OpenAI与亚马逊就100亿美元投资进行谈判，Waymo巨额融资，IPO分析 | 2025年12月17日

欢迎各位收看Informations TI TV。

Welcome everyone to the Informations TI TV.

我叫阿卡什·佩斯里查。

My name is Akash Pesritcha.

今天是12月17日，星期三。

It is Wednesday, December 17.

我们为大家准备了一场内容丰富的节目，包含独家报道。

We have got a jam packed show for you full of exclusive reporting.

信息频道率先报道，OpenAI正与亚马逊洽谈，拟筹集至少100亿美元，并使用其芯片。

The information is first to report that OpenAI is in talks to raise at least $10,000,000,000 from Amazon and use its chips.

我们将为您深入解析这条独家新闻背后的记者。

We'll break it all down with the reporter behind that scoop.

我们还将与我们的风险投资部门副主编对话，她独家披露了Waymo正在早期洽谈筹集数十亿美元的消息。

We're also speaking with our deputy bureau chief of venture capital about her scoop that Waymo is in early talks of raising billions of dollars.

接下来，OpenAI和Anthropic正在探索收购专业数据池，两者都在为下一轮AI训练做准备。

Next up, OpenAI and Anthropic are exploring acquisitions of specialized data pools as both prepare for the next wave of AI training.

我们还将关注华尔街如何为一系列大型IPO做准备，这些IPO可能包括OpenAI和Anthropic。

And we'll also look at how Wall Street is gearing up for a run of mega IPOs, which could include OpenAI and Anthropic.

最后，我们将回顾我们的每周财经专栏。

And we'll end with a look at our weekly finance column.

内容非常丰富，让我们马上开始。

There is a lot to get to, so let's get right on into things.

《信息》独家报道，亚马逊正在洽谈向OpenAI投资100亿美元或更多。

The information has exclusive reporting that Amazon is in talks to invest $10,000,000,000 or more in OpenAI.

现在加入我们的是报道这一消息的记者之一，阿尼萨·加尔迪齐。

Joining me now is one of the reporters who broke the story, Anissa Gardizzi.

阿尼萨，欢迎再次做客我们的节目。

Anissa, welcome back to the show.

很高兴你来到这里。

It's great to have you here.

谢谢，阿卡什。

Thanks, Akash.

昨晚可真忙啊。

It was a busy night.

确实是个忙碌的夜晚。

It was a busy night indeed.

这可是件大事。

This is a big deal.

能给我们讲讲你的报道中发现了什么吗？

Walk us through what you learned in your reporting.

我们从消息源了解到，亚马逊正在洽谈向OpenAI投资100亿美元或更多，这笔交易还可能涉及OpenAI使用亚马逊的Tranium芯片，而不仅仅是目前使用的英伟达GPU。

So we learned from our sources that Amazon is in talks to invest $10,000,000,000 or more into OpenAI in a deal that could also see OpenAI using Amazon's Tranium chips in addition to the GPUs from NVIDIA that OpenAI uses.

这将是亚马逊对OpenAI的一笔重大投资。

So this this would be a major, you know, investment by Amazon into OpenAI.

过去，亚马逊一直与Anthropic关系密切。

And Amazon historically has tied itself to Anthropic.

但正如我们在过去几周看到的，Anthropic已经与微软和谷歌达成了重大交易。

But as we've seen in the past couple of weeks, Anthropic has done big deals with Microsoft and Google.

所以亚马逊也在与OpenAI建立另一个联盟。

So Amazon is also forming another alliance with OpenAI.

我想进一步深入探讨这一点，但我一直在统计OpenAI目前与哪些公司有合作。

So I do want to dig into this a little further, but I was counting all the companies that OpenAI has deals with now.

现在真正的问题是，还有谁没有和OpenAI合作？

And I think the question now really is who doesn't OpenAI have a deal with?

因为他们已经和NVIDIA、AMD、博通达成了合作。

Because they have the deal with NVIDIA, AMD, Broadcom.

他们也和亚马逊在芯片方面达成了合作。

They've got Amazon for the chips.

我想他们目前还没有和谷歌的TPU达成合作。

I guess they don't have a Google TPU deal just yet.

Meta有，或者至少根据我们的报道是这样。

Meta has that, or at least according to our reporting.

天啊，OpenAI现在简直在和所有人合作。

Gosh, OpenAI is really just they're using everybody at this point.

是的，没错。

Yeah, that's right.

亚马逊是最大的云服务提供商。

And Amazon is the largest cloud provider.

所以，你知道，他们显然在环顾四周，看到OpenAI预计未来将在云计算上投入数百亿美元。

So, you know, they obviously are looking around the room and seeing that OpenAI is projecting that it's going to be spending, you know, potentially hundreds of billions on cloud computing in the future.

因此，我认为最大的云服务提供商想要分一杯羹是很合理的。

So I think it makes sense that the largest cloud provider does want a piece of that.

我觉得这其实早就该发生了。

And this is just a long time coming, I think.

所以这也是我们之前讨论过的那种循环交易之一。

And so this is also one of these circular deals that we've been talking about.

虽然在这里亚马逊既是芯片供应商又是云服务提供商，而有时我们看到这些角色是分开的。

Although Amazon here would be both the chip player and the cloud player, whereas sometimes we've seen that spread apart.

但这又是另一种供应商融资交易，对吧？

But it's another one of these vendor financing deals, right?

OpenAI 实际上是在拿钱付钱给亚马逊。

Where OpenAI is essentially getting paid to pay money back to Amazon.

