Anthropic’s Invasion of Slack, OpenAI Cuts Inference Costs in Half, Amazon’s Higher Anthropic Costs

欢迎各位收看《信息时报》的TITV节目。

Welcome everyone to the Information's TITV.

我是阿卡什·帕什里卡。

My name is Akash Pasricha.

今天是6月30日，星期二。

It is Tuesday, June 30.

今天首先带来一条独家报道：前AWS首席执行官亚当·塞利普斯基已经加盟一家数据中心公司，负责领导其在AI基础设施领域的竞争。

First up today, the information published exclusive reporting that former AWS CEO Adam Selipsky has a new job taking the helm of a data center company aiming to compete in the AI infrastructure race.

我们的AI与财经记者为此事专门采访了塞利普斯基。

Our AI and finance reporter spoke to Selipsky as part of the story.

稍后我们会直接听到他的发言。

We will hear from him shortly.

接下来，Anthropic公司重新谈判了与亚马逊的一项协议，可能让这家科技巨头支付更多费用。

Next up, Anthropic has renegotiated one of its deals with Amazon, potentially increasing what the tech giant pays.

我们的亚马逊特派记者带来最新内幕。

Our Amazon reporter joins us with the scoop.

说到Anthropic，这家公司推出了一款与Slack集成的AI工具，但让Salesforce的员工感到有些困惑。

Speaking of Anthropic, the company released an AI tool that integrates with Slack, but it is creating some confusion for Salesforce employees.

我们接下来会听到揭露这则新闻的记者的更多分享。

We will hear more from the reporter who broke that news.

最后，OpenAI似乎找到了大幅降低推理成本的方法。

And finally, OpenAI appears to have found a way to reduce the cost of inference by a significant margin.

我们会讨论客户是否能真正享受到这些成本下降的好处。

We'll talk about whether or not customers will see those cost savings.

今天这期节目内容很精彩，咱们马上开始。

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

Adam Selipsky这个名字，你可能听过。

Adam Selipsky is a name that you might recognize.

这位前AWS首席执行官现在加入了Helix Infrastructure Partners，一家专注于数据中心的公司。今天早上，《信息》的AI与财经记者达金·坎贝尔独家报道了相关细节。

The former CEO of Amazon Web Services has a new gig at Helix Infrastructure Partners, a data centric company, that the information's AI and finance reporter Dakin Campbell reported exclusive details about in a piece published this morning.

我请达金来给我们详细讲讲他发现了什么。

I want to bring on Dakin to walk us through what he found.

Dakin，欢迎再次做客节目。

Dakin, welcome back to the show.

很高兴你能来。

It's great to have you here.

是的，非常感谢。

Yeah, thanks so much.

Helix Infrastructure Partners 是家什么样的公司？你发现了什么？

What is Helix Infrastructure Partners, and what did you find?

这是一家刚进入数据中心市场的新兴企业。

So this is a new data center new entrant into the data center market.

他们目前还没有投入运营的资产，但很快就要收购一些了。

They don't have any operating assets yet, but they have plans to acquire some very shortly.

这家公司会建造数据中心，然后租给超大规模云服务商。

You know, it will be a company that builds data centers and then leases them to hyperscalers.

Selipsky 将担任这家公司的首席执行官。

Selipsky will be the CEO of the company.

他们几周前宣布筹集了100亿美元，还和英伟达、维斯特拉能源达成了战略合作。

You know, they they announced a couple weeks ago that they'd raised $10,000,000,000, with strategic partnerships with NVIDIA and Vistra Energy.

所以我们现在是在实时了解这家公司到底会变成什么样。

And so this is sort of we're learning about what this is gonna be a little bit in real time.

接下来很值得看看他们会怎么竞争，又会在整个生态里占据什么位置。

And so it'll be interesting to see sort of how they compete and where in the ecosystem they slot in.

那这么说，他们的想法到底是什么？

And so the idea here is what?

他们是打算收购数据中心吗？

They're going to buy up data centers?

这就是他们的计划吗？

Is that the game plan?

他们会收购一家现有的数据中心运营商，可能是规模较小、急需资金的公司，那些背后有私募股权或主权财富基金支持的，可能还没来得及注资，或者想套现退出。

They're going to buy up a data center operator, an already existing operator that might be small, that might need more money, that might have private equity or sovereign wealth backers who haven't been able to put the money in or may want their money out.

几周前宣布这个消息的时候，大家其实都在猜，他们到底是会从零开始建，还是直接收购。

So, you know, there was some question when this was announced a couple weeks ago whether they would build from scratch.

我们和他们聊过之后，最大的消息是，他们不会从零开始建了。

And the big news that that we've learned through talking to them is that they are not gonna build from scratch.

他们会去找至少一个重要的收购目标，作为基础，然后以此为起点扩展。

They're gonna be on the hunt for at least one significant deal that they can then use as a foundation to build from and scale from.

明白了。

Got it.

所以他们本质上不是一家控股公司，而是一家管理公司，专门收购那些已经运营数据中心的管理公司。

So they're basically not a holding company, but it's management company that is going to buy management companies that already manage data centers, essentially.

是的。

Yes.

说得一点不差。

That's exactly right.

