本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
欢迎各位来到《The Information》的TITB节目。
Welcome everyone to The Information's TITB.
我是阿卡什·佩里特沙。
My name is Akash Pesritsha.
今天是1月7日，星期三。
It is Wednesday, January 7.
我们今天有一期内容丰富的节目。
We have got a busy show ahead of us.
首先，《The Information》独家报道，本周中国要求国内科技公司暂停订购英伟达的H200芯片。
First up, The Information has exclusive reporting that China this week asked tech companies there to halt orders of NVIDIA's H200 chip.
我们将为您带来所有细节。
We will bring you all the details.
接着，我们将采访风险投资公司Conviction的创始人莎拉·郭。
Then we are talking to Sarah Guo, founder of the venture firm Conviction.
莎拉本周在CES上为黄仁勋开场，并发表了一篇关于企业如何在人工智能热潮中保持竞争力的深刻文章。
Sarah opened for Jensen Huang at CES this week and published a sharp essay on how companies can stay competitive in the AI boom.
今天是星期三，这意味着我们将深入探讨本周的财经简报，邀请我们的财经编辑一起讨论，随后我们会谈到彭博社的一则报道：Discord已秘密提交了IPO申请，最后我们将与OutSystems展开对话，这是一家为开发者提供AI平台的公司，我们将探讨在AI时代如何衡量投资回报率，以及在争夺最佳编码模型的激烈竞争中如何应对。
And it is Wednesday, which means it is time to dive deep into this week's finance newsletter with our finance editor, and we are also then going to talk about a Bloomberg report that Discord has filed confidentially for an IPO, and we will wrap with a conversation with OutSystems, an AI platform for developers to talk about ROI in the age of AI and the intensifying race for the best coding models.
事情很多，我们马上开始。
Got a lot going on, so let's get right on into things.
《The Information》亚洲团队今天早上发布了独家报道，称中国本周已要求科技公司暂停订购NVIDIA的H200芯片。
The Information's Asia Bureau published exclusive reporting this morning that China this week told tech companies to halt orders for NVIDIA's H200 chip.
现在邀请我们的NVIDIA记者韦恩·马与我一起讨论这一消息。
Joining me now to discuss the news is our NVIDIA reporter, Wayne Ma.
韦恩，欢迎再次做客我们的节目。
Wayne, welcome back to the show.
很高兴你能来。
It's great to have you here.
嗨，谢谢邀请我，阿卡什。
Hey, thanks for having me, Akash.
这是我们亚洲团队发布的一条重磅新闻。
So this was a big story that our Asia bureau published.
关于中国的监管政策、政府的表态以及局势的发展，我们了解到了什么？
What did we learn about the regulations of China, what the government is saying, and how the situation is evolving?
我认为关键在于，这只是一个临时的暂停。
Well, I think the key thing here is that this is like a temporary halt.
对吧？
Right?
之所以是临时暂停而非永久停止，是因为中国公司仍然需要英伟达的芯片来训练他们的模型。
And the reason why it's temporary and not permanent halt is that Chinese companies still need NVIDIA chips to train their models.
中国希望成为AI领域领导者的雄心，我认为仍然需要英伟达的芯片，至少因为中国本土的替代产品在这一领域还不够高效或强大。
And China's ambitions to become a leader in AI re still requires, I think, NVIDIA chips, if only because the Chinese equivalent still aren't as efficient or as capable in this sector.
所以‘临时’的意思，我理解为，我们正在理清情况，目前先别下订单。
Temp so temporary meaning the way I read it was, I mean, it's basically while we figure things out, hold off on placing your orders right now.
这就是我的理解。
That that was my read on it.
没错。
Yeah.
没错。
That's right.
我想关键问题是，这会持续多久呢？
And I guess the big question is is, like, how long is that going to be?
对吧？
Right?
会是一个月吗？
Is it gonna be a month?
会是六个月吗？
Is it gonna be six months?
会是一年吗？
Is it gonna be a year?
我觉得不会这么久。
I don't think it'll be that long.
但与此同时，你知道，这会对英伟达的收益产生实质性影响，具体取决于这种情况持续多久。
But at the same time, you know, it's gonna have a material impact on NVIDIA's earnings, know, depending on how long this is.
好吧，我马上会谈到这个机会的规模，但我觉得这有点令人惊讶，因为就在本周早些时候，我们刚听到詹森对记者说，对H200的需求非常高，而且从他的角度来看，似乎是全员投入。
Well, I want to get to the size of the opportunity in a second here, but this strikes me as a bit surprising because we just heard Jensen talk to reporters earlier this week talking about how demand is very high for h two hundred and, you know, from his end, seemed like it was all hands on deck.
是的。
Yeah.
我认为出现暂时停顿的原因是中国政府也不希望过度依赖英伟达。
I mean, I think the reason why there's a temporary halt is that the Chinese government also doesn't want to rely too much on NVIDIA.
对吧？
Right?
他们不希望过度依赖西方技术。
They don't want to rely too much on Western technology.
因此，他们一直在努力发展本土的AI芯片公司，以与英伟达竞争。
And so they've been kind of trying to develop their homegrown AI chip companies to try to compete with NVIDIA.
我认为，如果他们允许中国公司随意购买尽可能多的H200，就会失去推动本土技术发展的动力。
And I think there's a worry that if they just let Chinese companies buy, you know, as many h two hundreds as possible, there's no incentive to foster that local homegrown kind of technology.
中国公司如果一直落后，就永远无法与之竞争。
Chennai Wa was just lagged behind and never be able to compete.
你能再多谈谈这一点吗？
Can you talk a little bit more about that?
我的意思是，中国客户对这些芯片的需求有多大？
I mean, how how much do customers need these chips in China?
哦，他们确实需要这些芯片。
Oh, they definitely need them.
他们非常急需。
They need them badly.
目前，中国的AI芯片产业还远未达到英伟达的水平。
Like, right now, Chinese China's chip AI chip industry is nowhere near at the level of NVIDIA's.
我认为，即使他们能赶上，也需要很长时间。
And so it's gonna take a long time, I think, for them to catch up if if they ever will.
你提到了英伟达的收益。
And you talked about earnings from NVIDIA's end.
这有多大一个机会，或者说，是一个错失的机会？
How big of an opportunity or or, you know, a missed opportunity?
如果他们最终无法销售这些芯片，我们不知道长期会发生什么。
If if they can't sell these chips eventually, we don't know what's gonna happen in the long run.
但这对他们来说有多大一个市场？
But how how big a size of market is this for them?
因此，英伟达过去曾表示，就在去年年底，他们在中国的销售额每个季度可能在20亿到50亿美元之间。
So NVIDIA has said in the past, as recently as late last year, that their sales to China could range from anywhere from 2 to 5,000,000,000 a quarter.
所以这对他们来说几乎是每年200亿美元的机会。
So that's almost like a $20,000,000,000 a year opportunity for them.
他们去年采取了措施，停止报告或把中国销售额计入财报，只是为了给投资者一个更清晰的盈利预期。
They took steps last year to stop reporting or factoring in China sales into their earnings just to give investors a better idea of, you know, what they will make.
所以情况就是这样。
So there's that.
但同样，如果他们能再次打开中国市场，或者将中国销售额重新纳入财报，这将为他们和投资者带来巨大的机会。
But, again, like, if they can unlock the China market again or, you know, include that in their earnings, it'll be a great opportunity for them and for investors.
让我问你一个前瞻性的问题。
And let me just ask you a forward looking question here.
我们刚刚看到了詹森关于Rubin系列芯片的所有公告。
We just saw all of the announcements from Jensen about the Rubin family of chips.
我不禁想到，这个问题以不同芯片的形式持续了这么久。
And I can't help but think about how long this issue persists in different variations with different chips.
我的意思是，今天我们谈的是H200。
I mean, we're talking about the H200 today.
几个月前我们还在谈H20。
We were talking about the H20 a couple months ago.
当Rubin系列芯片发布时，你是否预见到这种猫鼠游戏在未来几年还会持续？
When the Rubin family trips comes out, mean, like, do you foresee this cat and mouse issue persisting for the next couple years even?
是的。
Yeah.
我的意思是，最终需要找到一个平衡点。
I mean, I think at the end of the there needs to be a balance.
对吧？
Right?
美国政府不希望中国在人工智能领域超越美国。
The US government doesn't want China to surpass, you know, The US in AI.
但与此同时，他们也不希望中国发展自己的国产芯片和人工智能产业，以免不再需要英伟达的芯片。
But at the same time, they don't want China to foster its domestic chip AI industry, so that they don't need NVIDIA chips anymore as well.
比如，英伟达的首席执行官黄仁勋一直表示，中国和世界都依赖美国技术是最好的。
Like, Jensen Huang, the CEO of NVIDIA, has always said that it's better that, China, and the world relies on US technology for everything.
因此，我认为每当英伟达推出新芯片时，这始终是个问题：到底该卖多少给中国，才能既让中国公司不至于全力竞争、自主研发芯片，又不至于多到让中国在人工智能领域超越美国。
And so they have to so I think this is always gonna be an issue with whenever NVIDIA comes out with a new chip is how much can you sell to China just enough so that China does Chinese companies don't try to compete and build their own chips, but not so much that China surpasses The US in AI.
对。
Right.
这确实是一个快速变化的故事，韦恩，感谢你前来讨论这个话题。
Well, it is a fast moving story, and Wayne, I want to thank you for coming on to discuss it.