是的。

Yeah.

正如你可能记得的，就在上个月，亚马逊和 OpenAI 宣布了一项战略合作，OpenAI 表示在未来几年内，将花费约 380 亿美元租用亚马逊的服务器。

As you might remember, just, you know, last month, Amazon and OpenAI announced a strategic partnership in which OpenAI said that over the next couple years, it's gonna spend somewhere around 38,000,000,000 on renting Amazon servers.

当然，这笔交易仅涵盖了来自英伟达的 GPU。

Of course, that and that deal only covered GPUs from NVIDIA.

所以，没错。

So yeah.

现在，如果 OpenAI 能从亚马逊获得一笔现金注入，这可能会帮助它们支付云服务账单，包括亚马逊的云账单。

Now now if OpenAI were to get a cash injection from from Amazon, it might help them pay their cloud bills, including their Amazon cloud bill.

你认为微软对此真的感到不满吗？还是说，到目前为止，这已经是预料之中的事了？

Do you think Microsoft is really upset about this or was this kind of expected in par for the course at this point?

我认为，既然我们早就预计到这种情况会发生，微软肯定也看到了风向，同样有所预期。

I think since we expected that this would happen, I'm sure Microsoft also saw the writing on the wall and was expecting it too.

亚马逊是其最大的云竞争对手，所以我相信这对他们来说并不是理想的情况。

Amazon is its largest cloud competitor so I'm sure this isn't an ideal scenario for them.

但如果这笔交易达成，微软仍然保持巨大的优势。

But Microsoft does still maintain a large advantage if this deal gets done.

微软是唯一被允许向云客户销售OpenAI模型的云公司。

Microsoft is the only cloud company that's allowed to sell OpenAI models to cloud customers.

即使OpenAI与其他云服务商达成了合作，微软仍保有独家销售权。

And that's they've maintained the exclusive right to do so even as OpenAI has been able to sign other cloud deals.

因此，这对微软来说无疑是一个转机。

So there definitely is a silver lining for Microsoft.

还有我想提到的另一个关联点，你前面也提到了，亚马逊与Anthropic合作得非常紧密。

Well, and the other link that I'm thinking about here, and you mentioned this up top, but Amazon has worked so closely with Anthropic.

我们知道，Anthropic在帮助亚马逊推出其Trainium芯片方面发挥了重要作用。

And we know that Anthropic has been big in helping Amazon get its Trainium chip out into the market.

我的意思是，我看Dario和Daniela说，我以为你已经拿到了Trainium的测试资格。

I mean, I'm looking at Dario and and Daniela saying, I I thought you were given Tranium test.

我的意思是，我知道大家都在和各方合作，但OpenAnthropic在这里是竞争对手。

I mean, I know everyone's working with everybody, but OpenAnthropic are rivals here.

所以我肯定安迪·贾西和达里奥之间有过一场艰难的对话。

And so I I'm sure there was maybe a tough conversation there between Andy Jassy and Dario.

是的。

Yeah.

我认为从安迪·贾西的角度来看，他非常需要一家领先的模型公司至少公开表示在使用Trainium芯片。

I think from Andy Jassy's perspective, he really needs a leading model company to at least publicly talk about using Trainium chips.

而且我们知道AWS与Anthropic已有现有合作。

And and we know that AWS has that existing deal with with Anthropic.

但最近几周，我们听到和爆出的所有新闻都是关于Anthropic在使用谷歌的TPU。

But in recent weeks, you know, all the news that we've been hearing and breaking are are about Anthropic using Google's TPUs.

没错。

Right.

对于安迪·贾西来说，这时候打电话给萨姆·阿尔特曼讨论训练芯片，时机似乎相当不错。

It's like seems like pretty good timing for Andy Jassy to be calling Sam Altman about training him chips.

这简直太疯狂了。

It's actually nuts.

我觉得我们应该有一个矩阵，列出所有公司之间的合作关系，也许你家的白板上已经画好了。

I feel like we should have, and maybe you have this on your whiteboard at home already, we should have a matrix of all the companies that work with all the companies.

作为记者，如果两家公司之间没有关联，那你就应该提出这个问题。

And as a reporter, if there is not a link tying two companies, that should be the question that you're asking really.

没错。

Yeah.

亚马逊和OpenAI之间发生的情况本质上就是这样。

That's essentially what happened here with Amazon and OpenAI.

所以这个关系矩阵变得越来越复杂了。

So the matrix is getting even more and more complicated.

在你走之前，我问你一个问题。

Let me ask you one question before you go.

关于这个故事，你现在还有什么问题要问吗？

What questions do you have now going forward here about the story?

是的。

Yeah.

我们从消息源那里了解到，亚马逊的这笔投资可能会触发OpenAI更大规模的融资行动。

One thing we heard from our sources was that this investment by Amazon could kick off a broader financing effort that OpenAI does.

所以我认为，在假期来临之前，我们肯定会就此展开跟进采访。

So I think heading into the holidays, that is what we will be making calls on for sure.

很好。

Great.

好了，阿尼萨，这真是个重磅新闻。

Well, Anisa, it was a big story.

恭喜你挖到这条独家消息，期待你后续有更多进展时再来做客。

Congrats on the scoop and look forward to having you on once again as you learn more.

这位是《信息》杂志负责云计算和计算领域的记者阿尼萨·加尔迪齐。

That is Anisa Gardizi, our cloud and compute reporter here at The Information.

好的。

Okay.

