will merchants participate in such a way that there's enough stuff out there for shoppers to want to do this?

AI公司、支付公司等各方能否理清这一切？

And can the AI companies, payment companies, everybody figure it out?

但就商家的采用而言，是的。

But for the merchant adoption specifically, yes.

这仍然是一个巨大的未解之谜。

Like, that's still kind of the big open question.

我们实际上有一名记者本周在现场报道纽约的NRF展会，昨天谷歌正是在那里发布了所有这些产品。

We actually have one of our reporters on the ground at NRF in New York this week, which is where, Google did announce all these products yesterday.

那是全国性的展会吗？是全国零售展吗？

That's the national was it is it national retail?

是的。

Yeah.

是的。

Yeah.

一些零售相关的展会。

Something retail.

这是纽约最大的零售行业大会。

It's the big retail conference in in in New York.

没错。

Exactly.

因此，对于这些较小的DTC品牌来说，它们最终会如何定位，仍然是一个开放的问题。

So those smaller DTC brands, think is really a bit of an open question to see where they land on this.

因为，你知道，反过来讲，如果你选择观望，而消费者行为已经转向了那里，那就很难再追赶上了。

Because, you know, on the flip side, if you sit it out and that's where shopper behavior goes, it's very difficult to catch up potentially.

而且，我想最后值得一提的是，我们的联合执行主编马丁·佩尔昨晚在简报中谈到了谷歌过去试图成为电子商务或购物平台的努力。

And and, you know, the the the last thing that I think is worth talking about is is Martin Pierce, our co executive editor, wrote last night in the briefing about Google's efforts to become more of an e commerce or shopping player in the past.

我想说的是，确实存在一个现实：尽管谷歌已经建立了这么多工具，比如谷歌购物，还有谷歌航班，你知道的。

And I guess, I mean, there is a reality that, yeah, even though they have all these tools established in place, I mean, Google Shopping, I guess, I mean, we do you know, there's Google Flights.

我在想，谷歌还有什么……

There's, you know, I'm trying to think Google.

我会通过谷歌地图订酒店，但我目前还不会通过谷歌购物。

I do hotels through Google Maps, but I'm not really shopping through Google just yet.

所以看起来，谷歌在这一领域过去确实遇到过一些困难。

So it seems to have had some trouble in that arena before.

没错。

Right.

但我觉得，过去这些赌注未必有现在这么高，这让我们回想起本段开头讨论的内容，即核心搜索业务可能正面临根本性威胁——因为人们购物和寻找产品的方式正在发生变化。

But though, I think in the past, the stakes weren't necessarily as high, kind of circling to what we were back talking about at the beginning of this segment, you know, with this idea that the core search business could be under more of a fundamental threat from just people shopping and thinking about looking at products in different ways.

你知道，这次他们必须更加努力才能做对。

You know, I think there's just more pressure there to get it right this time.

财报季即将来临，所以我想我们会得到更多这方面的答案。

Well, earnings season is coming up, so I guess we'll get more answers to that effect.

梅雷迪思，感谢你参加我们的节目。

Meredith, I want to thank you for coming on the show.

这位是梅雷迪思·马齐利，我们《信息》杂志的高级编辑。

That is Meredith Mazzili, our senior editor here at The Information.

好的。

All right.

谢谢，阿卡什。

Thanks Akash.

好的。

Okay.

Robinhood 和 Coinbase 之间的竞争正日益加剧，因为双方都在争夺相似的交易工具。

Robinhood versus Coinbase is a rivalry that increasingly has been heating up as they both compete on similar trading tools.

但去年，Robinhood的股价几乎翻了三倍，而Coinbase的股价却下跌了12%。

But last year, while Robinhood stock nearly tripled, Coinbase shares fell 12%.

我的同事杨悦驰上周末写了一篇深度报道，分析了这一差距，我想请她来谈谈这个话题。

My colleague Yuechi Yang wrote a deep dive looking at that delta over the weekend, and I want to bring her on to talk all about it.

Yuechi，欢迎再次做客节目。

Yuechi, welcome back to the show.

很高兴你来到这里。

It's great to have you here.