这位是我们在《信息》杂志的英伟达记者韦恩·马。
That is Wayne Ma, our NVIDIA reporter here at The Information.
谢谢你们邀请我。
Thanks for having me.
好的。
Okay.
2026年，人工智能的采用情况备受关注，关于为何其发展速度慢于许多科技公司的预期，已经出现了许多理论。
AI adoption is being closely watched in 2026, and there have been a lot of theories thrown around as to why it has been slower than many tech companies would have hoped.
本周早些时候，Conviction的莎拉·郭在X上发表了一篇关于这个问题的优秀文章。
Sarah Guo from Conviction wrote a good essay on X about that earlier this week.
她还发表了一篇关于风险投资机构如何在高度竞争的环境中保持相关性的文章，并且本周早些时候她在CES上为黄仁勋做了开场介绍。
She also published a piece about how VCs can remain relevant in a hypercompetitive landscape, and she was at CES opening for Jensen Huang earlier this week.
我想请她来谈谈所有这些内容。
I want to bring her on to talk about all of that.
莎拉，欢迎再次做客我们的节目。
Sarah, welcome back to the show.
很高兴你来到这里。
It's great to have you here.
谢谢。
Thanks.
很高兴见到你。
Great to see you.
你年初这一周挺安静的，是啊。
So quiet week for you to start off the year, Yeah.
我觉得整个行业现在都在全速运转。
Well, I think we are just running on full cylinders in the industry.
很棒。
Good stuff.
我想和你聊聊你写的一些内容。
Well, look, I want to talk to you about some of the stuff you wrote about.
我也想和你聊聊你在CES上主持的小组讨论。
I also want to talk to you about the panel that you led at CES.
你刚结束小组讨论后，黄仁勋就公布了关于Rubin芯片的一些新细节。
Right after you had your panel, Jensen Huang unveiled some new details about the Rubin chip.
我对你提出的广泛问题是：你认为NVIDIA的这一系列芯片和这些创新对AI的广泛采用会产生怎样的整体影响？
The broad question I have for you is what net impact you think that family of chips and those innovations from NVIDIA is going to have on AI adoption broadly?
是的。
Yeah.
我认为，对于更广泛的行业与芯片专家而言，有两个基本要点：第一，英伟达的创新速度对竞争对手来说真是极具压力。
I think the the two basic takeaways for the, you know, broader industry versus chip experts would be, one like the Nvidia pace of innovation is like really punishing for competitors.
从推理角度来看，这款芯片的性能比之前提升了5倍。
This is a 5X more powerful chip on a flops basis from an inference perspective.
这是一整套芯片，但系统规模非常庞大。
It is a huge like it's a family of chips, but the systems are huge.
因此，内存和内存带宽持续扩展，解决了人们此前在英伟达生态系统中担忧的许多瓶颈问题。
And so there's continued scaling of memory and memory bandwidth and a lot of the bottlenecks that people have been concerned about in the Nvidia ecosystem.
而到2026年底就推出如此重大的升级，时间非常近了。
And, you know, late twenty twenty six is a very near term timeline to be delivering such a big upgrade.
对吧？
Right?
所以，我们拭目以待，看看今年晚些时候会发生什么。
And, and so, you know, we'll see, we'll see what happens later this year.
但我认为，如果这些芯片能够按时在行业内广泛普及，应用开发者和最终用户就应该预期，我们可以用低得多的成本使用更多的令牌，对吧？
But I, I think if those chips actually get widespread in the industry on time, application developers and then end consumers should expect that we can use a lot more tokens for much less cost, right?
这正是
Which is
某种程度上
sort of
NVIDIA在这里的总体目标。
the overall goal for NVIDIA here.
当你说到如果这些芯片能按时进入市场时，我想我听到你的意思是，你假设它们最终会到位，最终会被购买。
And when you say if those chips sort of get in the market on time, I mean, I guess the assumption that I'm hearing you say is that they eventually will get there and they eventually will get bought.
只是时间问题？
It's just a question of when?
这就是你的意思吗？
Is that the idea?
当然。
Absolutely.
所以，正如之前提到的，它们已经进入全面生产阶段。
So, you know, it's been stated already that they're in full production.
是的。
Yeah.
我只是点头认同，因为这些芯片的规模极其庞大。
And I'm only nodding to the fact that these are extremely large scale, like.
对。
Right.
这是一个完整的系统。
It's the whole system.
这是一个完整的系统。
It's the whole system.
已经交付。
Delivered.
对吧？
Right?
因此，为了实现我们所讨论的这种大规模最终交付，你可能会遇到各种各样的限制。
And so you can end up with all sorts of constraints to to get to, you know, final delivery at the massive scale we're talking about.
但我确实认为，从英伟达芯片开始的推理效率这一理念，也体现在所有软件层面上，比如像Base10这样的云平台。
But I I do think this idea of inference efficiency that's starting with Nvidia chips, but it's also in all the software layers like these cloud platforms like base 10.
这对行业的影响是，许多看似更先进的AI应用，比如长周期智能体、更智能的整体应用，都将变得普及。
They, the impact for the industry is that a lot of things that look like, you know, much more advanced applications of AI, like long horizon agents, just smarter, you know, applications overall are going to be widespread.
在某些情况下，部署起来也容易得多。
And a lot easier to roll out in some cases.
这让我想到你在X上写的那两篇随笔之一，现在人们最喜欢通过在X上发文章来传达自己的观点。
It leads me to an essay that you one of two essays you wrote on X, which is articles on X are now people's favorite way to get their message across.
你写过关于AI采用的文章，我想读一下你文中的一段话。
You wrote about AI adoption, and I want to read a quick passage from what you wrote.
你说：AI带来的价值出现的位置，与组织寻找价值的位置之间存在不匹配。
You said, There's a mismatch between where AI value is showing up and where organizations are looking for it.
企业抱怨ROI缺失，尽管对模型和推理的需求正在迅速增长。
Enterprises complain about missing ROI even as demand for models and inference grows rapidly.
有人正在获得价值，只是它还不总是能在仪表板上显现出来。
Someone is getting the value, it just isn't always visible on a dashboard yet.
我想问问你关于这一点。
I want to ask you about that.
那么，谁在获得这些价值呢？
Who is getting the value then?
如果我们看不到它在仪表板上，它在公司里是从哪里体现出来的？
If we can't see it on the dashboard, where is it coming up in companies?
所以我认为，今天获得这些价值的人是多种多样的。
So I think there's like a mix of who's getting that value today.
如果你用人工智能做更好的新闻报道，你就能获得一部分价值。
If you do better journalism, right, using AI, you accrue some of the value.
也许你能早点下班。
Maybe you get to go home earlier.
也许信息获得了更好的阅读量和互动率？
Maybe the information gets better readership and engagement, right?
但我不是在轻率地说。
But I'm I'm, you know, not being glib.
实际上，生产力的提升是以不同方式实现的。
It's really like productivity gains get reaped in different ways.
这篇论文的一个要点是，我们之所以仍处于人工智能商业应用的早期阶段，原因有很多，但其中一个原因是产品层面的问题，对吧？
Part of the point of the essay is, you know, there are a lot of reasons why we're like still very early in sort of business deployment of AI, but one of them is a product question, right?
从产品角度来看，前沿能力在工作流程中的深度应用还处于非常非常早期的阶段。
Things are very, very early in getting frontier capabilities exploited from a product perspective, like deeply in the workflows.
但我这篇论文关注的原因是激励结构，对吧？
But the reason my essay focuses on is like what's the incentive structure, right?
如果我利用了所有这些生产力提升，比如说以9亿多使用聊天PT的用户为例，其他所有消费这些令牌的人实际上都在利用这些人工智能能力。
If I, you know, leverage all this productivity and like let's say 900,000,000 plus chatty PT users as an example and everybody else consuming these tokens is actually leveraging these AI capabilities.
那么，我可以把这些收益据为己有，也可以与我的企业分享。
Well, I can take those gains or I can share them with my business.
其中一些收益显然是个人层面的，比如获得了新的能力。
And some of those games are personal, obviously, in terms of new capability.
但这取决于激励结构，以及组织是否真正与员工作为所有者保持一致。
But it depends on the incentive structure and whether or not organizations really are aligned with their workers as owners.
对。
Right.
现在，你还谈到了这个故事中的变革管理角度，这一点你之前在詹森·黄上台前的小组讨论中也提到过。
Now, other thing you talked a little bit about is the change management angle to this story, and that was something that you also covered in your panel before Jensen Huang came up on stage.
另一点，我想读一下，因为我觉得这在文章中是个很好的观点：私下使用AI来提升自己的产出通常是安全的。
Another part, and I do want to read it because I thought it was a good point in the essays, you said, Using AI privately to improve your own output is generally safe.
但以公开、协调的方式推动自动化则不然。
Proposing automation in a visible coordinated way is not.
然后你进一步说，这常常被误解为对技术的恐惧。
And then you go on to say, This often gets misread as fear of technology.
更准确的理解是，人们对机构如何应对变革持谨慎态度。
It is better understood as caution about how institutions handle change.
几个月前，我和文德·霍斯拉谈过这个问题，他谈到的是那些在一线尝试使用这些工具的人。
So, I talked about this a little bit a couple months ago with Vinod Khosla, and he was talking about who the people are on the ground trying to use these tools.
你这里的评论让我想起一个我一直困惑的问题：是我们公司里没有合适的人来实施AI，还是这些人对AI有抵触情绪？
Your comments here reminded me of a question that I've been having, that, is it that we don't have the right people on the ground at these companies trying to implement AI, or just that those people are resistant to it?