《The Information》独家报道，Waymo正在与潜在投资者进行早期洽谈，旨在以至少1000亿美元的估值筹集更多资金。现在邀请到为我们揭发这一消息的风险投资部门副局长凯蒂·鲁夫。

The Information has exclusive reporting that Waymo is in early talks with potential investors to raise more money at a valuation of at least $100,000,000,000 Joining me now is our deputy bureau chief of venture capital, Katie Roof, who broke that story.

凯蒂，欢迎再次做客我们的节目。

Katie, welcome back to the show.

很高兴你来到这里。

It's great to have you here.

很高兴能来。

Good to be here.

这些关于Waymo的新谈判进行到什么阶段了？

So how far along are these new Waymo talks?

目前还处于早期阶段，但他们已经与投资者展开了对话。

So it's somewhat early, although they have already been in conversations with investors.

他们尚未确定确切的估值，但知道至少将达到1000亿美元。

They have not decided the exact valuation, but they know that it's going to be at least a 100,000,000,000.

要知道，Waymo部分由谷歌的母公司Alphabet持有。

Know, Waymo is partially owned by Alphabet, the parent of Google.

而且他们你知道吗？

And they you know?

但他们也有现有的外部投资者，比如安德森·霍洛维茨，他们还会继续吸纳更多投资。

But they also have existing outside investors such as Andreessen Horowitz, and they'll be taking more.

这一轮融资与公司以往的融资轮次相比如何？

And how does this round compare to past funding rounds that the company has done?

估值更高了。

Well, the valuation is higher.

比上一轮高出一倍多。

It's more than double the the last round.

去年我在前东家报道了450亿美元的融资，因此他们希望抓住这一势头。

I broke the $45,000,000,000 last year at my former employer, and so they are hoping to capitalize on this momentum.

他们不仅在旧金山、洛杉矶和其他城市扩大了乘客规模，还获得了多项监管批准，有望进一步扩展。

They have not only expanded their ridership in San Francisco, Los Angeles, and other cities, but they have also gotten a lot of regulatory green lights where they can can potentially expand further.

今年他们还计划新增几个城市，我认为他们希望投资者能基于这一势头进行投资。

They also have several new cities they're adding this year, and I think they're hoping that investors will invest based on that momentum.

你认为公司会用这笔钱做什么？

What what do you think the company is gonna do with all the money?

这是一项成本很高的业务，你知道，需要大量的研发用于这些汽车以及安全协议。

Well, it's a costly business, you know, to for a lot of r and d for the these the cars and, you know, safety protocols.

据我所知，他们目前还没有盈利。

And so I think that they, from what I understand, are not yet profitable.

所以，为了持续扩张，他们必须继续投入资金。

So they have to to keep that you know, to to expand, they have to keep spending.

但他们并没有披露任何具体计划。

But, you know, they haven't disclosed anything in particular.

你知道，公司融资有时是为了招聘，但有时也是为了收购。

You know, sometimes when companies raise money, it's often for hiring, but sometimes it's for acquisitions.

凯蒂，我可以问你一下，你以前坐过Waymo吗？

Can I ask you, Katie, have you taken a Waymo yet before?

我在这里坐过一次。

I took one here.

你在这里坐过一次？

You took one here?

我无意中向司机道谢了，然后我想，等等。

Accidentally, I thanked the driver, and I was like, wait.

没有司机。

There's no driver.

你来这里坐Waymo，就是为了讨论你乘坐的那辆车背后的公司即将融资，可你连个司机都没法问。

You took one here to discuss the scoop that the company behind the vehicle that you were in is gonna raise more money, and you didn't even have a driver to tell.

我的意思是，有司机在场总归是好事。

I mean, you know, it would have been nice to have a driver there.

所以你知道，我知道一些你老板们不知道的事。

So you know, I know something that, you know, your bosses don't know I know.

上帝说，好吧。

God says, Okay.

那么，我问你，你在这里乘坐Waymo时，是用Waymo应用叫的车吗？

Well, and let me ask you, when you took the Waymo here, did you take it with the Waymo app?

你用的是另一个网约车应用吗？

Did you take it with another ride sharing app?

是的，我用的是Waymo应用，但他们也与Uber合作。

Yeah, I took it with the Waymo app, but they also partner with Uber.

你知道，他们并不是唯一可用的无人驾驶网约车服务。

And, you know, they're not the only ride sharing, driverless ride sharing service available.

特斯拉正处于推出自动驾驶出租车服务的初期阶段。

Tesla's in the early phases of rolling out its robo taxi service.

然后，你知道，我们在旧金山的办公室经常看到Zooks车经过。

And then, you know, our office in San Francisco, we always see the Zooks going by.

它们是亚马逊旗下的。

They're owned by Amazon.

Waymo当然在这里拥有主导的市场份额和先发优势，但其他人最终能否赶上还有待观察。

You know, it's Waymo certainly has the dominant market share here and the early mover advantage, but it remains to be seen whether others will eventually catch up.

那么，我有一个快速而重要的问题：Waymo如何看待它与这些网约车服务在未来的关系演变？

Well, so one of the quick big questions I have is how Waymo sort of sees its relationship with these ride hailing services evolve in the future.

因为正如我们所说，他们拥有Waymo应用。

Because as we said, they have the Waymo app.

他们还有这些合作关系。

They also have these partnerships.

我有点好奇他们如何看待这种分工。

I'm sort of wondering how it thinks about that split.

关于这个业务的未来，你还有什么其他需要了解的问题吗？

What other reporting questions might you have about the future of this business?