嘿，Akash。

Hey, Akash.

这是本赛季最激动人心的对决——Robin对非Robin，Robinhood对Coinbase。

It's the most exciting rivalry of the season, Robin versus not Robin, Robinhood versus Coinbase.

为什么Robinhood的股价表现远优于Coinbase？

Why has Robinhood stock done so much better than Coinbase?

这简直太疯狂了。

This is not even like I mean, is crazy.

我之前不知道这些数字，直到你告诉我。

I didn't know the numbers until you did it.

下跌了12%，而Robinhood则上涨了一百多个百分点。

Down 12%, 100 and something percent for Robinhood.

到底发生了什么？

What's going on?

这是一场酝酿多年的竞争，但去年我们开始看到Coinbase和Robinhood的股价出现分化。

So this is a rivalry that's years in the making, but really last year we started to see the stocks of Coinbase and Robinhood to diverge.

投资者更喜欢Robinhood的股票，因为它是一个更多元化的投资选择。

And investors like Robinhood stock because it's a more diversified bet.

你知道，Robinhood提供股票交易。

As you know, Robinhood offers stocks trading.

他们也提供加密货币交易。

They offer crypto trading.

他们还有一个庞大的期权业务。

They have a big options business.

加密货币仅占罗宾汉总收入的20%，而Coinbase的所有收入都来自加密行业。

And crypto is only 20% of Robinhood's total revenue, whereas Coinbase generates all of its revenue from the crypto industry.

我还注意到，罗宾汉去年充分利用了特朗普政府下的监管环境，迅速行动并 aggressively 推出新产品。

And I also noticed that Robinhood last year really took advantage of the regulatory environment under Trump by moving very quickly and launching new products aggressively.

一个典型的例子就是预测市场。

And one case in point is prediction markets.

罗宾汉于2024年底推出了预测市场，最初提供对2024年美国总统大选的投注，并迅速扩大规模，增加了更多投注选项。

Robin Hoo launched prediction market in late twenty twenty four, initially offering the bet on the twenty twenty four US presidential election, and it quickly doubled down by adding more bets.

罗宾汉表示，预测市场自推出以来已成为其历史上收入增长最快的一项业务。

And Robin Hoo said that prediction market has since become its fastest growing business line by revenue in its entire history.

所以这里一个有趣的地方是，长期以来，当我几年前报道加密货币时，Coinbase的股价似乎一直与比特币的表现同步。

So one of the interesting parts here is for the longest time, when I covered crypto a couple years ago, Coinbase shares sort of seemed to have mirrored Bitcoin's performance.

我觉得Coinbase一直在努力通过其订阅服务摆脱这种关联，使收入和盈利能力变得不那么波动。

And I feel like Coinbase has tried to get away from that a little bit with its subscription offerings and make its revenue and its profitability a little bit less volatile.

但我感觉Coinbase仍然没能真正摆脱与比特币价格的紧密关联，对吧？

I feel like Coinbase still hasn't really been able to get away from that tying to Bitcoin prices at large, right?

是的。

Yes.

去年比特币全年下跌。

And Bitcoin was down on the year last.

对。

Right.

我觉得你说得对。

I think you're right.

Coinbase的所有收入都来自加密行业。

Coinbase generates all of its revenue from the crypto industry.

我认为它在发展订阅业务和服务方面做得不错，这些业务比单纯的交易手续费更稳定。

I would say that it has done a good job building up a bigger portion of revenue from subscription businesses and services, which are more sticky than just trading fees.

Coinbase现在也开始进入股票市场和预测市场。

And Coinbase now is entering stock markets stock offerings as well as prediction markets.

因此，我们明年需要重点关注的一点是，Coinbase在非加密产品上能取得多大的进展。

So one big thing we'll be watch out for next year is to what extent Coinbase can gain traction with their non crypto products.

那我问你一个问题。

So let me ask you this.

你认为这更多是Robinhood表现卓越的故事吗？

Do you think that this is more a story of Robinhood really excelling?

还是Coinbase落后了？

Or is it Coinbase falling behind?

我认为它们正在正面竞争，我们看到了

I think they're competing head on, we saw

我认为是Robinhood表现卓越。

I think it's Robinhood excelling.