我的意思是，你觉得问题出在哪里？
I mean, where do you see the problem coming up?
说实话，我不认为在某些情况下这纯粹是人员和适应能力的问题。
You know, I actually wouldn't say, you know, in some cases maybe it'll be a question of the people and the adaptability of people.
但我想问的是，所有人类，包括我，今天是否都在最大限度地使用AI？
But I'd say for all humans like MI or are you using AI to the maximum effectiveness today?
我没有。
I'm not.
我不知道你怎么样，但我真的很想做到。
I don't know about you yet, but I really want to be.
我认为这并不是因为我抗拒改变。
I don't think that's because I'm resistant to change.
我觉得改变很难，而且这种变化发生得非常、非常迅速，对吧？
I think change is hard and this changes happen really, really rapidly, right?
因此，这是我们应当具备的一个基本认知，即对于每个人来说，变革和技术采用都是困难的。
And so that's just like a fundamental recognition we should have, which is, you know, change and technology adoption is hard for every human being.
如果组织希望实现变革，就必须在这方面进行真正的投入。
And if organizations want to happen, they really have to invest in that.
我认为许多优秀的企业已经理解了这一点。
I think like a lot of great organizations already understand that.
但有时我们陷入了一种误解的漩涡，低估了这一切发生的速度。
But sometimes we like we're in this like vortex of misunderstanding the time scales of how quickly this is all happening.
它已经持续了
It's been
三年了，对吧？
happening for three years, right?
所以我认为这是一个认知，但恐惧，我认为恐惧其实是相当合理的。
And and so I I think that's one recognition, but the the fear, I think actually fear is really rational, right?
如果我利用人工智能，找到方法自动化我工作的一部分，并且我足够聪明和了解人工智能能力正在进步，那么自然会问：我的角色会改变吗？
If I think that I leverage AI and I figure out how to automate parts of my job and I am smart and informed enough to know that AI capabilities are progressing, it's a natural question to ask, well, does my role change?
对吧？
Right?
而且我是否
And do I And get
这会引起一些不安，我是否愿意将自己的名字与这个计划、这个损益挂钩？
that causes some nervousness around, do I want to attach my name to this initiative, to this P and L?
当然。
Absolutely.
而且我认为，如果这种变化也影响到其他人，当你在管理结构中成长时，这取决于组织内部的激励机制，对吧？
And I think if that change affects other people too, when you like grow up in the management structure, it depends on the incentives within organizations, right?
好的。
Okay.
那我是否想改变阿卡什的职位？
Well, do I want to change Akash's role?
我是否想改变萨莉的职位？
Do I want to change Sally's role?
那么，这对我在组织政治中的地位有什么影响？
And like what does that do for me within the politics of an organization?
我不知道。
Don't know.
这些组织就是有这么多东西。
It's just like these organizations have a bunch.
每个大型组织都有各种混合的激励机制，而在这场向AI原生公司转型的过程中，赢家必须找到方法，使这些激励机制与组织的成功保持一致。
Every large organization has a bunch of mixed incentives and the winners in this, transition to AI native companies are going to have to figure out how to align that with the success of the organization.
你在小组讨论中提到，医疗保健领域一直是一个被忽视的行业，但其采用AI的速度远超预期。
Now, you talked in your panel about how healthcare has been sort of a sleeper sector that has really adopted AI faster than some would have otherwise expected.
Open Evidence 是你投资组合中一家取得巨大成功的公司。
Open Evidence is a portfolio company of yours that has seen a lot of success.
你能谈谈为什么医疗保健会成为这样一个反常的案例吗？
Can you talk about why healthcare has been sort of this oddball case study?
Open Evidence 的产品究竟有什么特别之处，使其获得了如此大的市场反响？
And what is it about Open Evidence's product that has really gotten traction?
我认为，在Open Evidence的情况下，这里存在一种非常特别的商业模式契合。
I think there is a business model, you know, alignment here that is really special in the case of Open Evidence.
然后还有一个技术结构上的原因。
And then there's like a technical structural reason.
简单来说，Open Evidence是一个任何医生都可以作为消费者使用的应用，对吧？
So really quickly, open evidence is an app that any provider can use as a as a consumer, right?
你可以像在应用商店下载一个应用一样开始使用它，它是免费的，能帮助你应对日常工作中一个非常困难的挑战——理解医学研究，成为最知情、最专业的医生，并在此基础上扩展到协助各种其他互动。
You like downloaded an app store and you just start using it and it's free and it helps you with your day to day work against a really difficult challenge, which is understand medical research and be the best provider, the most informed provider you can and expanding capabilities from there into sort of helping with all sorts of interactions.
但这里没有任何利益冲突。
But like there's no misalignment.
它只是让我更好地完成工作，而Open Evidence通过广告模式来支持这一点，对吧？
It just makes me better at my job and open evidence has an ad based business model that supports that, right?
它不需要推动变革管理，直到全国一半的医生已经在使用它，这要容易得多，对吧？
It doesn't have to do change management until half the doctors in the country already using it, and that's a lot easier, right?
你知道，组织是被它的用户说服了它的价值，因此我认为这对整个行业来说是一个非常有趣的案例，即消费者和准专业人士的模式可以成为一种强大的采用机制。
You know the organization is convinced by its users that's useful, and so I I think this is just a really interesting case study for the industry, which is like consumer and prosumer can be a really powerful adoption mechanism.
AI 对医疗保健之所以可能非常有趣，是因为大量的数据和互动都是非结构化的，对吧？
The technical reason why AI could be really interesting for health care is that a lot of the data and a lot of the interactions are unstructured, right?
因此，大型语言模型和代理在这些碎片化的系统中处理自由文本、医学研究、电话中的医生笔记等方面表现得非常好。
And so, you know, large language models and agents like in these fragmented systems working on free text, medical research, doctor's notes in phone calls.
它非常擅长
It's really And good at
我的意思是，人们已经尝试了很长时间来让这些内容变得更结构化一些。
I mean, they've been trying to find ways to make that a little bit more structured for a while now.
我知道你得走了，但我很好奇。
I'm curious, I know you got to run here.
你几周前在你的播客中发布了这一期预测，你其中一个预测是，你预计机器人领域的整体情绪会发生某种变化，或者至少是机器人某些部分的情绪。
You put out this predictions episode on your podcast a couple weeks ago, and one of the predictions you made was that you are forecasting some kind of a change in the sentiment around robotics, or at least parts of robotics.
我觉得这很有趣，因为关于物理AI和机器人的讨论正变得越来越热门——今年年初的头几周，这已经成为每个人关注的焦点。
I thought that was interesting because the conversation around physical AI and robotics is becoming so much more- it's much more of a focus for everyone really, even in the first few weeks of the year here.
谈谈你为什么认为机器人领域的情绪正在变化，以及你预期会发生什么吧。
I mean, talk to me a little bit about why you see the sentiment around robotics changing and what you're expecting there.
今年对人工智能机器人行业来说是高风险的一年。
This is a very high risk year for the AI robotics industry.
但这也正是我感到兴奋的原因，对吧？
But that's why I'm excited, right?
我认为，包括我们投资组合中的星期日机器人在内的多家公司，今年将真正面对现实。
I think that there are a number of companies, including Sunday Robotics in our portfolio, that are like meeting reality this year.
它们计划向消费者家庭交付真实产品进行测试，这些产品的功能与以往的产品不同，因为它们是通用型的，对吧？
They intend to ship real products to consumer homes for testing, and these are like their sub function different than the products that have existed before because they are, you know, generally capable, right?
它们不是为单一任务设计的，就像大语言模型与之前的特定任务模型有所不同一样。
They're not designed for a single task, much like the you know LMS were different from the pre foundation model task specific models.
因此，我认为市场情绪正在改善。
And so I think there's sentiment improving.
或者说，人们对这一领域产生了极大的兴趣，原因有两个。
Or you know, there's like all this heightened interest for two reasons.
一是所有研究人员都相信，在模型、数据和部署方面都将取得巨大进展。
One is like all the researchers believe that there can be a huge amount of progress on both the you know, models, data and deployment side.
当然，如果这一切成真，今年将会有大量投资和许多真实的测试、现实世界评估发生。
And then of course, like if that is real, there's a lot of investment and a lot of big tests, real world evaluations that are going to happen this year.
这也是詹森在CES主题演讲中的一个重要部分。
And it was a huge part of Jensen's CES keynote as well.
他一直是物理AI领域的早期信仰者之一，部分原因在于。
He's been one of the earliest believers in the physical AI world, in part because that you know.
这放大了商业机会的规模。
Multiplies the size of the opportunity for.
我知道。
I know.
我的意思是，商业。
I mean the business.
商业前景确实存在，
The business case is certainly there,
你知道，他长期投资于此，这是一项极具信心的押注。
you know, it's a really high conviction bet to be investing in it for as long as he has.
我也是一个坚定的信徒，但今年我们将会面对现实。
And I'm a big believer, but we'll sort of meet reality this year.
但那种情绪的崩溃会发生在什么时候呢？
But where does that sort of collapse in sentiment happen then?
我的意思是，你预测它会在哪里崩溃？
Mean, where do you forecast it sort of breaking down?
所以这些技术目前还没有在实际生产中应用，对吧？
So none of this works yet in production, right?
如果问我，是否存在一个能在任何现实环境中运作的通用机器人？像我和其他业内人士都会说：不，不可能。
If you ask me, is there a generalist robot in any real world environment like me and other people from this industry are going to say no, right?