是的，我认为当我与投资者交谈时，有些人非常希望投资，因为他们看到了这里的势头，也看到了未来的前景。

Yeah, well, I think when I talk to investors, some of them are really, really hoping to invest because of the momentum and because of the, you know, they see the future here.

但你知道，目前他们的估值看起来非常高。

But, you know, right now it looks like a very, very high multiple that they're going to be getting.

有估计称他们的收入约为3亿至3.5亿美元。

There's estimates out there that their revenue is like 300, maybe $350,000,000.

因此估值超过1000亿美元。

And so north of 100,000,000,000 valuation.

我的意思是，你们确实为这家公司注入了大量预期。

I mean, they're really you really are baking in a lot of expectations for this company.

所以我认为一些投资者想知道，他们能否保持这种势头？

And so I think some investors want to know, will they be able to keep up this momentum?

而且，你知道，它们的单位经济效益是否可持续？

And, you know, are the unit economics sustainable?

也就是说，它们目前是否在亏损？能否最终改善到实现盈利或接近盈利？

Meaning, are they losing, you know, can they eventually improve to the point where they are at profitability or close?

对。

Right.

很好。

Great.

好了，凯蒂，这真是个大新闻。

Well, Katie, it was a big scoop.

恭喜。

Congrats.

谢谢你参加节目，期待很快再和你交谈。

Thanks for coming on the show, and look forward to talk to you again very soon.

谢谢。

Thank you.

好的。

Okay.

随着科技行业等待SpaceX、Anthropic和OpenAI在未来几年的三次重大IPO，银行家们正试图找到方法，限制因私人和早期融资金额巨大而导致的大量股份抛售。

As the tech sector awaits three big IPOs in the coming years from SpaceX, Anthropic, and OpenAI, bankers are trying to find ways to limit mass selling of shares given how much money has been raised privately and early.

现在邀请我们的《The Information》交易记者瓦利塔·波伊来讨论这一困境。

Joining me now to discuss the dilemma is Valita Poe, our deals reporter at The Information.

瓦利塔，欢迎再次回到节目。

Valita, welcome back to the show.

很高兴你能来。

It's great to have you here.

很高兴能再次来这里。

It's nice to be here again.

那么，银行家们面临的困境到底是什么？

So what exactly is the dilemma here that bankers are facing?

因为像SpaceX、Anthropic和OpenAI这样的科技公司已经筹集了大量私人资本，并且它们希望在未来十八个月内通过潜在的IPO进一步融资。

So because tech companies like SpaceX, Anthropic and OpenAI has raised so much like private capital and they want to raise more in the potential IPOs that could come in the next eighteen months.

这些公司都拥有庞大的股东群体，以及在成立时设立的特殊目的载体。

And these companies all have like large shareholder bases and special purpose vehicles that created in their formation.

因此，银行家们正在思考如何逐步释放股份的出售。

That's why bankers are trying to see how do they phase out, like, you know, the selling of the shares.

通常，IPO过程会有九十天或一百八十天的标准禁售期，但银行家们正在排除这种做法，转而考虑如何分阶段释放股份，比如每二十天或三十天释放一批，如何更长时间地安排这些释放？

So usually an IPO process is when companies have standard lockup period of like ninety or like one hundred eighty days and bankers are largely ruling that out and thinking about how they would do like staggered release, like, you know, maybe every twenty or like thirty days, how can we space it out for like longer and for like structuring those releases?

这就是他们正在与这些公司讨论的问题。

So this is what they're thinking about, like, with the companies.

这其实很有趣，因为我从未见过如此多的IPO公司或潜在IPO公司筹集了如此巨额的资金。

Well, and this is kind of interesting because never before do I feel like we've seen a slate of IPO companies, or IPO hopeful companies, that have raised so much money.

我的意思是，它们拥有大量员工和巨额资本，对吧？

I mean, they have so many employees, they have so much capital, right?

我的意思是，我们看到OpenAI每两个月就进行一次融资。

I mean, we see OpenAI as fundraising every two months.

那么，我们这里讨论的大型银行是哪些呢？

Hear a Which new are the big banks that we're talking about here that are at the center of the story?

通常，摩根士丹利和高盛主导着科技公司的IPO业务，任何一家大型科技公司的上市都由它们其中一家牵头。

So usually, like Morgan Stanley and Goldman Sachs dominate the Emtek IPO business, and one or the other kind of led all the major tech listings.

当然，其他所有银行也想参与IPO，但主导科技IPO业务的基本上还是这两家银行。

Of course, like all the other banks may want to get a seat in the IPO as well, but mostly it's the two banks that dominated the tech IPO business.

所以我很想知道，对于面临这一挑战的银行来说，它们有哪些不同的选择。

So I'm curious, for these banks who are faced with this challenge here, they have a number of different alternatives.

它们可以像你提到的那样，采取分批释放的方式。

They can do this staggered release, as you talked about.

从银行的角度来看，这样做有什么好处吗？

Is there anything to gain from the bankers' end?

如果操作得当，它们能获得更多的费用吗？

Do they get more fees here if they get it right?

我的意思是，我甚至都没研究过这些银行的薪酬结构，但它们能从中获得什么好处呢？

I mean, I haven't even studied how the compensation for these banks worked, but what do they stand to gain?

对于承销IPO的银行来说，它们实际上会获得所筹资金单个位数百分比的份额。

So for like the banks who are underwriting the IPOs, they will actually get like, you know, a single digit percent share from like, you know, how much money they're raising.