他们相信，收购之后，凭借自己独特的竞争优势，能更快、更大规模地把这家公司发展起来。

And then they believe that they can take that company that they buy and grow it much quicker, much bigger because of their unique set of what they believe are competitive advantages.

现在，塞利普斯基是这家公司的核心人物。

Now, Selipsky is the man at the center of this company.

你有机会和他聊过。

You had the chance to speak with him.

我想知道，你去采访他之前，最想问哪些关键问题？

I wondered what were the big questions that you went into that interview with?

是的。

Yes.

我其实一直在想这个问题。

I you know, I was thinking about this.

我可能不该这么直说，但说实话，亚当·塞利普斯基这个名字，对我而言没那么响亮，不像对你们观众或者新闻部其他同事那样。

I probably shouldn't be totally honest about this, but, you know, Adam Selipsky, the name doesn't mean quite as much to me as it does maybe to our viewers or certainly to other colleagues, in the newsroom.

我是个搞财务的。

I'm the I'm the finance guy.

所以我去采访前，问题都集中在那条公告太模糊了。

So, you know, my questions going into it were very much, you know, the announcement was pretty vague.

根本说不清楚。

It wasn't clear.

所以我想知道，一百亿美元可不是个小数目。

So I wanted the you know, $10,000,000,000 is a lot of money.

我想了解他们打算怎么用这笔钱。

So I wanted to know sort of what they were planning to do with it.

我还想弄清楚，亚当·塞利普斯基掌舵到底意味着什么，他们觉得他能带来什么，他自己又觉得能做些什么。

I wanted to know a little bit about what, you know, Selipsky being at the helm really meant for them and and and what they thought he could bring and what he thought he could bring.

所以我的问题更多是围绕这家公司未来会做什么。

So question much more around, you know, what is this company gonna be doing?

他们能在这上面成功吗？

And, you know, are they gonna be successful at it?

明白了。

Okay.

那他怎么说的？

And and what did he say?

他说他们现在特别关注与英伟达和维斯特拉的战略合作。

You know, they're they're really focused on their these strategic partnerships that they've got with NVIDIA and Vistra.

这两家公司加上科威特投资局，各自投入了十亿美元或更多，凑成了这百亿规模。

Both companies plus the Kuwait Investment Authority each put a billion dollars in or more into this 10,000,000,000.

Vistra是一家能源公司。

And so, know, Vistra is an energy company.

据我所知，目前还没有哪家数据中心运营商能像他们这样，跟能源供应商有这么紧密的联系。

There are no, to my knowledge, data center operators out there that have such a close connection to an energy provider.

我们都知道，电力供应是这个行业的一大瓶颈。

We know power is a big bottleneck in the industry.

所以他们的想法是，因为能随时跟Vistra的CEO或者董事会直接通话，

And so the idea I think is because they've got Vistra on on a phone call and can talk to, you know, the CEO or the board.

如果他们需要建一个数据中心，还得配套建个发电厂，直接打个电话给Vistra就行。

If they've got a data center that they need a a power plant or something built for, they'll call Vistra.

或者在选址时，他们可能会特意选在Vistra已有的发电厂或电池设施附近。

Or if they are looking to cite a data center, they might, you know, choose to cite it nor near some power plants or some battery facilities that Vistra already owns.

所以核心就是，电力供应是关键瓶颈。

So the idea there is power is a big bottleneck.

如果你能随时联系到国内最大的能源公司之一，那应该就能稍微打破这个瓶颈。

If if you've got one of the biggest energy companies in the country, you know, on speed dial, then you should be able to, you know, break up that bottleneck a little bit.

另一个战略合作伙伴是英伟达。

The other strategic partner is NVIDIA.

这一点不能忽视。

We can't overlook them.

据我了解，英伟达最近推出了一款新产品，能更高效地建设数据中心，让电力更精准地供给他们的芯片。

You know, NVIDIA has a somewhat new product out from what I can tell, you know, helping to build data centers more efficiently so that the power in a data center most efficiently powers their chips.

所以，Helix 将和英伟达合作，从零开始打造超高效的数据中心，他们相信这样不仅能加快建造速度，还能为未来的大型客户推出更具吸引力的方案。

And so Helix will be working with NVIDIA to to build super efficient data centers from the ground up, which they believe will allow them to do them, to build them more quickly, and make a much more compelling package for the hyperscalers who are gonna be their clients.

他有没有提到他们可能考虑收购哪些公司？

Did he give you a sense for any of the companies that they could look to acquire?

我们有线索吗？

Do we have any idea there?

他没说。

He didn't.

你明白吧？

You know?

我觉得这正是这家公司的核心优势。

I anticipate that as the secret sauce of the whole company.

没错。

Yes.

就是这样。

Exactly.

他们主要做的是业内说的‘定制建设’。

I mean, the place they're gonna play is is what's known in the industry as build to suit.

所以他们会跟超大规模客户沟通。

So, you know, they're gonna talk to hyperscalers.

超大规模客户会明确告诉他们需要什么样的冷却系统、电力供应，还有备用发电机，然后他们就按这些要求来建造。

Hyperscalers will tell them exactly what they want in terms of cooling, in terms of power to the facility, in terms of, you know, backup generators, and they'll build them then for the hyperscalers.