我直接说吧。

I'm be going upfront with it.

我的意思是，很难否认一家公司在一年内实现180%以上的增长不是在卓越表现。

Know, I mean, hard to argue that 180 something percent growth over the course of a year is not a company excelling.

但你知道，我觉得它的创新更快，产品也更多。

But, you know, I feel like it's a little faster to innovate, there's more products.

我的意思是，他们也有信用卡。

I mean, they have the credit cards too.

他们确实吸引了更多元化的交易者群体，对吧？

They're really getting a much more diverse pool of traders, right?

是的。

Yeah.

但我要说，Coinbase和Robinhood的用户群体有很大重叠，因为它们都瞄准了热爱零售投机和交易的年轻用户，且主要位于美国。

But I will say that Coinbase and Robinhood actually have a lot of overlap with your users because they're both going after young users who love retail users who love speculations and trading, who are largely US based.

所以这场竞争只会愈发激烈。

So this is a competition that will only intensify.

那么，他们的盈利方式完全一样吗？

Now, do they make money the exact same way?

因为我知道，Robinhood的订单流支付是其让交易者几乎无需付费的方式，对吧？

Because I know that Robinhood payment for order flow was that way of Robinhood keeping charges basically to nothing, right, for traders.

他们并不是从交易者身上赚钱的。

They weren't making money off their traders.

他们更多是从他们将交易路由给的做市商那里获利。

It was more off the, I guess, the market makers that they were routing the trades to.

我认为，Coinbase 仍然会对他们的交易收取交易费。

Coinbase, I think, still takes a transaction fee off their trade.

所以这里的商业模式是完全不同的，还是仍然相当相似？

So are the business models entirely different here or are they quite similar still?

Coinbase 的加密货币交易仍然是其最大的收入来源，而且 Coinbase 对零售用户提供的加密货币交易收取高额费用。

Coinbase crypto trading is still their biggest revenue generator, and Coinbase does charge a high fee, especially from retail users for offering crypto tradings.

他们多年来一直维持这一费用，仅仅因为他们在美国内部非常受欢迎且占据主导地位。

And they have been able to maintain that fee for years just because they're so popular and so dominant in The U.

S，他们几乎没有面临降低费用的价格压力。

S, and they haven't really faced too much pricing pressure to reduce their fee.

对于 Coinbase 的新业务——股票产品，Coinbase 也将提供零手续费交易。

For the stock product, which is a new product at Coinbase, Coinbase will also offer zero fee trading.

因此，在股票产品方面，他们的收费模式与 Robinhood 的股票产品商业模式相同。

So in that sense, on the stock product that they're charging is the same business model as Robinhood's stock product.

回到你一开始提到的内容，预测市场似乎是这两家公司都在密切关注的新趋势。

And going back to what you were talking about at the very start, prediction market seems to be the new in thing that both of these companies are exploring closely.

还有这种说法，你知道，Robinhood 正在尝试，尽其所能地让投资者接触到私营公司。

There's also this notion, you know, Robinhood is getting into, I guess, giving exposure to private companies to whatever extent it can.

你觉得今年这个故事会如何发展？

How do you think this story plays out this year?

预测季现在已经结束了。

And prediction season is over now.

我们现在处于一月份，但你认为这两家公司在业绩上会变得更加紧密相关吗？

We're in the January, but do you see the companies being a little bit more closely tied together in terms of performance?

你认为预测市场会彻底改变整个格局吗？

Do you see prediction markets changing the landscape altogether?

也许，我不知道，Polymarket 会出现，但我们真的不知道人们最终会去哪里。

And maybe, I don't know, maybe polymarket comes and you know, we don't know how where people will go, really.

我认为预测市场在2026年之前至关重要。

I think prediction market is important going into 2026.

有一些重大事件是人们关注的焦点。

There are some big events that people are watching out for.

今年将举办世界杯，许多人对在预测市场上投注这场比赛感到兴奋。

The World Cup is going to happen this year, and that's something that a lot of people are excited about betting on prediction markets for.

美国中期选举也将举行。

The US midterm elections will also happen.