例如，如果你进行一笔500亿美元的IPO，银行如果拿到3%的费用，那就是15亿美元的收入，用于运作这场巨大的IPO。

So for example, if you're raising a 50,000,000,000 IPO at bankers, if they get 3%, that's like, you know, 1,500,000,000.0 in the fees they can get for, running this huge IPO.

这对银行来说是一笔巨额资金的博弈。

So that's lots of money at stake for the bankers.

这同时也是它们难得的历史性IPO机会。

This is also like a historic IPO's opportunities for them.

那么，它们的报酬是否与IPO后几天内股价的表现有任何关联呢？

And then does it typically at all does their compensation matter with respect to what happens to the stock price in the days after the IPO at all?

我认为它们只拿承销费。

They just get the underwriting fee, I think.

但这确实是一个两难的困境。

But it certainly is a dilemma.

根据你所看到的一切，你觉得这件事可能会走向哪里？

Wonder, based on everything you're seeing, do you have any sense on where this might go?

你有信心他们会做对吗？

Are you confident that they'll get it right?

我的意思是，银行家们似乎相当有创新精神。

I mean, bankers, they seem to be pretty innovative.

对吧？

Right?

科技领域的IPO圈已经见证了从拍卖到直接上市、再到可以募资的直接上市等多种创新。

The tech IPO circles have seen lots of innovations from the auctions and into direct listing, and direct listing where you can raise money.

所以我相信银行家们会提出最全面或最具创新性的解决方案，以应对这些公司可能面临的大问题。

So I'm sure bankers would try to come out of like the most comprehensive or innovative solutions to prevent whatever big issues may face these companies.

很好。

Great.

瓦莉塔，这是一个很棒的故事。

Well, Valita, it was a great story.

帮个忙，再留一分钟。

Do me a favor, stick around for one minute.

我们马上回来，进入下一个环节。

We're going to bring you right back for our next segment.

如今，似乎你可以向聊天机器人问任何问题，它都会给出答案，提供对海量知识的即时访问。

These days, it seems like you can ask a chatbot almost anything, and it will serve up an answer, offering instant access to a vast range of knowledge.

但根据相关报道，这些机器人能够抓取的公开数据资源正开始枯竭。

But according to reporting from the information, the well is starting to run dry in terms of publicly available data that these bots can scrape.

因此，像OpenAI、Anthropic和谷歌这样的公司正在与他们认为能填补特定知识空白的企业洽谈数据合作。

And so companies like OpenAI and Anthropic and Google are discussing data deals with companies they feel might be able to fill specific knowledge gaps.

接下来为我们解释这些合作的是我们的财经记者瓦莉塔·包和我们的AI记者斯蒂芬妮·帕拉佐洛。

Here to here to explain that are our deals reporter, Valita Pao, and our AI reporter, Stephanie Palazzolo.

欢迎回来，瓦莉塔。

Welcome back, Valita.

欢迎，斯蒂芬妮。

Welcome, Steph.

瓦利塔，我想先从你开始。

Valita, I wanna start with you.

OpenAI、Anthropic 和 Google 正在与哪些公司接洽？

Which companies are OpenAI, Anthropic, and Google starting to speak to?

我们听说 OpenAI 的员工曾与生命科学公司如 Reviti 或 Siro 接触过，后者是一家会计软件公司。

So we have heard that staff at OpenAI have spoke to life science companies like Reviti or Siro, which is an accounting software companies.

他们还广泛接触了生物技术公司、消费医疗领域，以及软件或金融服务领域的初创公司或企业，以试探是否存在潜在合作机会。

And they also have mass outreach to kind of biotech companies, consumer healthcare, or like software or financial services, startups or companies just like to feel out if there's a potential opportunity there.

这些公司中有任何一家同意出售它们的数据吗？

And are any of those companies, are they agreeing to sell their data at all?

目前情况如何？

What's the status?

我认为大多数公司都拒绝了，因为它们拥有独特且至关重要的数据。

I think most of the companies have said no because they have such priority data that are unique to them.

这些数据在隐私权和使用权方面存在许多复杂问题。

And there's lots of complicated issues regarding privacy rights or usage rights of those data.

所以我认为公司通常倾向于拒绝。

So I think companies are usually leaning against no.

斯蒂芬，这看起来显然是个答案，对吧？

Steph, it feels like that's an obvious answer here, right?

我的意思是，为什么一家拥有这些数据的公司会愿意放弃呢？我们之前已经讨论过这一点。

I mean, why would a company sitting on this data and we've talked about this.

比如Shopify，它们掌握着OpenAI和聊天机器人公司所需的所有数据。

I mean, like Shopify, for example, they're sitting on all this data that OpenAI and the chatbot companies would need.

这可是皇冠上的明珠，对吧？

That's the crown jewel, right?

我看不出任何一家公司会真正愿意放弃这些数据的情景。

I don't see any scenario why a company would actually give it up.

是的。

Yeah.

我的意思是，对很多公司来说，总有一个价格能让它们动心。

Well, I mean, I think for a lot of companies, there is always a price that they might become interested in.

所以在某些情况下，可能只是因为价格合适，公司才会考虑这个想法，但他们的要价可能高于AI实验室愿意支付的水平。

So in some cases, it might just be that for the right price, the company would entertain the idea, but their price might just be higher than what the AI labs are kind of willing to pay.

我认为另一个可能的情况是，在某些情况下，公司可能会关闭。

I think another thing that could be going on too is in some cases, companies might be shutting down.

我今天早上写到，有一些公司基本上正在关闭，它们有兴趣把代码库卖给AI实验室。

And so what I wrote about this morning is there are some companies that are basically shutting down that are interested in selling their code bases to AI labs.