我觉得他们也会做一点‘动力壳’，就是先把数据中心的四面墙建好，通上电，然后把一个空仓库交给超大规模客户。

And I think they'll be doing a little bit of powered shell, which is basically you build the four walls of the data center and get power to it and then give a empty warehouse basically over to a hyperscaler.

这是一种很多公司都在做的商业模式。

So that is a business model that a lot of companies are in.

像这种程度的私募股权，更多是投向那些更成熟的企业。

Private equity to this degree has really weighed in higher, like, you know, more established companies.

看起来他们是打算收购一家中型公司，然后把它打造成行业领军者。

This looks like they'll be they'll buy somebody in the in the mid tier and then, you know, look to grow that into an industry leader.

没错。

Right.

能再详细说说私募股权在数据中心领域的布局吗？

Talk a little bit more about private equity and the extent to which it has moved into data centers.

你有听到这方面的消息吗？

Is this something that you're hearing?

对了，我们还没提到KKR在这笔交易里的角色，KKR可是这家新公司的投资人吧？

By the way, we haven't even acknowledged the KKR component of this venture KKR is an investor in the new venture?

是的。

Yes.

是的，他们确实如此。

They they are.

他们实际上是Helix的拥有者，并且拥有四家合作伙伴共同投入的100亿美元，用于在发现目标公司时进行收购。

And they they are basically the owner of Helix, and they'll have the $10,000,000,000 that was contributed by all four partners to draw down when they see a company that they want to acquire.

那么，KKR是唯一一家在关注数据中心的公司吗？

So Is KKR the only company here looking at data center?

我们之前聊过其他融资交易。

We've talked about other financing deals.

我觉得阿波罗也参与其中。

I think Apollo is in the mix.

所以，这是一种可能会变得更常见的交易模式吗？

So is this a framework for a deal that is likely to be more common?

这是个好问题。

It's a good question.

某种程度上，KKR在这个领域算是来晚了。

I mean, to some degree KKR is late in this.

你知道吗，黑石在2021年收购了QTS。

You know, Blackstone bought QTS in 2021.

今年早些时候，他们收购了罗万公司49%的股份，这家公司的发展方向，跟Helix未来要走的路很像。

Earlier this year, they they acquired 49% of a company called Rowan that, you know, looks to be what Helix is going to become.

不过，确实如此。

But but yeah.

我的意思是，现在资金多得是。

I mean, there are lot there's lots of money.

现在还在不断有新资金进来。

There's still money being raised.

所以，他们可能算不上领先，也算不上落后。

So, you know, maybe they're middle of the pack.

我觉得在他们之后，还会有很多新玩家进场。

I think we will see other entrants come in after them.

有大量资金已经准备好出手，而且还在继续募集更多。

There's a ton of money that's ready to be deployed and more money currently being raised.

你知道吗，Helix认为他们的竞争优势在于这些战略合作伙伴关系。

You know, what Helix what they believe their competitive advantage to be is these strategic partnerships.

所以我不觉得NVIDIA或者Vistra会大力支持太多其他由私募股权 backing 的公司。

So I'm not sure you'll see NVIDIA or Vistra lend their support to a lot of a lot more, you know, private equity backed players.

从这个角度看，Helix的做法可能比较独特，甚至相当少见。

So in that regard, this might be unique or or somewhat rare.

而且，还有更多资金正在涌入。

Well, and I also want More money's coming in.

这点毫无疑问。

That's for sure.

更多资金确实在进来，但我也想提一下亚当·塞利普斯基这部分，我知道你可能不像金融记者那样熟悉这个名字，但AWS的CEO可不是个小角色，他在这一行里举足轻重。

More money's coming in, but, you know, I sort of I don't wanna overlook the Adam Selipsky portion of this, and I know it's not a name that you might have followed as closely as a finance reporter, but the CEO of AWS, a one time CEO, that's not a small character to play in this space.

所以我想，KKR之所以这么做，可能是觉得他们在等一个对的人，或者一个合适的时机。

So I assume there is some confidence there from KKR in saying, well, maybe we were waiting for the right person to come along or the right venture.

你从塞利普斯基那里感受到什么了吗？

Did you get a sense from Selipsky at all?

我是说，从个人角度来说，他有没有跟你提过，为什么在执掌AWS这么久之后，还要把职业生涯的下一阶段投入到基础设施这个领域？

I mean, on a personal level, did he tell you at all why he had chosen to commit the next stage of his career to this infrastructure category, I mean, after leading AWS for so long?

是的。

Yeah.

他去年九月以高级顾问的身份加入KKR的。

I mean, he started at KKR in September of last year as a senior advisor.

据我了解，Helix这个想法很快就成型了。

And from what I understand, the idea for Helix came together pretty quickly.

我觉得他觉得，数据中心、电力供应，还有支撑AI所需的基础设施能力，是当下商业世界里最大，甚至是最关键的挑战。

I mean, I think he sees data centers, access to power, and basically the capacity needed to power AI as the, you know, one of the biggest, if not the biggest challenge in the in the business world right now.

所以从这个角度看，这件事让他特别感兴趣。

So I think that from from that extent, was exciting to him.

