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。
OpenAI的首席财务官莎拉·弗莱尔。
OpenAI's CFO, Sarah Friar.
我们得马上进入正题。
We gotta get right to it.
你刚刚完成的，是我所见过最成功的融资轮次。
You have just completed what I regard as the most successful fundraising round in history.
我们实际上要筹集超过一千二百亿美元。
We're gonna raise actually north of a $120,000,000,000.
我们认为，人工智能是我们迄今见过的最大时代。
We think AI is the biggest era that we've seen to date.
我们才刚刚开始理解它对全球生产力意味着什么，而由此带来的，或许是更高的富裕水平，以及每个人更好的生活。
We're just starting to understand what it's gonna mean for global productivity, and with that, you know, hopefully more affluence, better lives for everyone.
运气就是准备遇上机会的时候，但你必须抓住它。
Luck is whatever the preparation meets opportunity, but you gotta grab it.
听了这么久，第一次打电话进来。
Long time listener, first time caller.
能跟在座各位一起聊聊，真让人兴奋。
Quite exciting to get to hang out with all the bros here.
大家好。
Hello.
我们
We
一开始有点拿不准该怎么开场，所以我觉得最好的方式，是让咱们之前那位加密领域的老将先说几句。
weren't sure how to start this off, but I thought the best thing was to allow our erstwhile cryptozar to maybe Nice.
别急着评论。
Save comments.
今天我看了一篇文章，好像是《华尔街日报》上的，说如果是一家AI公司，早点上市反而更有优势。
I saw an article today, think it might have been in the Wall Street Journal that the perception is that there's an advantage to IPO ing earlier if you're an AI company.
现在我们知道SpaceX要上市了，接下来大家自然会问，OpenAI和Anthropic什么时候上市？
So now we know SpaceX is going And then the question is when's when are OpenAI and Anthropic gonna go?
我很想知道，你们是怎么看这个问题的？
And I'm curious, how do you think about that?
你觉得现在有点像在赛跑吗？还是说你们还没拿定主意？
Do you think there is a little bit of a race on or, you know, you haven't made a decision about that yet?
说到底，IPO，我经常跟团队说，它只是一个里程碑。
Like, in the end, an IPO, I say this to the team all the time, it's a milestone.
它不是终点。
It is not a destination.
别把公司当成是为了冲向这个终点在跑。
Do not run your company as if that's some sort of destination.
它不过是一种新的融资方式。
It's just another way to fundraise.
我们刚在三月完成了融资，你可能听我说过，筹了1220亿美元，就是为了给自己最大的灵活性。
We just did, you heard me on the the sizzle reel, raise a $122,000,000,000 in March, and that was to give ourselves maximum flexibility.
我觉得我作为CFO的职责，是为这家公司、甚至为我们所处的这个时代创造更多选择空间。
I feel like my job as a CFO is create optionality for this, not just this company, but just this era that we're living in.
所以，萨拉，那次融资是不是在SpaceX上市之前，史上最大的一轮私人融资？
So, Sarah, was that that point in fundraising, is that the biggest private republic up until the SpaceX IPO?
确实是的。
It is.
差距大得不是一星半点。
It is by orders of magnitude.
我觉得迄今为止最大的IPO是沙特阿美那年，大约300亿美元。
I think the largest IPO to date was the year of Saudi Aramco, which was about $30,000,000,000.
没错。
Right.
你能有三个IPO的规模超过2000到2001年那会儿，这简直不可思议，那时候市场也挺热闹的。
So it is actually incredible that you're going to have potentially three IPOs at a scale that will be bigger even than 2001, 2000 that that time frame, there was a lot that went on in the market too.
但现在的市场已经大不一样了。
But the market has grown.
另外，现在市场上的情况是，你看股票回购、并购这些，大量资金其实都以现金形式回流给股东了，所以市面上其实积压着不少钱。
And by the way, the other thing going on in the market is like, if you look at buybacks, m and a, and so on, it's actually a lot of capital keeps being returned back to shareholders in cash, so there is a lot of money sitting on the sidelines.
不过，回到你刚才的问题，大卫，我觉得归根结底，你最终还是要接受市场的检验。
But in the spirit of like the question, David, I think in the end, you wanna you'll be measured.
对吧？
Right?
归根结底，市场是个称重器，不是个投票器。
It's the in the end, the market is a weighing machine, not a popularity machine.
没人记得谁先上市的，是谷歌还是雅虎，是Lyft还是Uber。我说这个不是因为我想争第一还是第二，只是觉得媒体喜欢制造点戏剧性，但最终我们得打造规模大、能持续、经得起时间考验的公司，而融资正是实现这一点的关键。
No one remembers who went first, Google or Yahoo, Lyft or Uber, and I say that not because whether I wanna be first or second, but I just think it you know, the the press loves a bit of drama, but in the end, we're gonna have to build big, sustainable, durable companies and fundraising will be a key component of doing exactly that.