所以在这些情况下，它们实际上并没有放弃太多，因为无论怎样，它们都会关闭。

And so in those cases, they're not really giving up much because they would have, you know, shut down either way.

我们是否知道在这种数据市场中，一个代码库能卖多少钱？

Do we know anything about how much a code base might fetch in this data market?

是的。

Yeah.

所以我们并不确切清楚。

So we're not exactly sure.

所以目前的情况是，有一些像Terrain或Afterquery这样的数据标注和整理初创公司。

So kind of what's going on is there's a couple of these data labeling curation startups like Terrain or Afterquery.

这是另一个较新的公司，我们之前没怎么写过。

It's another newer one that we haven't written about as much.

他们基本上是在寻找那些失败的初创公司，或者已经关闭的公司，试图购买它们的代码库。

And they are basically going out to these kind of failed startups basically or sort of sort of closed closed down and asking to buy their their code bases.

所以我们不确定他们之后会以什么价格把这些代码库转卖给AI实验室。

So we're not sure how much they then turn around and sell these code bases for to the AI labs.

但根据我们今天早上的报道，我们了解到他们会以每份代码库1万到5万甚至6万美元的价格收购。

But from our story from from this morning, we've heard that they will buy the code bases anywhere from, you know, 10,000 to 50,000 or $60,000 per code base.

平均下来，大约是数万美元的水平。

On average, kind of the tens of thousands of dollars.

对。

Right.

瓦莉塔，这些数据交易中存在隐私问题吗？

Valita, are there any privacy concerns here with these data selling deals?

我的意思是，我们知道我们的数据无处不在，但他们是如何应对这个问题的？

I mean, we know that our data is everywhere, but how are they approaching that?

他们通常处于并购讨论中，比如，公司可能会有兴趣直接收购一家公司，以获得这些数据，但用户层面有不同的隐私协议，可能不允许公司按其预期方式使用这些数据。

They're usually, like, in M and A discussions, like, you know, companies would be interested in, like, you know, buying a company outright because they would have access to those data, but there's different privacy agreements, like with the users that may not allow, you know, companies to use the way they may intend to.

例如，当人工智能实验室希望使用某家公司数据来训练模型时，他们可能并不具备相应的授权。

For example, like when the AI lab wants to use the data in a company that they want to train their models on, they may not necessarily have that permissions.

如果涉及生物科技或医疗公司以及患者的_data_或医疗记录，监管会更加严格，甚至可能受到联邦法规的约束。

And if it's including like a biotech or healthcare companies and patients' data or medical records are involved, it's much more rigorous and it could be federally regulated.

因此，关于他们能共享多少、能使用多少、以及使用这些数据的参数，情况相当复杂。

So it's kind of complicated in terms of like, you know, how much they can share, like how much they can use and what the parameters they might be able to use those data.

所以，如果这些复杂的交易真的发生，情况会很复杂。

So these complicated deals, if they would happen.

对。

Right.

斯蒂芬，当你思考未来与此相关的报道问题时，有哪些主要问题浮现在你脑海中？

And, Steph, as you think about the reporting questions here moving forward, what are the big questions that come to mind for you?

而我想到的一个问题是：这有多昂贵？

And one that comes to mind for me is how expensive is it?

作为模型开发者，自己收集这些数据需要多长时间？

How long would it take to generate that data, to collect that data on your own as a model maker?

在什么情况下，购买数据比自己构建更划算？

When does it become better to buy versus to build?

你正在考虑哪些问题？

What sorts of questions are you thinking about?

是的。

Yeah.

关于这一点，有很多方面要考虑。

I mean, lots of things on that.

我认为在某些情况下，像OpenAI这样的AI实验室自己生成这些数据的成本非常高。

I think in some cases, yes, it's very expensive for an AI lab like OpenAI to come up with this data on their own.

但在其他情况下，他们可能根本无法生成出与现实世界相匹配的、真实的自有数据集。

But in other cases, it might just be impossible for them to come up with their own realistic data sets that kind of match what you see in the real world.

因此，在医疗或生物技术等专业领域，OpenAI的研究人员虽然非常聪明，但他们未必是生物技术科学家或医生。

So in niches like healthcare or biotech, OpenAI researchers are super smart, but they're not necessarily biotech scientists or doctors.

所以无论他们做什么，我认为他们自己很难生成这类数据。

And so no matter what they do, I think it's difficult for them to come up with that sort of data by themselves.

因此，对于这类数据，他们可能不得不通过收购途径，从现有公司购买这些数据。

And so for that sort of thing, they might just have to go through the acquisition route of trying to buy that data from existing companies.

我认为这引发了我的一些问题，其中一个问题是：未来，AI实验室最感兴趣的是哪些类型的数据？这些数据能告诉我们他们未来的产品重点是什么？

I think in terms of questions that this raises for me, I think one question is just, you know, moving forward, like what types of data are the AI labs most interested in and what can that tell us about their product focuses moving forward?

所以，如果他们对消费医疗或生物技术数据有特别的兴趣，这是否意味着像OpenAI、谷歌或Anthropic这样的公司将来会进入这些领域？无论是向普通消费者销售医疗应用，还是向医生或科学家销售工具。

So, if there is this kind of special interest in consumer healthcare or biotech data, does that mean that companies like OpenAI or Google or Anthropic will want to get into those spaces later, whether that means selling, you know, healthcare apps to everyday consumers like us or selling tools to doctors or scientists.