而且他希望能在高层级上参与其中，而这个项目正好给了他这样的机会。

And and, you know, he wants to he wants to be in the mix at a high level, and and this gives him the opportunity.

你懂的。

You know?

很明显，他会去联系其他超大规模云服务商的首席执行官和部门主管，向他们推介Helix。

And it's very clear he's gonna be calling on, you know, CEOs, division leaders at the other hyperscalers pitching Helix.

他手上握着一大把这样的资源，而且打算好好利用。

So, you know, those are that's a Rolodex that he's got in spades and and plans to use.

没错。

Right.

我觉得我们以后会更多地听到这个名字。

Well, I anticipate it's gonna be a name that we're gonna hear even more about.

Dakin，非常感谢你来做客。

Dakin, I wanna thank you for coming on.

这位是达金·坎贝尔，我们TITV频道报道AI领域资本动态的记者。

That is Dakin Campbell, our reporting our reporter covering all the money behind the AI boom here on TITV.

非常感谢你。

Thanks so much.

好的。

Okay.

尽管亚马逊是Anthropic的重要投资者，但Anthropic在定价上对亚马逊的态度变得更严格了。

Anthropic is taking a tougher stance on Amazon in terms of pricing, despite the fact that Amazon is a major investor in the company.

我的同事凯瑟琳·佩洛夫本周报道了这段合作关系的内部细节，我想请她来分享一下她发现的情况。

My colleague Catherine Perloff reported inside details of that partnership this week, and I want to bring her on to share more about what she found out.

凯瑟琳，欢迎再次做客节目。

Catherine, welcome back to the show.

很高兴你能来。

It's great to have you here.

嗨，阿卡什。

Hi, Akash.

凯瑟琳，你了解到Anthropic和亚马逊在定价上的具体安排了吗？

What did you find out about Anthropic's pricing plan with Amazon, Catherine?

你知道，Anthropic最近调整了对很多客户的收费方式，不少客户现在都要为他们的模型支付更多费用。

So, you know, Anthropic has, been charge changing how they charge a lot of their customers, and a lot of those customers are now paying more for their models.

亚马逊也不例外。

Amazon is no exception.

但很明显，亚马逊可不是普通的客户。

But obviously, Amazon is not just any customer.

所以我们对这个故事特别感兴趣。

So that's why we were interested in this story.

大概几个月前，他们改了合作方式，不再按计算时长收费，而是改按令牌数计费——也就是亚马逊使用Anthropic模型时所消耗的令牌数量。

Basically, they, a couple months ago, changed the deal so that instead of Amazon paying, Anthropic in a unit called compute hours, so basically, you know, the numb the sort of hours of compute taken up when Amazon is using Anthropic models, they shift to tokens.

这样一来，亚马逊使用Anthropic模型的成本就会上升，这可非同小可，因为Anthropic的模型支撑着亚马逊很多核心AI产品，比如购物网站上的聊天机器人Alexa。

And this is gonna make it more expensive for, Amazon to use Anthropic models, which is sort of a big deal because Anthropic models power a lot of Amazon's big AI products that includes Alexa for shopping, which is their chatbot on their retail site.

还有AWS推出的几款软件，像编程工具Curo和办公助手Quick，也都依赖它。

Some of the AWS software they're selling like Curo, a coding product, and Quick, a workplace assistant.

所以，如果他们还想继续用的话……

So, yeah, if they wanna you know?

Anthropic并不是亚马逊唯一的底层模型，但绝对是其中最重要的一个。

Anthropic isn't the only model underneath the hood, but it is a big one.

现在，他们的开销要涨了。

So now their costs are going go up.

是的。

Yeah.

说实话，从Anthropic的角度来看，这确实合理。

Look, I mean, from an Anthropic end, it certainly makes sense.

我们之前聊过，现在代理工具使用的token数量有多庞大。

We've talked about how many more tokens are being used right now with agents.

既然用户在单位时间内消耗的token更多了，按使用量收费对他们来说当然更有利。

And so the idea that you can consume more tokens in a per hour basis, it would certainly be beneficial for them to charge based on the consumption model there.

我们有没有任何信息说明，到底是什么原因导致了这次调整？

Do we have any reporting at all on why this changed and what flipped the switch here?

我不清楚当时每个人是怎么想的，但据我了解，亚马逊员工早就意识到自己享受了很优惠的定价，担心成本会上涨。

You know, I'm not exactly sure what was in everyone's heads, but do my understanding is that Amazon employees for a while were sort of aware they were getting a pretty good deal, and they were worried that the cost could increase.

还有一点值得注意，就是之前用的计费单位——计算小时。

And, you know, another thing to kind of point out is the the previous unit compute hours.

亚马逊AWS显然是个庞大的基础设施公司，用计算小时计费，更方便他们管理和优化计算资源。

Amazon AWS is obviously a massive infrastructure company that's something easier for, AWS to manage the cost and, you know, optimize their compute.

所以当Anthropic需要付费时，他们可以优化一下，成本可能更低。

So Anthropic could, so when they had to pay Anthropic, you know, that could be cheaper, they could optimize it.