所以这是一个非常重大的开放性问题。
So it's a really big open question.
机器人这类产品真的很难交付，对吧？
They're like amongst robots are really hard products to deliver, right?
比如，可以把自动驾驶汽车当作一个例子。
Like, you know, think of autonomous vehicles as an example of it.
是的，安全论证非常重要，而现实世界总是比人们希望的更加多变，对吧？
Yeah, the safety case is really important and the real world is always more varied than one would hope, right?
就像你知道的，
Like you know,
它是
it's
很容易想象在家庭环境中，但也在工厂或数据中心里。
easy to imagine like in the home, but also you know in factories or in data centers.
所以这些机器人必须对这些环境非常稳健，而这些产品本身就很难以交付。
So you know these robots have to be really robust to those environments and the products are really hard to deliver to begin with.
因此，我们会看到这些公司失败多于成功，而且它们的运输成本非常高，对吧？
And so we are going to see more failures than successes of these companies, and they're very expensive to ship, right?
我认为，如果它们能成功，它们可以成为硬件公司。
I think if they work, they can be their hardware companies.
可以是，也许还是具有广泛应用场景的通用型机器人。
Can be and perhaps general ones with really expansive applications.
可能会出现像特斯拉、苹果、通用电气、西门子这样的硬件公司，它们对制造业有重大影响，但这并不是我们关注的领域。
Could have hardware like Tesla, Apple, you know, GE Siemens type companies that are really impactful to the manufacturing world, which is not their area of interest for us.
但未来六个月
But there's going to
会有很多失败。
be a
在六个月内也会有很多失败。
lot of failure as well in six months.
这让我想到一个有趣的问题，那就是，这个循环究竟从哪里开始？
And I think it kind of leads to an interesting question for me, which is that, you know, where does that cycle sort of start?
因为，我认为你所指出的是，投资者会意识到，嘿，这些东西并没有像预期那样有效。
Because, you know, I think what you're pointing to is that investors will wake up to the reality that, Hey, this stuff is not working as well.
我们不应该继续给予如此高的估值，这些估值已经开始攀升。
We shouldn't be affording it, the valuations that we've seen sort of start to creep up.
另一方面，大型公众科技公司仍在将其作为应用场景进行宣传。
On the other hand, you have big public tech companies continuing to talk about it as an application.
在我看来，我在想什么是鸡，什么是蛋？
I guess in my mind, I'm sort of wondering what's the chicken, what's the egg?
如果英伟达继续谈论它，我不确定，什么才是先发生的？
What comes first in terms of if NVIDIA continues to keep talking about it, I don't know.
投资者要花更长时间才能意识到这个现实吗？
Does it take longer for investors to wake up to that reality?
我想知道你怎么看。
I wonder what you think.
哦，我不知道。
Oh, I don't.
我完全不这么认为。
I don't think so at all.
我会反驳说，我不觉得这些价格是错的。
I would push back and say, I don't know that the prices are wrong.
就像那些人一样，我认为其中一些人会在极高的价格上是对的，而另一些人会错，就像任何其他高价值市场一样，对吧？
Like some of those people, I think some of those people are going to be right at massive prices and some of them are going to be wrong, like in any other really high value market, right?
这是关于挑选赢家，而英伟达一直是这个生态系统的重要贡献者。
It's about picking winners and Nvidia has been a huge contributor to this ecosystem.
比如，提供模型、开源数据框架以及硬件本身。
Know, delivering models, open source data frameworks, the hardware itself.
因此，我认为他们无疑在支持这一成功案例，正如你所预期的那样。
And so I think they're definitely supporting the success case here, as you would expect.
我只是想说，这其实是一个问题：它到底能不能行得通？
I would just say I think it's a it's a question of like, does it work or not?
作为投资者，我们当然会经历情绪的波动，你知道的。
And we can certainly have waves of sentiment as you know, an investor.
当某个具体的投资失败时，你的反应可能是这个投资失败了，或者整个行业失败了。
A specific bet doesn't work, your reaction can be that bet didn't work or the sector didn't work.
所以我认为这可能是情绪转变的原因。
So I think that could be the change in sentiment.
很好。
Great.
萨拉，感谢你再次做客我们的节目。
Well, Sarah, I want to thank you for coming on the show again.
每次和你聊天都特别精彩。
It's always a great conversation.
这位是萨拉·郭，Conviction的创始人兼投资人，欢迎收看TI TV。
That is Sarah Guo, founder and investor at Conviction here on TI TV.
好的。
Okay.
在某些方面，如今投资AI从未如此简单，因为每一家上市公司都在抓住机会，利用这项新技术全面改造自己的业务。
In some ways, it's never been easier to invest in AI with every public company jumping at the opportunity to overhaul its business in the face of the new technology.
但另一方面，鉴于股市对AI新闻的反应如此剧烈，想要通过一项新颖的AI交易跑赢市场，正变得越来越困难。
On the other hand, given how much of the stock market is reacting to AI news, it's becoming a whole lot harder to beat the market with a novel AI trade, should say.
这一现象正是我们的财经编辑肯·布朗在他今天发表的每周财经专栏中所探讨的内容，我请他来为我们详细解读一下这篇专栏。
That is a phenomenon that Ken Brown, our finance editor, wrote about in his weekly finance column out today, and I to bring him on to tell us more about his piece.
肯，欢迎再次来到节目。
Ken, welcome back to the show.
很高兴你来到这里。
It's great to have you here.
嗨，阿卡什。
Hi, Akash.
那么，是什么让你选择在这一年第一周、当人们展望投资组合时写这篇专栏呢？
So what made you want to write this column on this day, the first week of the year as people are looking ahead to their investment portfolios.
我的意思是，我们通常不会写这类信息。
I mean, is not something we typically write about, the information.
没错。
Right.
有两个原因。
So there's two reasons.
一是今年我们可能会看到一些AI公司的IPO。
One is this is going to be a year where we're going to probably see some AI IPOs.
有传言称Anthropic今年将上市，而OpenAI可能明年跟进。
There's talk that Anthropic will go public this year and OpenAI will follow maybe next year.
投资者们急于进入这个领域。
And investors are eager to get into this stuff.
另一件事是我有几位亲戚问我，因为我一直从事人工智能和金融相关的工作，他们问我应该如何投资人工智能？
The other thing is I got a question from a relative of mine, a couple of relatives, because I deal with AI all the time and I deal with money and how should we invest in AI?
于是，我做了一些研究，查看了市场上有什么，并决定现在是写一篇专栏的好时机。
And so, I did a little research and looked at what was out there and decided the timing was right to write a column.
好的。
Okay.
这篇专栏很好地深入探讨了一些公司为投资者提供参与人工智能方式的创新做法——我说的公司，是指那些主营业务就是为散户投资者创造盈利机会的公司。
So, the column was great at diving into some of the more creative ways that companies are offering exposure to And by companies, I just say, companies whose business it is finding retail investors ways for them to make money.
谈谈我们看到的市场上的一些交易，以及它们是如何发展的。
Talk a little bit about some of the deals that we've seen in the market and how that's shaped up.
是的。
Yeah.
目前，这些大型人工智能公司大多仍是私营企业。
So, most of these big AI companies are still private.
像谷歌这样的公司是上市的，但纯粹的AI初创公司都是私营的。
There are public ones like Google and all that, but the pure plays, the startups are private.
因此，嘉信理财、Robinhood和其他经纪公司都在试图找到方法，帮助投资者进入这些私营公司。
So both Schwab and Robinhood and other brokers are trying to figure out ways to help investors get into these private companies.
我的研究显示，个人这样做通常不是一个好主意。
My research shows it's generally not a very good idea for individuals to do this.
费用很高。
The fees are high.
他们总是在错误的时间入场。
They get in at the wrong time.
价格很高。
The prices are high.
现实是，他们总是最后一个拿到这些机会的人。
The reality is they're last in line to get this stuff.
所有那些强大的投资者——风险投资公司、私募股权公司、主权财富基金——都早已抢先入局了。
All the big powerful investors, the VCs, the private equity firms, the sovereign wealth funds, they're all in there first.
所以我们这些小投资者并不是优先考虑的对象，也得不到最好的交易。
So, us little guys, we are not a priority and we don't get the best deals.
所以，我只是想提醒一下，如果有人对此感到兴奋，我想至少能让他们冷静一点。
So, just wanted to If people got excited about it, I wanted to sort of at least sober them up a little.
为了明确一下，这指的是个人投资者通过这些新兴平台寻找投资这些私人公司的途径，这些平台似乎越来越受欢迎。
And just to be clear, this is individual investors finding ways to get exposure to these private companies through these newer platforms, I guess, that have become more popular.
这正是你持谨慎态度的原因。
That's sort of where your caution is coming into play here.
我的谨慎主要基于两点。
Well, so my caution is two things.
一方面，确实有经纪商在推动，投资者也要求获得投资这些公司的机会，比如Anthropic、OpenAI等等。
So one is yes, there is a push by brokers and investors are asking for it to get access to these companies, Anthropic, OpenAI, all that stuff.
我想投资人工智能。
I want to invest in AI.
我想进入这个领域。
I want to get into this stuff.
我该怎么做到呢？
How do I do it?
所以，这种情况正在发生。
And so, that's happening.
另一件事是，人们正在以各种新颖的方式对公开上市的公司进行押注，或者通过其他他们认为可以获得AI投资机会的方式。
The other thing is people are making bets in all kinds of novel ways on the stock market with the companies that existed publicly and other ways that they feel like they can get AI exposure.