因此，我认为这在某种程度上让我们提前窥见了AI实验室接下来可能进入的产品或行业。

And so I think it kind of gives us like a sneak peek into what types of products or industries AI Labs might want to move into next.

对。

Right.

好了，瓦莉塔和斯特凡，感谢你们的到来。

Well, Valita and Stefan, I want to thank you for coming on.

这位是我们的交易记者瓦莉塔·潘，这位是我们在《The Information》的AI记者斯蒂芬妮·帕拉佐洛。

That is Valita Pao, our deals reporter, and Stephanie Palazzolo, our AI reporter here at The Information.

好的。

All right.

我们刚刚讨论了明年可能迎来的IPO热潮。

Well, we just talked about the avalanche of IPOs that could be coming next year.

今天，在我们的每周财经专栏中，我们的编辑肯·布朗探讨了这一趋势可能对这些公司整体估值产生的影响，我想请他来详细谈谈。

Today, in our weekly finance column, our editor Ken Brown discusses how that could have implications on the valuations of these companies, broadly speaking, and I want to bring him on to talk all about it.

肯，欢迎再次做客我们的节目。

Ken, welcome back to the show.

很高兴你来这里。

It's great to have you here.

嗨，阿卡什。

Hi, Akash.

你今天写的专栏核心观点是什么？

So what was the thrust of the column that you wrote today?

核心观点是一个小小的警告，即聚焦于稀缺价值。

The thrust was a little bit of a warning, which is it focuses on scarcity value.

所以，这些之前提到的大型私营公司，比如SpaceX、OpenAI和Anthropic，可能会上市。它们之所以价值如此之高，能够以数千亿美元的估值融资，正是因为它们稀缺。

And so the idea is these giant private companies that were talking about, might go Public, SpaceX, and OpenAI, and Anthropic, one of the reasons they're so valuable, they're raising money at these hundreds of billions of dollars valuation, is because they're scarce.

你很难进入这些公司。

You it's hard to get into them.

有一大群人会说：嘿，阿卡什，你想加入这家很棒的公司吗？

There's a whole army of people who say, you know, hey, Akash, you wanna get into this great company?

我就是那个能帮你进入这家优秀公司的关键人物。

I'm the guy who can get you into this great company.

因此，由于这些公司是私营的且难以进入，它们的估值中就包含了一种溢价。

And so there's a premium in the valuation because they're private and they're not easy to get into.

好的。

Okay.

最终，这些公司会上市。

So eventually, these companies go public.

突然间，它们的股票变得可以自由流通了。

Suddenly, their stock is free flowing.

每个人都能接触到它。

Everybody can get their hands on it.

这就是人们所说的私有与公开之间的溢价。

This is the private public premium that people talk about.

谈谈我们见过的一些例子吧。

Talk a little bit about examples where we've seen this.

你之前提到了几个，比如加密货币和房地产。

You laid out a couple of them, crypto, real estate.

这两者发生了什么？

What happened in those two?

没错。

So, right.

当这些东西进入市场后，稀缺性带来的溢价就会降低。

So the idea is when things get out there in the market, there's a less of a premium on the scarcity.

加密货币多年来，自从加密货币出现以来，你必须是加密货币交易者，拥有加密货币账户才能购买加密货币。

So crypto, for a bunch of years, for as long as crypto has been around, you had to be a crypto trader, you had to have a crypto account to buy crypto.

而且没多少人这么做。

And not a lot of people did.

我的意思是，有些人确实做了，但普通投资者通常只有股票和债券账户，也就是典型的经纪账户。

I mean, some people did, but your average investor, they have a stock market and bond market account, typical brokerage account.

所以几年前，我们有了第一只比特币ETF，随后大量ETF涌现，像MicroStrategy（现更名为Strategy）这样的公司也开始大量持有加密货币，并在股票交易所上市交易。

So then a couple of years ago, we had the first Bitcoin ETF, and then there was this flood of all these ETFs and all these companies like MicroStrategy, now called Strategy, that basically stuff themselves with crypto and trade on the stock exchange.

于是，那些通常只投资股票和债券的投资者，现在可以轻松购买加密货币了。

And so all of a sudden, all these investors who typically invest in just stocks and bonds could easily buy crypto.

因此，加密货币的稀缺性消失了。

And so, the scarcity of crypto disappeared.

那么，发生了什么？

So, what happened?

嗯，通常会出现一波繁荣。

Well, so there was a boom, which often happens.

当这些资产进入公开市场时，往往会迎来一波热潮，人人都兴奋不已，纷纷买入。

These things come on the public markets and there's a boom, everyone gets excited, everyone buys them.

然后，通常还会发生崩盘。

And then also what often happens is there's been a crash.

因此，这些资产的价格大幅下跌。

And so a lot of these things have fallen a lot.

它们总是在普通投资者兴趣达到顶峰时达到峰值，然后就崩盘了。

And they peak just at the time of maximum interest of average investors and then they crash.

所以，Strategy现在的股价已经下跌了三分之二，比特币也大幅下跌。

So strategy now is down by two thirds and Bitcoin is down a lot.

还有这些稀奇古怪的项目， literally 数以百计的也都崩盘了。

And then all these wacky ones, there's literally hundreds have crashed as well.

这已经成为我们一再看到的模式。

And so it's a pattern that we've seen over and over again.

房地产也是如此。

So real estate, same thing.