但使用令牌的话，这个单位是由Anthropic掌控的。

But with tokens, that's the unit that Anthropic controls.

亚马逊在这方面能主动降低成本的空间就小多了。

There's sort of less, leeway Amazon has to sort of, you know, do their best to optimize costs down.

不过这里还有一点很重要，Anthropic现在已经成长为一家大公司了。

I think the other thing to really point out here though is, Anthropic is becoming, like, a really big company.

你知道的。

You know?

它正逐渐成为一家主导性的企业科技公司，而且很快就要上市了。

It's sort of becoming this dominant enterprise tech company, and, they're gonna go public soon.

尽管亚马逊曾是它最大也是最早的投资者之一，但现在Anthropic的议价能力比以前强多了。

And despite the fact that, like, Amazon was one of their biggest and earliest backers, they have a bit more leverage than maybe they used to.

没错。

Right.

而且你刚才提到，亚马逊和OpenAI也有合作关系，就是通过他们最新的融资。

And and, I mean, you pointed out in the story that and that Amazon also has a relationship with OpenAI here through its latest funding.

我记不清是不是500亿美元的投资了，我觉得是亚马逊

I can't remember if it was a $50,000,000,000 investment, I think, that Amazon

加起来达到了500亿美元

It's made up in to 50,000,000,000 Up

500亿美元

to 50,000,000,000

是的。

Right.

他们还没全投完，但已经砸了大笔资金进去。

Think they haven't done it all yet, but a lot of money on the table.

是的。

Right.

我记得很清楚，当初亚马逊说要向Anthropic投资最多40亿美元，后来又追加了更多。

And I certainly recall when I think Amazon said we will invest up to 4,000,000,000 in Anthropic, and then they came up with another one and stuff like that.

好吧，行吧。

So, okay, fine.

亚马逊正在同时拉近与OpenAI和Anthropic的关系。

So Amazon is cozying up to both OpenAI and Anthropic.

这可能会带来一些影响。

Maybe that has repercussions here.

我们还不确定。

We're not sure.

亚马逊跟你们说过他们的成本上升了吗？

What did Amazon tell you about their costs going up?

他们有没有承认，这个新定价模式可能导致成本上涨？

Did they acknowledge that their costs could go up through this new pricing model?

亚马逊对我们的说法是，定价模式可能变了，但他们不觉得自己的成本在上升。

Amazon's statement to us was that pricing model may have changed, but they don't feel like their costs aren't going up.

我的消息源说这会让成本更高，所以咱们只能各持己见了。

My sources said it would make it more expensive, so gonna have to agree to disagree.

但你知道，这也没关系。

But, you know, that's okay.

你是懂的。

You know?

他们确实也有一定的道理。

And they and they probably have some fair points there.

没错。

Right.

Anthropic 告诉我们的看法是，虽然定价模式可能在调整，但他们清楚自己对不同客户收的费用不一样，越来越偏向按使用量计费。

And Anthropic's POV that they shared with us was like, you know, while the pricing model may have, you know, may be changing, maybe they're they know they're charging all customers different, you know, with more usage based pricing.

总的来说，AI 模型的成本正在下降。

And overall, AI models are getting cheaper.

所以就算单位经济模型在变，AI 模型的实际成本确实在降低。

So even if, like, the unit economics is changing, you know, AI the actual cost of the models is going down.

而且从大方向看，亚马逊对成本是很敏感的。

So, and I think also big picture, Amazon is conscious of costs.

我跟他们的一些高管聊过。

Like, they I I've spoken to some of their executives.

比如在他们的产品Quick里，如果任务允许，他们会引导用户使用更便宜的模型。

And for example, in their product, Quick, they kinda help the customers of that product use a cheaper model when the task allows for it.

所以他们在后台悄悄把用户导向成本更低的模型，帮这些用户降低AI软件的使用开销。

So they're kind of under the hood routing customers to a cheaper model to help those customers of their AI software keep keep costs down.

他们确实一直在这么做。

So they they are definitely

他们可不是坐视不管、任由成本上涨。

They're not just they're not just sitting idly by basically and and and taking costs.

他们正在想办法降低开支。

They're finding ways to

没错。

Right.

是的，没错。

Yeah, yeah.

我想知道，在这段长期合作关系中，你觉得亚马逊还有什么其他谈判筹码？我敢肯定这些协议以后还会不断重新谈判。

I wondered what other bargaining chips do you think Amazon has here, knowing that this relationship is a long one, and I'm sure these deals will continue to get renegotiated.

我的意思是，考虑到Anthropic确实依赖亚马逊的云服务，你觉得亚马逊还能施加哪些其他影响？

I mean, are there other levers do you think Amazon could pull on here given that Anthropic certainly needs Amazon for cloud services?

你觉得这些谈判长远来看会怎么发展？

How do you see these negotiations playing out in the long run?

是的。

Yeah.

这挺有意思的。

It's interesting.

我的意思是，亚马逊内部肯定也考虑过，面对不断上涨的成本，我们能不能改用OpenAI的模型？

I mean, so Amazon has internally thought about, like, with these rising costs, could we use OpenAI's models?

能不能更多地用我们自己的Nova模型来驱动一些产品？

Could we use our own, Nova models more to power some of our products?