我也提醒人们注意这一点。
I also warn people about that.
所以这有点像个人理财专栏，因为它是
So this is a little bit of a personal finance column because it
这是一场金融对话。
would It's a finance call.
这纯粹是一场金融对话，肯。
It's just a finance call, Ken.
它可以是任何东西，而且你
It can be whatever And you
通常不会那样做，但这是一系列事件的汇合。
don't tend to do that, but it is a confluence of events.
所以人们现在购买的这类基金的历史记录非常糟糕。
So the track record of these kinds of funds that people are buying now is terrible.
这些基金表现很差。
The funds perform badly.
它们可能在一段时间内表现很好。
They can do great for a while.
然后当所有人都把钱投进去时，他们表现得很糟糕。
And then when everyone puts the money into it, they do terribly.
很多这样的公司都倒闭了。
A lot of them shut down.
大多数都跑输市场。
Most of them trail the market.
我的意思是，当你查看数据时，大约80%的公司会在三年后倒闭或落后于市场。
I mean, when you look at the data, like 80% of them shut down or lag the market after three years.
所以我只是想让大家意识到，这确实令人兴奋。
And so I just want people to be aware that this is exciting.
我们整天都在写和谈论它，但对于个人投资者来说，进入这个领域必须保持明智和谨慎。
We spend all day writing and talking about it, but for an individual investor to get into it, you just gotta be smart about it and careful.
另一个问题是，每个人都已经持有这些东西了。
The other problem is everyone owns this stuff already.
谷歌、微软、Meta，这些公司都是上市公司。
Google, Microsoft, Meta, all these companies are public.
它们占了股市的约25%到30%。
They make up around 25 to 30% of the stock market.
所以，每个人都已经暴露在其中了。
So, everyone's exposed to this.
那么，你为什么还需要更多呢？
So, why do you need any more?
哇。
Wow.
因为更多就是更多，肯。
Because more is more, Ken.
意思是，这正是关键所在。
Mean, that's the whole point.
没错。
Exactly.
这是一个非常令人兴奋的领域，将会涌现出大量公司和大量赚钱的机会。
And it's a very exciting area and there's going to be lots of companies and lots of money to be made.
因此，在我结束这篇专栏时，我说：我已经给出了所有这些建议，明年每个人都会回来对我说：我们都没听你的建议，却赚了这么多钱。
And so, as I end the column, I say, I've given all this advice and next year, everyone's going to come back to me and say, We made all this money ignoring your advice.
对，这通常是节日聚餐的结局。
Right, that's typically how holiday dinners end up going.
是啊，是啊，是啊。
Yeah, yeah, yeah.
所以你也可以回来找我
And so you too can come back to me
展开剩余字幕（还有 285 条）
然后吃吧。
and eat.
不，不，不，听好了，我从不投资这些东西。
No, no, no, look, I don't invest in any of this stuff.
我们只是写作。
We just write.
我们就做这个。
That's all we do.
没错。
Exactly.
但我确实想和你聊聊你刚才简要提到的另一个领域——能源行业。
But I do want to talk to you about another pocket that you touched on briefly, which is the energy sector.
能源行业是人们一直谈论的一个机会，但当我更深入地思考，并读了你的专栏后，我又开始想，如果你只是买入标普500指数，那些能源公司本来也都是跟着这些交易在走的。
And the energy sector is another area that people have been talking about as an opportunity, But then as I thought about that more and I read your column, then I got to thinking again, like if you just buy the S and P, I mean, the energy companies are all moving according to these deals anyway.
没错。
Right.
因此，指数基金的魅力在于，很多人都持有它，可以接触到所有资产。
And so the beauty of index funds, which is what a lot of people own, is you get exposure to everything.
所以你永远不会错过任何一项大幅上涨的投资，这非常棒。
And so you never miss a big run up in something, which is fantastic.
是的，我知道能源公司，它们是一个波动性很大的行业，人们正把大量资金投入人工智能，但那些记忆更久远的人——记忆超过一年的人——还记得两年前最热门的能源投资是转向太阳能、风能和电池等。
Yeah, I know the energy companies, I mean, they and it's a volatile sector and people are pouring a lot of money into AI, but people with longer memories, memories longer than a year, remember that the hottest energy trade two years ago was the energy transition to solar and wind and batteries and all that stuff.
这项投资对投资者来说结果非常糟糕。
That trade has turned out to be a terrible trade for investors.
他们亏了很多钱。
They lost a bunch of money.
所以现在人们又在想，能源，哦，为人工智能提供动力的能源。
And so now people are thinking energy, oh, energy to power AI.
这确实可能非常有趣，但这是一个波动性很大的行业，过去的表现并不理想。
And it's true, it could be really interesting, but it is a volatile sector and the track record in the past is It just ain't great.
没错。
Right.
我再问你一个问题。
Let me ask you one more question.
正如我们之前讨论的投资或获取私营公司敞口时，我不禁想到，公司通常在准备好上市时才上市，也就是说，它们对自己的现金流和收入增长有了一定的可预测性。
As we talked about investing or finding exposure to private companies, I couldn't help but think about the notion that companies go public when they're ready to go public in a sense that they have some sort of predictability about what their cash flows, what their revenue growth will look like.
这让我意识到另一个需要警惕的点：如果一家公司是私营的，虽然可能有各种财务原因导致它尚未筹集到所需资金，但也可能是因为公司自己觉得：‘嘿，我还没准备好成为一家上市公司。’
I mean, it struck me as another point of caution here that if a company is private, I mean, there is all of the financial reasons why maybe they haven't had to raise all the money, but there also might be sort of this reality that the company says, Dude, I'm not ready to be a public company.
我们还不想让散户投资者参与。
We don't want retail investors yet.
我想知道，你是否认为这也是一个需要考虑的因素。
I wonder if you think that that's another consideration of this.
确实如此。
Well, that's been true.
像Stripe、SpaceX等一些公司已经保持私有状态很长时间了，远超人们根据其规模和成功程度的预期。
A bunch of companies like Stripe and a few others, SpaceX have stayed private a long time, longer than you would think given their size and success.
是的。
Yeah.
那么，谁会走出去呢？
And then who goes out?
谁会上市？
Who goes public?
我的意思是，一个问题在于，他们是否迫切需要资金？
I mean, one question is, do they desperately need the cash?
他们是否已经用尽了私募市场中所有其他融资方式？
Have they exhausted every other way to raise cash in private markets?
所以，他们选择上市。
So, they're going public.
那么，你愿意成为那个购买这种急需资金的公司的投资者吗？
And then, do you want to be that person buying that company that's desperate for cash?
它们的增长率非常出色，但股市中的公众投资者关心的是利润。
And the growth rates are fantastic, but public investors in the stock market care about profits.
最终他们关心的是利润，而我们短期内还看不到这一点。
Eventually they care about profits and we're not seeing that for a while.
所以，它们将会非常有趣。
So, they're gonna be really interesting.
这对我们来说将是一个精彩的故事。
It's gonna be a great story for us.
我很高兴能报道它，但我不确定是否想投资它。
I'm happy to be covering it, but I don't know if I want to invest in it.
很好。
Great.
好了，肯，感谢你前来做客。
Well, Ken, I want to thank you for coming on.
这位是肯·布朗，我们《The Information》的财经编辑。
That is Ken Brown, our finance editor here at The Information.
好的。
Okay.
说到IPO，据彭博社的一篇报道，Discord已秘密提交了上市申请。
Speaking of IPOs, Discord has confidentially filed to go public according to a report from Bloomberg.
这将是最新一家最终在股市上市的后期阶段私营公司，尽管报道称该公司也可能决定不推进上市。
It would be the latest in a string of late stage private companies finally listing on the stock market, although the report said the company could also decide not to go ahead with things.
我想邀请一位对消费者通讯领域有密切跟踪的风险投资人。
I want to bring on a venture capitalist who has followed the consumer messaging sector fairly closely.
梅赛德斯·本特是Premise的联合创始人。
Mercedes Bent is a co founder of Premise.
梅赛德斯，欢迎再次做客我们的节目。
Mercedes, welcome back to the show.
很高兴你能来。
It's great to have you here.
谢谢你的邀请。
Thanks for having me.
那我们来聊聊Discord吧。
So let's talk about Discord.
这是一家我们之前在节目中还没讨论过的公司，所以我很高兴你能帮助我们理解它。
It's a business that we haven't talked about yet here on the show, actually, so I'm excited to have you help us make sense of it.
直截了当地说，Discord 是怎么赚钱的？
Brass tacks here, I mean, how does Discord make money?
是通过高级订阅吗？
Is it a premium subscription?
是靠广告吗？
Is it ads?
它是怎么运作的？
How does it work?
Discord 通过三种方式赚钱。
Discord makes money in three ways.
其中一种是很多人熟悉的订阅服务，也就是他们为用户提供的 Nitro 高级订阅模式。
There's kind of the subscriptions that a lot of folks are familiar with, Nitro being the premium subscription model that they have for users.
他们还有我所说的创作者收入。
They also have their, what I call creator revenue.
这是来自社区组织者的收入，也就是那些实际负责管理和创建社区 Discord 的人。
This is where a lot of the revenue from their community organizers, the people who actually moderate and create the community Discords.
然后是第三种方式：广告。
And then they have the third, which is advertising.
还有其他一些方式，但这些都是他们盈利的三大主要类别。
So there's a couple of others, but those are kind of the three largest categories of how they're making money.