预测市场非常依赖这些重大新闻、文化事件和体育赛事，今年有几场这样的事件已经安排好了。

So prediction market is really rely on these big news event and cultural events and sports event, and there are several that are scheduled to occur this year.

对于Coinbase来说，我们将看到其股票交易产品表现如何。

And for Coinbase, so we'll see how their stock trading product perform.

许多用户已经在其他地方拥有股票投资功能，因此我们将看到他们在这一领域能取得多大进展。

A lot of users already have their stock investing capability elsewhere, so we'll see to what extent they're able to gain traction there.

很好。

Great.

我昨晚刚看了金球奖，发现他们也与Polymarket达成了合作，我想是这样的。

I was just watching the Golden Globes last night and I saw that they too had their polymarket partnership, I guess.

他们本来就在预测谁会赢得下一个奖项。

They were predicting who was going to win the next award anyway.

一切都发展得非常迅速，非常感谢你来帮助我们理解这一切。

It's all moving very quickly, and I want to thank you for coming on to help us make sense of it.

这位是《The Information》的加密货币记者杨月驰。

That is Yuechi Yang, our crypto reporter here at The Information.

好的。

Okay.

一家新的数据中心冷却公司正在筹集新一轮融资。

A new data center cooling company is raising a new funding round.

专注于液冷业务的Excelsius公司正在筹集6500万美元。《The Information》在我们的AI基础设施通讯中报道了这一消息，其中还包含对该业务的深入分析。

Excelsius, which specializes in the business of liquid cooling, is raising $65,000,000 The information wrote about that news in our AI infrastructure newsletter, where we get into some analysis around the business as well.

我想邀请Excelsius的首席执行官乔什·克拉曼来帮我们解读这一重大新闻。

I want to bring on Josh Klaman, CEO of Excelsius, to help us unpack the big news.

乔什，欢迎来到节目。

Josh, welcome to the show.

很高兴你能来。

It's great to have you here.

能来这里很好。

It's good to be here.

谢谢。

Thanks.

你今天有一些大新闻。

So you had some big news today.

给我们讲讲你们公司是做什么的。

Tell us about what your company does.

我们制造两相定向芯片液冷系统。

So we make two phase directed chip liquid cooling.

你知道，随着芯片越来越热，数据中心的液冷现在是一个巨大的挑战。

You know, liquid cooling in the data center is a huge challenge now as chips get hotter, the water chip chips continues to increase.

在机架或机架内的服务器中，功率密度以及因此产生的残余热量持续增加。

And within a rack or a server in a rack, the density of power and therefore heat, residual heat, continues to increase.

这对数据中心来说是一个巨大的管理挑战。

This is a huge challenge for data centers to manage this.

几年前，曾出现过一种关于冷却新技术的竞赛，那就是液冷。

You know, a few years ago, there was a kind of a competition of new architectures for cooling, that is liquid cooling.

共识认为，未来几年液冷将成为数据中心市场的必需技术。

The consensus was building the liquid cooling to be required for the data center market over the next few years.

其中一种是浸没式冷却，当我们公司成立时，我们认为这很有趣，但可能不适合大规模应用。

And one was immersion, you know, and when we founded our company, we thought that was interesting but probably an impractical architecture for broad adoption.

浸没式冷却是他们过去使用的方式。

Immersion was the way that they used to.

那到底是什么意思？

What does that even mean?

浸没式冷却意味着将服务器设备直接放入非导电液体中。

Immersion means you actually put the server equipment in a pool of nonconductive fluid.

这就像是一个非导电液体的浴缸。

So, it's kind of a bathtub of nonconductive fluid.

所以整个服务器都放进去。

So, the whole server goes in.

三四年前，这曾引起很多炒作。

This had a lot of hype, you know, three or four years ago.

我认为它吸引了普通公众的注意或想象力。

I think it captured sort of the attention or the imagination of the lay public.

你知道，这看起来有点像科幻小说。

You know, it seemed like sort of science fiction y.

是的。

Yeah.

所以那已经被淘汰了。

So that was out

他们把它放在一个池子里。

They put it in a pool.

是的。

Yeah.