还有待观察。

You know, remains to be seen.

当然，他们还有其他供应商可以选择。

So, obviously, they can there's other providers out there.

Anthropic 依然高度依赖 AWS 提供基础设施，同时也靠 AWS 来拓展新客户。

I mean, am Anthropic, still really relies on AWS for infrastructure, and they also really rely on AWS to help them reach new customers still.

他们通过 AWS 销售 Anthropic 的模型。

They sell Anthropic.

很多模型的销售都是在 AWS 上完成的，这帮助 Anthropic 获得了大量企业客户。

A lot of the model sales happen on AWS, and that was a big way for Anthropic to access a lot more business customers.

通过这种合作模式，Anthropic 在 AWS 上销售模型，要将毛利润的 50% 支付给 AWS。

And through that business relationship where Anthropic sells their models in AWS, they pay, like, 50% of the gross, profits.

你今年早些时候就报道过这件事，是给亚马逊的。

You reported this earlier this year back to Amazon.

也就是说，即便付完云服务费用后，他们还得把一半的利润再交给亚马逊。

So, you know, they're basically, that means after they pay their cloud computing bill, they still have to fork over 50% of the profits back to Amazon.

所以，亚马逊掌握了基础设施。

So, you know, Amazon had the infrastructure.

他们还提供一些Anthropic正在使用的芯片。

They also have some chips that Anthropic is using.

所以，绝对不是说Anthropic离不开亚马逊。

So definitely not like, Anthropic doesn't need Amazon.

但现在，Anthropic可以说，我们自己也有了不少企业客户。

But, you know, now Anthropic can say we've got some business customers of our own.

你知道，计算成本真的很高。

You know, we also, compute is really expensive.

我们得让商业模式跑得通，大家也都得互相配合。

You know, we have to make our economics work, and, everyone kinda has to play ball.

很好。

Great.

Catherine，非常感谢你今天来做客。

Well, Catherine, I wanna thank you for coming on.

这位是Catherine Perloff，我们《The Information》的亚马逊记者。

That is Catherine Perloff, our Amazon reporter here at The Information.

Anthropic 最近推出了一款与 Slack 集成的 AI 工具，但当 Slack 的母公司 Salesforce 开始推广这款工具时，让一些员工感到困惑，担心会削弱自家产品的市场。

Anthropic released an AI tool that integrates with Slack recently, but when Slack's parent company Salesforce started promoting the Anthropic tool, that confused some of the company's own employees over risks of cannibalization.

为了更清楚地说明这一点，我请来了我们《应用人工智能》通讯的作者劳拉·布拉顿。

To explain that more, I want to bring on Laura Bratton, author of our Applied AI newsletter.

劳拉，欢迎来到节目。

Laura, welcome to the show.

很高兴你又回来了。

It's great to have you back.

嘿，阿卡什。

Hey, Akash.

最近怎么样？

What's up?

你那边怎么样？

What's up?

你来说吧。

You tell me.

我来告诉你。

I'll tell you.

好的。

Okay.

说说这个新工具吧。

Tell me about the new tool.

Claude Tag 是什么？

What is Claude Tag?

是的。

Yeah.

Claude Tag 就像是企业用户在 Slack 里能召唤的一个共享助手。

So Claude Tag is basically a shared teammate that enterprise users can access in Slack.

你只要在 Slack 频道里直接 @ Claude，让他回答问题，或者帮忙处理一些管理或编程任务，所有团队成员都能看到。

So you tag Claude literally at Claude in your Slack channel that you want Claude to answer a question in or do some sort of administrative or coding task in and all your teammates can see it.

随着时间推移，Claude Tag 会逐渐积累该频道的上下文和记忆，还能主动跟进你，替你发起一些操作。

Over time, Claude tag builds context and memory from your Slack channel, and it can follow-up with you and sort of initiate actions on your behalf.

你用过吗？

Have you used it?

你有没有在 Slack 里 @ 过 Claude？

Have you at clauded yet on Slack?

我还没在 Slack 里用过 Claude，不过我们或许该问问编辑团队，要不要在频道里试试这些工具。

I have not at clauded yet, but I guess we should ask our editors if we should be testing these tools in our Slack channels.

说得对。

True.

确实如此。

Very true.

而且，你刚才提到的一点，我们待会儿再聊，但这一切确实涉及隐私问题。

And, well, because one of the things that you mentioned, we'll talk about in a second, but there are privacy implications to all this.

不过我用过的一个工具是 Slackbot。

But one of the tools that I have used is Slackbot, okay?

我试过一些基本功能，比如设置提醒。

And I have tried to, you know, basic functions, reminders, okay?

它们做提醒还挺靠谱的，但我试过设置一些自动化功能。

And they do reminders pretty well, but I've tried to create some automations.

效果最多也就马马虎虎吧。

It works like okay at best.

说实话，以我的使用经验来看，这工具真不算出色。

It's not really the greatest tool, I will say, in my experience.

但是你刚才说的那些，我有点没听清。

What But are I

我想问一下，你从一月份开始用了吗？就是Slack改用Anthropic模型之后那次更新？

just want to ask, have you been using this since January, since they did the reimagining of Slack powered by Anthropic Models and stuff?