根据你所了解的情况，我们能否判断这三者中哪一个规模最大，或者增长最快？
And do we have a sense as to which of those three categories is the biggest or the fastest growing even from what you've seen?
嗯，我最后看到的信息是内部数据，所以我们就不分享了。
Well, I think the last thing I've seen was private info, so we'll share.
但我觉得他们——你知道的，我们喜欢保持私密。
But I think they, you know, it's- We like private
顺便说一下，你在节目中谈私密信息时要注意。
info when you're on the show, by the way.
你知道，对于消息和社交平台来说，广告收入的一个有趣之处在于，它在很大程度上取决于广告商认为你的用户有多大的价值。
And you know, I do think that one of the things that's interesting about advertising revenue generally for messaging and social platforms is it's largely dependent on how valuable advertisers think your users are.
而许多公司遇到问题的地方，往往在于你的用户在地理分布上处于哪里。
And a lot of the ways that this can come to trip up different companies is where are your users located geographically?
例如，我读过一些网络报告，称Discord大约有25%的用户在美国。
For example, I've read reports online that, you know, Discord has about 25% of their user base in The US.
如果你将此与Reddit对比，我看到Reddit的美国用户比例接近48%到50%，这是一个巨大的差异，因为一旦超出美国，从广告角度看，用户的价值就会下降，因为他们消费水平不同。
If you compare that to Reddit, where I've seen their user US based percentages closer to 48 to 50%, that's a really big difference because when you start to go outside of The US, users are worth less from an advertising perspective because they don't spend the same amount.
有时候，我一直使用一个大致的经验法则。
Sometimes, you know, I've always used a rule of thumb around.
这几乎与该国相对于美国的GDP或货币差异成正比，例如，巴西的用户价值可能只有美国的五分之一，尽管该国采用社交和通讯产品的倾向非常高。
It's almost proportional to the GDP or currency difference from that country to The U S so for example, for Brazil, it might be worth one fifth, even though that country has very high propensity to adopt social and messaging products at a very high rate.
因此，我认为，如果他们真的推进此次IPO，这将是关于收入质量的一个关键问题。
So I think that's one of the types of quality of revenue questions that would come up if they reportedly go forward with this IPO.
所以，这是我想要关注的一点。
And so that's something that I would be looking for.
那么，目前的用户规模与Reddit相比如何呢？
And what about the user base right now, the size of it compared to something like Reddit?
我的意思是，你对这方面熟悉吗？
I mean, is that something that you're familiar with?
据报道，它要小得多。
It's reportedly quite a bit smaller.
我看到过报道，Discord 的用户数大约有两亿多，那是去年的数据。
I've seen that Discord has around 200,000,000 plus well, that was what was reported last year.
所以我们看看接下来几个月他们能增长多少。
So we'll see how much they might improve in the last couple of months.
但相比之下，Reddit 的月活跃用户超过12亿。
But compared to, you know, Reddit has over a 1,200,000,000 of monthly active users.
当你开始思考这对广告收入意味着什么时，规模上的差异就非常显著了。
There's a significant difference in scale when you start to think about what that means for advertising dollars.
但Reddit就是你所说的对比对象吗？你笑了。
But is Reddit the comp then that you would say is the you're smiling.
我知道这是每个人都想问的问题。
I know this is the question everyone wants to know.
我和谁竞争？这会如何影响事情？
Who do I compete against and how does that affect things?
但你知道，除了Reddit，你还会提到哪些其他平台呢？
But, you know, other than Reddit, which other names would you put up there?
我认为这是一个非常好的问题，因为并非所有网络都相同。
I think this is a really great question, and not all networks are the same.
Discord本质上是以兴趣为导向，而不是以身份为导向的网络。
Discord is really an interest first, not identity first network.
如果你思考一下，我们该如何将那些大型平台归类到这两类中呢？
And if you think about, okay, how do we categorize the different big platforms that fall into those two groups?
Reddit、WhatsApp、Telegram可能更偏向兴趣这一端。
Reddit, WhatsApp, Telegram might be more on the interest side of things.
而Instagram、Snapchat、Facebook、TikTok则更偏向个人资料这一端。
Whereas Instagram, Snapchat, Facebook, TikTok would be more on the profile side.
这对我来说意味着，作为消费者投资者或消费者风投，用户是因为兴趣而来平台，而不是为了构建自己的个人资料。
What this means to me is that as a consumer investor, a consumer VC is users are coming to the platform because of their interest, not because of what profile I am building.
LinkedIn也同样是一个以个人资料为导向的平台。
It's not LinkedIn would be more of a profile oriented platform as well.
这极大地改变了你构建平台的方式以及如何实现盈利。
And that changes a lot about how you build the platform, how you can monetize it.
因此，我认为更合适的比较对象可能是像Reddit这样的消息平台和兴趣导向型平台。
And so I think the better comps are probably more the messaging platforms and interest based platforms like Reddit.
有趣的是，Discord在2015年早期正是通过Reddit大量分发用户而起步的。
Funny enough, Discord in the early days in 2015 actually got its start by getting a lot of users distributing through Reddit.
所以这简直是一个完美的闭环。
So this is a bit of a full circle moment.
但我认为它更接近Reddit和WhatsApp，而不是Instagram或TikTok。
But I think it might be a little bit more akin to what if to Reddit and WhatsApp than to a Instagram or a TikTok.
或者Snap。
Or a Snap
就此而言。
for that matter.
没错。
Exactly.
所以如果我们以Reddit作为参照，我正在思考Discord未来的商业模式以及它可能的发展方向。
So if we go with maybe Reddit as a comp, I'm sort of trying to think about the future of Discord's business and where you think it could go.
Reddit在数据方面已经做了很多工作，并且也涉足了自己的AI领域。
Reddit obviously has done a lot with its data, and it's sort of had its own AI play.
你刚才提到的Discord的三种收入来源，你认为还可能出现新的收入来源吗？
I mean, those three revenue streams you talked about with Discord, could you see there being new revenue streams?
或者你认为未来的业务形态会如何变化？
Or how do you think the shape of the business changes in the future?
这是个非常好的问题。
It's an excellent question.
我认为Reddit有很多值得借鉴的经验。
I think there's a lot of lessons that can be learned from Reddit.
目前，大规模消费平台最大的优势之一就是它们的数据可以成为新的变现途径。
Mean, one of the biggest advantages that scaled consumer platforms have right now is that their data can be a new monetization stream.
在过去十年里，当一家消费公司或任何公司告诉我‘我们要通过数据变现’时。
This is something that the last decade when a consumer company or any company told me, we're gonna monetize our data.
我以前常常翻白眼，说：当然当然。
I used to sort of roll my eyes and say, sure, sure.
你知道，人人都这么说，但我现在不得不收回前言，因为过去几年里，由于基础模型的出现，这真的变成了一项切实可行的能力。
You know, everyone says that, but I'm now eating my words because in the last few years, it actually has become a real capability due to the foundation models.
所以如果你看看Reddit与一些大型基础模型建立的合作关系，以及Reddit在2025年8月之前一直是OpenAI及其他几个聊天基础模型最主要的信息和资源来源——尽管此后OpenAI大幅降低了其排名，Reddit确实曾是第一信息源。
And so if you look at the partnerships that Reddit has formed with some of the larger foundation models, and now Reddit being the number one, at least until August 2025 when OpenAI began deep ranking and down ranking them quite a bit, Reddit really was the number one source of information and resources on OpenAI and quite another, a few other chat foundation models.
因此，我认为Discord也可以做类似的事情。
So I do think Discord could do something like this.
显然，多年来他们多次与用户群体发生冲突，而这些用户非常直言不讳且充满热情，他们清楚自己为什么使用Discord以及来这里的目的。
It obviously, they've had a lot of instances over the years where they've had clashes with their user base who is very vocal and passionate about why they use Discord and what they're here for.
所以我想象这将是另一个需要谨慎应对的雷区。
So I imagine that would be another fun minefield to navigate.
但我确实认为，对于他们来说，这无疑是一个真正的机遇：利用基础模型庞大且迫切的资金预算，来获取他们的数据进行使用。
But I do think it certainly could be a real opportunity for them to monetize via the large, large hungry budgets of the foundation models to use their data for that.
那么，让我问你一个问题。
Well, and so let me ask you this.
我正在这里查一下。
I'm just looking it up here.
所以我正在看Reddit的收入倍数。
So I was looking at Reddit's revenue multiple.
现在我正在Kofin上查看。
And so right I'm on Kofin here.
他们目前的股价是未来十二个月收入的17倍，过去十二个月收入的24倍。
So they're trading at about 17 times next twelve months revenue, 24 times last twelve months revenue.
所以如果你随便说说，你觉得Discord会获得和Reddit一样的估值倍数吗？
So if you were to just throw it out there, do you think Discord gets a Reddit size multiple?
是更高吗？
Is it higher?
还是更低？
Is it lower?
哦，你是要我随便说点完全错误的东西吗？
Oh, you're asking me to be just totally wrong on whatever I say in terms of
现在是预测季节。
what It's prediction season.
听起来就是这样。
That's what it sounds
就是预测季节，你知道的？
like It's prediction season, you know?
在我看来，如果我把这些线索串起来，你提到了用户基础，顺便说一句，用户群体的地理分布也是我们之前在节目中讨论ChatGPT时提到过的话题，即其用户主要分布在北美还是海外。
It sounded to me, if I just connect the dots, mean, you talked about the user base, and by the way, the geography of the user base is something we've talked about on the show with ChatGPT as well, and the extent to which its users are located in North America or abroad.