是的。

Yeah.

然后通过水泵将水输送到安装在芯片顶部的散热板上。

And then water, simply pumping water over plates that are installed on top of the chips.

基本上，那里有几个变量。

And basically, you have a couple variables there.

你可以加快水流速度来应对更高的热量，或者使用更冷的水，但这是有限制的。

You can pump water faster if it's higher heat or you can pump colder water, but it's limited.

我们看到，随着下一代芯片以及再下一代芯片的出现，热量密度将高到水通过对流或所谓的显热冷却方式难以有效散热。

And we see with the next generation of chips and the one after that, the concentration of heat is going to be such that water is going to struggle through convection or sensible, what it's called sensible cooling.

它将难以带走这些热量。

It's going to struggle to carry that heat off.

因此，Excelsius 成立了一家公司，旨在为未来的数据中心开发一种专门适用的冷却架构。

So, Excelsius founded a company to develop a really fit for purpose cooling architecture for data centers going forward.

它基于——抱歉，你先说。

It's based on- sorry, go ahead.

它是怎么工作的？

How does it work?

是的，给我们讲讲。

Yeah, tell us.

是的，它基于两相物理原理，意味着你泵送一种非导电液体经过芯片顶部的板子，并控制其发生相变，从液态变为气态。

Yeah, it's based on two phase physics, which means you pump a fluid, a non conductive fluid, over a plate on top of the chip and you orchestrate a phase change, so from fluid to vapor.

好的。

Okay.

如果你回想一下你的高中物理课

And if you think about your high school physics class

我现在正在想呢，是的。

I'm thinking about it right now, yeah.

好吧，就是这样。

Okay, there you go.

通过这种相变，它能吸收大量的热能，远超单纯的对流冷却或不发生相变时带走热量的能力。

It captures a huge amount of heat energy by creating that phase change, much more than just convection cooling or carrying that heat off without that phase change could do.

这有很多优势。

There are a lot of advantages to this.

第一，这是一种非常高效的冷却方式。

One, it's a very efficient cooling.

因此，我们可以将这一点与过去几天詹森引发的市场震荡联系起来，他说：

So, we can tie this into some of the conversation over the last few days that Jensen created this of shake in the market because he said

对，没错。

Right, yeah.

因为，如果我没理解错的话，他说新的鲁宾芯片系列不需要那么多制冷设备，对吧？

Because he, I mean, correct me if I'm wrong, he said, he basically said that the new Rubin chip family was not going to need as much of the chilling equipment, I think, right?

他就是这么说的，这震动了全球大型制冷设备制造商，如江森自控、英格维特等。

That's what said, and it shook up sort of the big chiller manufacturers in the world, JCI, Invertev, etcetera.

我们认为这并不正确。

We don't think that's true.

我们认为，数据中心的冷却将是一个挑战，你需要高效地排出这些热量。

We think that data center cooling is going to be a challenge and you need to reject that heat efficiently.

对冷却器的依赖将继续减少，因为你可以将数据中心中所谓的主水循环加热。

The reliance on chillers will continue to decrease because you will be able to warm that primary, what's called the primary water loop in the data center up.

有趣的是，每升高一摄氏度，你就无需冷却那部分水。

The interesting thing is that for every one degree C, you don't have to cool that water.

每年可以节省约4%的能源，这对于大规模的数据中心来说意义重大。

You save about 4% energy a year, which obviously for a scaled out data center is incredibly material.

所以我想弄清楚这一点。

So I just want to understand this.

你是在暗示，詹森所说的设备不会那么热、我们不需要那么多冷却设备的观点，其实并不成立，它们仍然会产生大量热量？

Are you suggesting that basically what Jensen was saying about things not getting as hot and us not needing as much cooling equipment, Is it your view that that's not going to be the case, and that they're still going to produce as much heat?

这仍然是个问题吗？

This is going to still be a problem?

好吧，让我这么说吧。

Well, think that the here's what I'll say.

我对基于水的解决方案（比如直接用水冷却芯片）能否实现这种效率并不那么有信心。

I am not as confident that water based solutions, so just pumping water over chips, is going to create that efficiency.