你是在那之后开始试用的吗？

That's when you're trying it?

是的，我确实试过设置，之所以我知道，是因为我创建的那个提醒，现在在通知里显示的是‘正在思考’。

So I tried to create, yes is the answer, because and the reason I know this is because now the reminder that I've created, I see in the notification, it says thinking.

所以我猜后台应该有模型在运行。

So I assume there's some models working in the background.

没那么好用。

It's not that great.

我不确定。

I don't know.

我是不是加载中了

Am I loading I

说实话，我本人没用过Slack机器人，但我接触过很多用过它的人。

mean, I haven't tried Slack bot personally, but I've talked to a lot of people who have used Slack bot.

他们说，用Slack机器人的主要目的，是跟Salesforce里的客户关系数据互动。

And they said the ways that they're using Slack bot are primarily to interact with their CRM data in Salesforce.

也就是客户关系管理软件。

So that's customer relationship management software.

我接触的那些团队，会用Slack机器人从CRM里提取商机进展、客户信息，汇总客户的人口统计特征，然后在会议里用这些生成的报告。

So teams that I talked to will use Slack bot to pull data from their CRM about the deals they have in the pipe pipeline, or their customers to summarize the sort of demographics of their customers, and use the reports that they can generate from that data in meetings.

很多行政类的工作，而Salesforce也希望Slack机器人能像个人助理一样，接入你用的任何商业软件。

So a lot of administrative tasks and Salesforce certainly wants Slackbot to be kind of like a personal assistant that can hook up to any business software applications that you use.

因为他们有MCP服务器，可以连接客户使用的任何商业软件。

Because they have MCP servers that can connect with any business software that a client uses.

然后他们希望Slack机器人帮你完成这些软件系统里的各种行政任务。

And then they want Slack bot to help you do whatever administrative tasks you want to do in those software systems.

但客户最终会不会选Slackbot来做这些，还有待观察。

But I think it remains to be seen if customers will choose Slackbot for that.

现在又有好多其他AI代理被集成到Slack里，客户可能会选它们代替。

And now you have all these other AI agents that they're loading into Slack that customers might choose instead.

所以这不只是Claude Tag一个。

So it's not just Claude Tag.

还有Perplexity、Linear、Cursor这些公司的AI代理。

There's also AI agents from Perplexity, Linear, Cursor.

一家叫Victor的小公司也推出了一个代理，功能跟Claude Tag几乎一模一样。

This small startup Victor has an agent that essentially does all the same things that Claude Tag does.

不过我觉得Claude Tag特别引人注目，是因为Salesforce对它的推广力度比其他Slack上的AI代理都大，还在X上发了专门的帖子，高管们也在领英上转发了。

But I think in Claude Tag particularly made a splash because Salesforce also pushed it a little bit more heavily than it did other AI agents that have been launched in Slack and had a whole post on X and Salesforce executives posted about it on LinkedIn.

我觉得这正是造成困惑的原因——我们为什么要大力推广这个和自家专有AI代理竞争的工具呢？

And I think that's what created some of the confusion is, you know, why are we heavily pushing this AI agent that kind of competes with our own proprietary AI agents?

没错。

Right.

我得说，也许我还没给Slackbot足够多的时间去证明自己。

And I should say, I mean, look, maybe I haven't given Slack bot the time that it deserves.

好吧。

Okay.

我本意也不是要贬低Slackbot。

Maybe I I I don't mean to to, you know, talk down on the Slack bot.

我们该去问问业务团队，毕竟才是他们直接在使用这些工具。

We should talk to our business teams because they're the ones actually interacting with us.

没错。

Right.

我目前也没在用任何CRM系统。

And I don't have a CRM that I'm pulling from.

我并不是他们理想中的目标客户。

I'm not exactly the ICP here that they are looking for.

但我的意思是

But mean

但它对你来说还是应该能用得很好，你知道的，没错。

But it still should work well for you in a way that, you know, yeah.

应该很简单，对吧？

It should be easy, right?

更简单一点。

Easier.

你刚才提到了这一点，但问题的核心在于，现在Claude要进场了，对吧？

So now you touched on this now, but here the crux of this issue is that you have Claude tag coming along, right?

这就是Claude的功能。

And that's the Claude function.

而Salesforce的员工会问：既然我们的Slack机器人也能实现Claude助手的不少功能，为什么我们要大力推广这个Claude集成？

And you have Salesforce staffers that are saying, Well, why are we Salesforce promoting this Claude integration when our Slack bot conceivably could do some of what the Claude assistant could do?

我想确认一件事。

I just want to confirm one thing.

所以我用 Slack 机器人的时候，背后是 Anthropic 的模型在驱动吗？

So when I use Slack bot, is that powered by Anthropic models under the hood?

是这个意思吗？

Is that the idea?

是的。

Yes.

对。

Yes.

明白了。

Okay.

那我想问你，Salesforce 为什么要推广 Claude？你们觉得背后的理由是什么？

So then the question that I have for you is then what did you find as the rationale here for why Salesforce would be promoting Claude?

这里面有商业考量吗？

Is there a business?

这里头是不是有什么幕后交易，他们因此获得报酬？

Is there a deal under the hood here that they're getting paid for this?