你知道，25%的用户在美国，我想你提到过这个数字。
You know, 25% of users The US, I think, is what you mentioned.
这是最新的估算。
That's the latest estimate.
所以从倍数角度来看，它可能没那么值钱。
So maybe it's not worth as much on a multiple basis.
我不确定。
I don't know.
是的，可能不是。
Yeah, possibly not.
很多时候，市盈率不仅是绝对规模的函数，还取决于你的增长率和盈利能力。
A lot of times multiples are also a factor of your rate of growth and of your profitability levels, not just the absolute
规模。
scale.
是的。
Yeah.
是的。
Yeah.
但我同意你的观点。
But I agree with you.
我认为，公开市场通常奖励规模、增长率和盈利能力的综合表现。
I think typically the the public markets are rewarding a total combination of scale plus rate of growth plus profitability levels.
因此，对于其中一些数据点，我们没有相关信息，但大致来看，基于用户规模，我认为这将是一次规模较小的IPO。
And so some of these and for some of these data points, don't have information about, but broad strokes, it it probably would be a smaller overall IPO would be my guess just based on this size of the user base.
很好。
Great.
但我不太确定具体的收入倍数。
But And won't say exactly on revenue multiples.
对。
Right.
还有，Mercedes，我想快速澄清一下。
And and, Mercedes, I just wanted to clarify very quickly.
你并不是 Discord 的投资者，对吧？
You you are not an investor in in Discord.
对吧？
Right?
不是。
No.
好的。
Okay.
好的。
Okay.
只是想确认一下。
Just wanted to make sure.
我猜，即使没有参与公司的人，私人文件也会流传出去
I guess private documents float around even for people who aren't involved
在公司里。
in the company.
没错。
Exactly.
当这些私人文件公之于众时，我们会再邀请你上节目，进一步聊聊这件事。
When those private documents become public, we'll bring you back on the show and you can talk more about it.
太好了。
Wonderful.
好的。
All right.
好吧，梅赛德斯，感谢你加入我们。
Well, Mercedes, I want thank you for joining us.
这位是梅赛德斯·本特，Premise公司的联合创始人兼合伙人，欢迎来到TI TV。
That is Mercedes Bent, co founder and partner at Premise here on TI TV.
好的。
Okay.
接下来的环节由我们的赞助商OutSystems带来，这是一个面向开发者的AI平台。
This next segment is brought to you by our sponsor, OutSystems, an AI platform for developers.
丰田、罗氏制药等公司都使用OutSystems在全公司范围内构建自定义应用和AI代理。
Companies like Toyota and Roche Pharmaceuticals use OutSystems to build custom apps and AI agents across their organizations.
我与首席执行官伍德森·马丁交谈过，了解了这家公司的业务、他如何评估企业在寻求真实AI影响力时的投资回报率，以及他对最佳编码模型竞赛的看法。
I spoke with CEO Woodson Martin about what the company does, how he evaluates ROI as businesses look for real AI impact, and also how he sees the race for the best coding models unfolding.
以下是他们的对话。
Here is that conversation.
伍德森，欢迎来到TI TV。
Woodson, welcome to TI TV.
很高兴你来到这里。
It's great to have you here.
嘿，阿卡什，很高兴能来这里。
Hey Akash, great to be here.
OutSystems 是一个面向开发者的 AI 平台，我希望你能帮助我们理解它与目前市面上其他众多正在使用 AI 的开发者平台有何不同。
So OutSystems is an AI platform for developers, and I'm hoping you can help us understand how it's different from all the other developer platforms that are out there that are working with AI that we hear so much about.
是的。
Yeah.
OutSystems 是专为企业打造的 AI 开发平台。
OutSystems is the AI development platform built for the enterprise.
像丰田汽车、罗氏制药、Axos 银行这样的公司，使用 OutSystems 快速构建自定义应用和智能代理，用于关键核心功能，同时也利用 AI 和智能解决方案现代化传统流程，并在一个统一平台上管理并治理整个应用和智能代理组合的生命周期。
So companies like Toyota Motors and Roche Pharmaceuticals, Axos Bank, they use OutSystems to rapidly build custom apps, agents, for mission critical core functions, but also to modernize kind of legacy processes with AI and agentic solutions and manage and govern that full life cycle of a portfolio of apps and agents all on a unified platform.
OutSystems 是唯一一个统一、敏捷且经过企业验证的 AI 平台。
AltSystems is the only AI platform that's unified, agile, and enterprise proven.
我们在这档节目中一直讨论的一个话题，就是企业采用 AI 时所面临的障碍。
Now one of the topics that we've been talking so much about on this show are the barriers to adoption that enterprises see with AI.
这周特别提到的一个问题是，当你甚至还不确定该测量什么、如何评估购买AI后的效果时，该如何衡量投资回报率。
One of the things that has come up this week in particular is how do you measure ROI when you're not even really sure what to measure and how to assess the effectiveness of AI after you buy it.
我想知道，当你们向客户推介平台时，公司是如何看待这个问题的。
I wonder how you think about that at your company when you're pitching your platform to customers.
是的。
Yeah.
我想说，也许从宏观层面来看，今年企业对AI的投资感觉与以往非常不同。
I would say, you know, maybe at the headline level, this year feels very different for enterprise AI investment.
随着企业逐渐从实验心态转向对实际业务影响的问责，客户思考问题的方式正在发生变化。
You know, as organizations are kind of shifting from a mindset of experimentation, toward accountability for real business impact, it's kinda changing the story in the way our customers are thinking.
通过迄今为止的实验，一个明确的结论是：我们必须预期早期会失败，并为此做好规划，因为大多数AI代理在投入生产后都会失败。
And I think one of the things that's become clear through the experimentation so far is that, yeah, kind of have to expect failure early and plan for it because most AI agents fail when they get into production.
我不是说它们会崩溃或损坏。
And I don't mean they fall down and break.
我的意思是，它们没能兑现承诺，而这其实很正常。
I mean, they don't deliver on the promise, and that's kinda normal.
要让事情完全到位，需要反复迭代、实验和大量调整。
It takes iteration, experimentation, and a lot of change to get things exactly right.
你知道，你希望AI代理完成的工作，与你真正希望依赖人工介入的工作，这在最初设计这类系统时很难预测。
You know, the work that you want AI agents to do versus the work you really want to rely on the human in the loop, that's kinda hard to predict when you're first designing one of these systems.
因此，你需要能够进行实验，这就是为什么你必须把AI当作一个系统，而不仅仅是一个模型。
And so you need to be able to experiment, and that's why you gotta really kinda treat AI as a system and not just a model.
所以你的意思是，不要指望这东西一上来就能正常工作。
So you're saying you're saying, you know, don't expect the thing to work out of the box.
我的意思是，你得经过几次使用这个工具并培训员工使用它的过程。
I mean, you know, it's gonna have to take a couple iterations of of using the tool and and training your people on the tool.
整个理念是，这些都是学习型系统，会迅速演变和变化。
The whole idea is these are learning systems, and they rapidly evolve and change.
对吧？
Right?
没错。
Right.
这种变化的一部分源于模型的学习，但更多的变化发生在企业中，因为我们正在改进上下文工程，或者正在演变用户——可能是我们的客户或员工——与系统互动的方式，因为我们正在推动变革。
Part of that change happens because the models learn, but a lot of that change happens in the enterprise because we're improving the context engineering, or we are evolving the way that users, maybe our customers or our employees are interacting with the system because we're driving change.
我们试图推动行为上的改变。
We're trying to drive behavioral change.
我们试图推动成果的改变。
We're trying to drive change in outcomes.
因此，整个系统发展得非常快。
And so this whole systems evolve fast.
但整个系统不仅仅是一个大语言模型。
But the whole system's not just the LLM.
它不仅仅是智能体的推理过程。
It's not just the reasoning of the agent.
对吧？
Right?
它还包括其他所有部分。
It's also everything else.
是数据。
It's the data.
是工作流程，是应用程序的用户体验。
It's the workflows, it's the UX of the applications.
这些可能是面向客户的移动应用，也可能是面向员工的网页应用。
Maybe those are mobile apps for your customers, maybe it's web apps for your employees.
所有这些都需要在企业中快速演进，以便迭代并实现投资回报率。
And all of those things need to be able to evolve quickly in an enterprise to be able to iterate to get toward that ROI.
因此，平台在人工智能的这一组合中至关重要。
And that's why platforms are so important as part of the mix here, in AI.
这正是我们所看到的真正差异。
And that's really a difference that we're seeing.
但需要多长时间呢？
How long though?
我们经常问人们一个问题：好吧，这需要一段时间，有些人说，广泛采用还要五年左右，对吧？
This is a question that we ask people is, okay, so it's going to take a while and we've had some people say, broad adoption is still five years away, right?
我想知道，当您与客户讨论投资回报率时，他们肯定也会问您：我应该等多久才能看到回报？
And I wonder, just when you're thinking about ROI with your clients, surely they must be asking you, how long should I wait to expect to see that?
我认为
I think
这是个很好的问题。
it's a great question.
我们对客户迅速获得实际价值的进度感到非常兴奋。
We're super excited about the pace at which customers are starting to see real value.
我们最早使用AgenTeq平台的客户之一，我们称之为Agent OutSystems工作台，是一家名为Travel Essence的公司。
One of the first customers to work with our AgenTeq platform, we call that agent OutSystems Agent Workbench, was a company called Travel Essence.