大多数房地产掌握在私人手中，而Brookfield以及一些非Brookfield、Blackstone和其他一些私募股权公司推出了这些基金，让普通人也能参与，资金大量涌入，人人都很高兴。

Most real estate is in private hands and Brookfield and some Not Brookfield, Blackstone and some other private equity firms offered these funds that got people gave people access to this and money poured in and everyone was happy.

然后就不再是这样了。

And then they weren't anymore.

下降了。

Declined.

在房地产方面，每个人都想退出。

With real estate, everyone tried to get out.

房地产是缺乏流动性的，结果他们被困在那里，不得不减记资产。

Real estate is illiquid, and it turns out they were stuck there and they had to write it down.

所以这些只是警示：一旦事物变得更加普及，或者说每个人都能购买，有时估值就会下降，因为不再具有稀缺性。

So these are just warnings that once things become more democratized or more, you know, that everyone can buy, sometimes valuation can go down just because there's not a scarcity value.

那我问你一个问题。

Well, let me ask you this.

加密货币的例子很有趣，因为一方面，当我2022年报道加密货币时，大家普遍认为，一旦ETF获批并上市，人们就会开始广泛了解加密货币，即使他们不必在交易所购买。

Crypto example is interesting because on one hand, when I was covering crypto back in 2022, the whole thing was once we get the ETFs approved and once we get them out there, I mean, then people will really start to become familiar with cryptocurrencies broadly, even if they don't have to buy it on exchange.

我的意思是，这将有助于提高认知度，进而推动价格上涨，也许人们未来也会更愿意接受它。

I mean, it will help with awareness, and that will, in turn, drive the price up, and maybe even people will become more comfortable with adopting it down the road.

你说这反而产生了相反的效果，现在它变得很容易获得了。

You're saying it's kind of has the opposite effect, which is that now it's readily available.

我的意思是，你能不能从另一个角度看待这个问题，比如加密货币就是这样走向主流的？

I mean, could you see the argument made the other way and saying, I mean, this is this is how crypto, for example, how how it goes mainstream?

是的。

Yeah.

听好了，我们不知道未来会发生什么。

Look, we don't know what's gonna happen in the future.

对吧？

Right?

有时候，随着时间推移，当人们逐渐适应之后，事情确实会如预期那样发展。

And sometimes things do play out, you know, over time when people do become comfortable.

我的意思是，你不可能预测这么长时间以来的所有情况。

I mean, you know, there's you know, certainly, you can't you're not gonna predict that all these years.

事实上，加密货币刚出现并上市时，人人都说：哦，我能靠这个赚钱。

I mean, the reality with crypto was it came out and went public and everyone said, oh, I can make money on this.

对。

Right.

我不认为当时有人认真考虑过这会是一种新货币，或者是一种新的投资或资产管理方式。

I don't think there was much thought about this as a new currency or this is a new way to invest or a new way to manage investments.

只是觉得它在上涨，我就去买。

Was just like, this is going up, I'm gonna buy it.

可惜的是，我们从共同基金、ETF中早就知道这一点：人们总是购买那些已经上涨的东西。

And sadly, we know this from mutual funds, we know this from ETFs, people buy what's been going up.

可惜的是，这是一种糟糕的投资方式，但人们就是这么做的。

Sadly, that is a bad way to invest, but that's what people do.

所以，这就是发生的事情，对吧？

And so this is what happened, right?

它上涨了，然后某种程度上公开了，人们可以买到了。

It went up, it went public sort of, you could get at it.

它上涨了，于是人人都买了。

It went up and everyone bought it.

然后它下跌了。

And then it went down.

对。

Right.

很正常。

Happens.

那么，再来说说这里的类比，你在文章中提到一个关键点：如果人工智能和太空等领域真的成为巨大市场，企业的基本面会蓬勃发展，我不禁想回到那些SaaS公司纷纷上市的时期。

Well, and so, again, trying to think about parallels here, you mentioned in your column the idea that the caveat is, look, if AI and space, for example, become huge markets, and the fundamentals of the businesses will boom, I wonder if you go back to the period where all these SaaS companies started going public.

我的意思是，那时候是不是也有类似的故事？当时有大量私营公司？

I mean, was it kind of a similar story there, where you had all these private companies?

不过，它们筹集的资金并没有那么多。

They weren't raising as much money, mind you.

我的意思是，这些企业的资本规模都比较小。

I mean, you know, these were less capitalized businesses.

但我一直在思考那个故事是如何发展的，因为当时我们确实看到了估值的飙升，随后又大幅回落。

But, I'm sort of trying to think about how that story played out, because we kind of did see a boom in the valuations, and then they very much did come down.

是的，我认为当时的SaaS公司作为私营企业并没有像我们现在看到的这样高的估值。

Yeah, so I don't think the SaaS companies had these high valuations as private companies, as crazy as we're seeing now.

所以并没有那么大的稀缺价值。

So there wasn't that much of a scarcity value.

我的意思是，也许有一些，但它们当时并没有筹集数十亿甚至两百亿美元的资金。

I mean, maybe there was some, but they were not raising tens and $20,000,000,000 the time.

因此，这个等式的第一部分并不存在。

And so, I think the first part of the equation wasn't there.

当这些SaaS公司上市时，它们已经实现了真实的利润和现金流。

And then these SaaS companies, when they did go public, they had real profits and real cash flow.

这是一门好生意，对吧？

That's a good business, right?

OpenAI的情况，你根本说不准。

OpenAI, you just don't know.

它们能赚多少钱、如何管理巨额支出和现金消耗，都极具投机性。

It's really speculative how much money they're gonna make, how they're gonna manage all the spending and the cash burn.