Salesforce 和 Anthropic 都不会向我透露他们协议的任何财务细节。

So Salesforce nor Anthropic would tell me any of the financial terms of their agreement.

但我从其他提供 AI 代理服务的公司，或者那些在 Slack 上部署代理的人那里了解到，他们并没有向 Salesforce 付费。

I did find out from some other providers or people close to the other providers of AI agents that have agents in Slack that they're not paying Salesforce.

这些消息来自维克托和佩洛夫。

That was Victor and Perloff.

我从熟悉这些公司的消息源得知，他们并没有向 Slack 或 Salesforce 付费，这让我对 Salesforce 的说法产生了怀疑——他们声称想把 Slack 打造成一个类似苹果应用商店的 AI 代理平台。

I found out, from sources familiar with those companies that they're not paying Slack Salesforce, which sort of puts a dent in the argument I heard from Salesforce, which is that they want Slack to be kind of like an Apple App Store for all AI agents.

我觉得这个举动其实表明，他们更看重 Slack 这个聊天平台的用户粘性，而不是它的单个 AI 机器人能否成功。

And I think this move just shows that they're prioritizing the stickiness of Slack, their messaging app, over the success of their individual agent's Slackbot.

即使 Slackbot 并不是每个企业客户首选的 AI 代理，但如果他们能在 Slack 里自由使用任何想用的 AI 代理，可能更愿意留在这个平台上。

So even if Slackbot isn't necessarily the AI agent of choice choice for every enterprise customer, they might be more willing to come to Slack if they can use any AI agent that they want to in the messaging platform.

但接下来

But then

因为他们不用额外付费就能使用Claude助手或者Perloff助手，对吧？

Because they don't they don't have to pay to access the Claude assistant or the Perloff assistant, right?

这对他们来说不是额外的成本。

This is not an added cost for them.

但在他们与Perloff或Claude的账户里，其实是算钱的。

Well, it is in their accounts with Perloff or Claude.

Claude Tag目前处于测试阶段，企业客户和团队客户可以通过Anthropic访问。

So Claude tag is available in beta to enterprise and team customers through Anthropic.

这些客户在测试期间可以使用一些积分来试用Claude Tag，看看是否合用。

And there's some credits that those customers can use while it's still in beta to test Claude Tag and see if they like it.

但等它正式上线后，这些客户就会用完他们账户里原有的灵活积分或信用额度来使用Claude Tag。

But then once it's generally available, these customers are gonna or they're gonna burn through flex credits that they have or credits that they have with these accounts to use Claude Tag.

你可以想象一下，如果Claude Tag真像外界说的那样，是个能跟多个团队成员互动、在后台运行、自动跟进任务的多角色助手。

And you can only imagine if Claude Tag is this quote, unquote multiplayer agent that can interact with lots of different team members and operate in the background and re up tasks.

我猜它消耗令牌的速度会非常快——我确实听一个在Slack里用Claude的用户说过，他得提醒同事，有些事只能用Claude，其他事必须用Slackbot，因为Claude在Slack里耗令牌实在太快了。

I assume that it's gonna burn through tokens pretty quickly, which I already did hear from a user of Claude inside of Slack who particularly tells his teammates that they can only use Claude in Slack for certain things and have to use Slackbot for other things because Claude burns through tokens so quickly in Slack.

对。

Right.

现在我们来谈谈之前提到的隐私问题。

Now let's go to the privacy concerns that

我们讨论过。

we talked

关于这一点。

about.

Anthropic 能接触到所有这些 Slack 数据，这让人担心吗？

Is there reason to be concerned if Anthropic has access to all this Slack data?

我仔细看了 Anthropic 的服务条款，发现它明确禁止使用与客户签订商业协议时所获取的数据。

So what I learned by looking through Anthropic terms of service is that it does not permit the data from customers with whom it has commercial agreements.

也就是说，比如企业版和团队账户的数据。

So that is, you know, it's enterprise and team accounts, for example.

它不能用这些数据来训练模型，所以完全没必要担心它会拿这些数据去训练模型。

It can't use that data to train its models, so there's no reason to think that it's going to use this data to train its models.

不过我认为，这让Anthropic得以窥见团队在Slack上的互动方式，而这原本是Salesforce的竞争优势。

I do think, though, this gives Anthropic itself a look under the hood at how teams are interacting on Slack in a way that felt like Salesforce's competitive advantage.

现在Anthropic能看清Slack上所有这些互动是如何发生的，未来可能会据此调整产品策略。

Now Anthropic gets to see how all that's taking place in Slack, and it could inform its product decisions going forward.

对。

Right.

我问你一个问题。

Let me ask you this.

你刚才提到了Perplexity这个应用。

So, you mentioned the Perplexity app.

听起来Slack这是在下长远的棋，就像你说的，想增强用户粘性。

Sounds like Slack is playing the long game here, trying to grow the stickiness, as you said.

你觉得这个问题长远来看会怎么发展？

How do you see this issue playing out in the long run?

你认为这些整合最终会怎样？我的意思是，客户现在其实已经要为此付钱了。

Do you think eventually Do you think these integrations I mean, they already cost money to a certain extent for the customers?