他们位于荷兰。
They're based in The Netherlands.
他们经营高端旅行服务。
They run luxury travel trips.
从项目启动到实现投资回报，仅用了三周时间，这是一个关于快速创新的惊人故事。
Three weeks from the very beginning of the project to the achievement of ROI, an incredible story of innovation at pace.
但他们真正做的是，并没有重新发明整个流程。
But really what they did, they didn't reinvent the process.
他们一直使用我们平台构建的应用程序来管理旅行规划师的工作，与客户互动以安排这些定制假期。
They have applications built on our platform that they've used for years to manage the work of their travel planners, in interacting with their customers to put together these custom holidays.
这个过程过去需要他们为每位客户花费两到三个小时来研究目的地、预订酒店、餐厅预约等，而现在他们在我们的平台上构建了一系列AI代理，后端使用了多种不同的大语言模型。
That was a process that would take them two to three hours per customer to research destinations, book hotels, restaurant reservations, etcetera, and they built a series of AI agents now on our platform that use a variety of different LLMs on the back end.
他们有一个预订代理。
You know, they've got a booking agent.
这个预订代理负责完成预订。
That booking agent makes the reservations.
他们还有一个研究代理。
They've got a research agent.
这个研究代理会探索各种场所，并根据客户提供的偏好进行匹配。
That research agent, explores the venues and matches those to the preferences of the customers they've collected.
他们将过去为每位客户制定定制假期方案需要两到三个小时的过程，缩短到了仅需三分钟。
They've taken a process that used to take two or three hours per client to develop a proposal for a custom holiday and turn that into a process that takes three minutes.
现在，这位旅行顾问只需审核一下，做些最终的调整，然后呈现给客户。
And now that travel adviser just reviews that, makes any final tweaks, and presents it to the customer.
这种显著的节省帮助他们在最初的几个月内实现了业务收入20%的增长加速。
A dramatic savings does help them achieve a 20% acceleration in growth of that business in terms of revenue in just the first few months.
对。
Right.
所以这就是快速的投资回报。
So that's rapid ROI.
所以这种情况可以发生得非常快。
So it can happen quite fast.
让我问你关于AI编程的总体情况。
Let me ask you about AI coding broadly speaking.
当我思考这些机会时，工作完成得更快，这对一线人员来说是非常明确的一个优势。
When I think about the opportunities, the work gets done faster, that's very clear, as one opportunity for people on the ground.
你对这里的风险怎么看？
What do you think about the risks here?
我的意思是，随着这些AI编码工具变得越来越普及，人们开始大量使用它们，这在某些情况下会不会令人担忧呢？
I mean, as these AI coding tools become more and more ubiquitous and people start to use them a lot, I mean, is it concerning at all in some cases at all?
我的意思是，当然会。
I mean, sure.
在大规模应用时，AgenTic AI可能会推动持续进步，也可能带来真正的运营风险，这完全取决于它是如何被有意识地设计、治理和运行的。
At scale, you know, AgenTic AI can either kinda drive sustained progress or introduce real operational risk kinda depending on how deliberately it's designed, governed, and run.
所以，我要说，今天生成代码更容易了，并不意味着生成的代码就一定好。
So, you know, I would say just because it's easier to generate code today doesn't mean that's any good.
对吧？
Right?
想象一下，这些代理是结果导向的。
Like, just imagine that these agents are outcome driven.
对吧？
Right?
想象一下，代码决定把你们客户的所有机密数据卖给你们的竞争对手。
So imagine that code decides to sell all your confidential data about your customers to your competitors.
也许它认为这是让你的业务最快盈利的方式。
Maybe it thinks that's the fastest way to turn a profit in your business.
然后也许
Then maybe
这可能是它被编码出来的方式，本质上就是这么做的，但没人真正留意过。
that's fastest way it's been that's the way it's been vibe coded essentially, and nobody was really paying attention to it.
有可能。
Could be.
对吧？
Right?
那么问题来了，我们该如何围绕这项新技术构建真正的企业级架构，以获得其好处而不承担所有这些风险？
And so the question is, how do we develop real enterprise grade architectures around this new technology so we get the benefits without all those risks?
因此，我们需要内置的防护措施，在实现规模化的同时避免混乱，而这正是OutSystems等平台真正发挥作用的地方。
So we need the built in safeguards to unlock scale without the chaos, and that's really where platforms like OutSystems help.
我们多年来一直是一个确定性的平台。
We're a deterministic platform for years.
对吧？
Right?
客户一直在构建确定性的工作流，当你输入相同的内容十次，你会得到十次相同的输出。
Customers have been building deterministic workflows where you get the same input 10 times, you'll get the same output 10 times.
现在我们有了大语言模型，而它们的本质是概率性的，这意味着你输入相同的内容十次，会得到十种不同的回答。
Now we have LLMs, and by their very nature, they're probabilistic, which means that you put the same input in 10 times, you'll get 10 different responses.
对吧？
Right?
这在推理和创意方面可能非常出色，这很棒。
And that can be great for reasoning and creativity, and that's awesome.
但它并不完全符合我们地球上最受监管的行业的需求。
But it's not exactly what we want in our most, like, say, regulated businesses on the planet.
我们实际上希望将两者结合。
We actually wanna mix.
对吧？
Right?
你希望有一些推理能力来加速决策，但向客户传达结果时，你需要确定性的流程。
You want some of the reasoning to help accelerate decision making, but you want deterministic processes for things like communicating outcomes to customers.
想象一下你是一家银行，正在向消费者发放贷款。
Imagine you're a bank, right, and you're making loans to consumers.
这是一个受到严格监管的行业。
That's a well regulated industry.
你不希望客户关于贷款申请被拒原因的回复，是由一个完全自主的LLM生成的。
You don't want the responses to your customer about why their loan application was denied written by an LLM in a fully autonomous process.
这可能会给你带来法律风险。
That could create legal risk for you.
当你想到贷款时，你想要的是正式的信函。
When you think of loan, you you wanna form letters.
对吧？
Right?
我问你一个问题，关于我们现在看到的这些不同的编码模型。
Me ask you a question about all these different coding models then that we see coming out.
我的意思是，每个月都有新的工具让人去玩，对吧？
I mean, every month there's just there's a new toy for people to play with, right?
你看到各种基准分数上下波动。
And you see the benchmarks moving up and down.
人们都在谈论谁在领跑这场比赛。
You see people talking about who is winning the race.
考虑到你与所有这些厂商集成，而且你的许多客户也都在使用它们，你如何看待这些模型相关的讨论在长期的发展趋势？
How do you see the conversation around these models playing out in the long term given that you integrate with all these players and a lot of your customers do so many of them?
是的。
Yeah.
我的意思是，我们关注的一点是，对客户选择的模型保持中立。
I mean, one of the things we're focused on is, like, being agnostic to the model that customers choose.
因此，在客户基于我们的平台构建的智能体系统中，通常会在一个系统中协调四个或五个智能体。
So in the agentic systems customers build on our platform, it's typical to have four or five agents being orchestrated in a single system.
而且这些智能体经常使用针对不同应用场景优化的不同大语言模型。
And often those agents are using different LLMs that are optimized for the use case.
我们开始看到一些客户避免使用某些大语言模型，转而选择更针对特定任务、成本更低的小语言模型。
And we're starting to see customers actually avoid LLMs for some of them and go to small language models that are more tuned to the specific task at a lower token cost.
对吧？
Right?
因此，我们认为客户会希望不断演进并更换他们使用的模型。
And so we think that customers and we see that customers are going to want to evolve and change the models they use.
你知道的。
You know?
当谷歌发布新的Gemini版本，或者OpenAI推出新的ChatGPT版本时。
As Gemini releases a new Google releases a new version of Gemini or OpenEye comes out with a new version of ChatGPT.
你希望能够尝试这些新模型，无需重新设计整个AgenTex系统，只需进行热替换、微调和优化。
You wanna be able to play with that and hot swap those models without reinventing the entire AgenTex system and just tune and optimize.
而我们通过允许您使用自己的模型、自己的令牌，并将我们的平台用作编排工具，实现了这一点。
And that's what we're making possible by allowing you to bring your own model, bring your own tokens, and essentially use our platform for orchestration.
因此，我们坚信，这个世界将演进到这样一个阶段：真正重要的是平台和编排能力，而不是具体的模型。
So we really think that this world evolves to the point where what matters is the platform and the orchestration and not the specific model.
对。
Right.
好了，伍德森，感谢你参加我们的节目。
Well, Woodson, I wanna thank you for joining us on the show.
我们很快再和你聊。
We will talk to you again very soon.
这位是OutSystems的首席执行官伍德森·马丁，正在TI TV上。
That is Woodson Martin, CEO of OutSystems here on TI TV.
谢谢，阿卡什。
Thanks, Akash.
好的。
Okay.
今天的节目就到这里。
That does it for today's show.
提醒一下，我们周一至周五每天上午10点（太平洋时间），下午1点（东部时间）直播。
A reminder, we are on the stream Monday through Friday at 10AM Pacific, 1PM Eastern.
感谢大家的参与。
I want to thank you all for joining us.
我们非常感谢你们的观看。
We really do appreciate your viewership.
我已经迫不及待期待明天的下一期节目了。
I'm already excited for our next show tomorrow.
祝你们周三剩下的时间愉快。
Have a great rest of your Wednesday.
暂时再见了。
Bye bye for now.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。