莎拉，有新闻。
Sarah, news.
天哪。
Oh my god.
这么多人涌进来。
So many people coming
我知道。
out I know.
确实是。
It is
同时应付四个面试官，确实挺难的。
it is it is hard balancing four interviewers at the same time.
没事的。
It's okay.
这可是我的日常。
This is my world by the way.
所以我...
So I'm
我挺适应的。
good with this.
杰森。
Jason.
Anthropic 刚悄悄提交了 S-1 文件。
Anthropic just confidentially filed their s one.
那是不是说你在申报顺序上排第三？
So does that mean you're third place in terms of the filing?
这还什么都算不上，因为你得走过SEC这一关，谁知道得花多长时间。
It does not mean anything yet because you have to run now the gauntlet of the SEC and who knows how long that takes for anyone.
是的。
Yeah.
不过他们未来会不会有优势呢？我想大家心里都在琢磨和Anthropic的竞争关系。
Is it is there though a benefit to them going forward and I think unpacking the rivalry with Anthropic is on everybody's minds.
我知道你不能多谈IPO的事，那我换个话题吧。Anthropic之前远远落后，但现在，我觉得整个行业都公认，他们在开发者、企业数量，还有收入上，已经远远甩开OpenAI了。
So, just I guess you can't talk too much about IPOs, so I'll just pivot to Anthropic was far behind and now they've really, I think everybody would agree in the industry now blown past OpenAI in terms of developers and corporations and it seems revenue.
那你们当初优势那么大，Anthropic是怎么追上来的？
So did how did that happen at OpenAI when you had such a tremendous lead?
Anthropic到底是怎么超越你们的？
How did Anthropic blow past you guys?
咱们聊聊战略吧。
So let's talk a little bit about a strategy.
我们的策略是不一样的。
Our strategy is different.
对吧？
Right?
我们在打造AI底层架构，关键是要有一个统一的基础，但对外要支持多种交互界面。
So we are building the AI layer, the infrastructure, and it's really important that there's a single foundation, but then with many interfaces out into the world.
比如ChatGPT就是面向普通用户的那个入口。
So ChatGPT is one to the consumer.
每周有超过九亿人使用ChatGPT，它已经成了一个名词，也成了一个动词。
Over 900,000,000 people use ChatGPT weekly, and it's become the noun and the verb.
对大多数人来说，这就是他们第一次接触AI的方式。
It's how most people experience AI for the first time.
有个挺有意思的数据，我们经济研究团队刚给我看的：现在增长最快的地区是非洲，这其实也不意外，毕竟基数本来就不大。
Kind of fun fact, our economic research team just showed me the fastest growing continents now are Africa, probably not totally surprising since it started a small base.
增长最快的几种语言是阿塞拜疆语和哈萨克语。
Fastest growing languages are Azerbaijani and what Kazakhstani.
应该是哈萨克语。
What is it's Kazakh.
想想它未来的发展，真是让人难以置信。
Which is kind of incredible to talk about where it's going.
所以有多个接口，比如ChatGPT，当然还有Codex，上周末用户数刚突破五百万。
So multiple interfaces, ChatGPT, of course there's Codex, just hit 5,000,000 over the weekend.
我们对此非常自豪。
We're really proud of that.
从去年一月几乎从零开始起步的。
Coming from almost zero in January.
五百万用户。
5,000,000 users.
好。
Go
Codex。
Codex.
也帮我准备一下上面这个特别环节吧。
Help me prepare for this little special up here too.
当然还有我们的企业产品Frontier，以及我们为各类规模企业提供的其他所有服务。
There's of course Frontier, our enterprise offering and everything every other way that we can get out there to reach businesses of all sizes.
这完全是另一种策略。
That is a very different strategy.
我们觉得，因为所有服务都基于同一个模型，所以能产生叠加的优势。
We think that because it's served up on one model, there's a compounding element of advantage that comes from that.
用户越多，数据越多，个性化能力越强，ChatGPT就成了入口。
More users, more data, more ability to personalize, ChatGeePTS as a front door.
随着模型越来越大，效率也会提升，从而降低每个令牌的全球成本。
As we as models get bigger, there's more efficiency that should lower the overall cost to give you a token in the world.
这会带来更高的毛利，最终有更多方式来支付算力成本，而算力的获取，目前正是最大的竞争优势之一。
That should compound to higher gross margins, ultimately more ways to pay for compute, and then access to compute is one of the really big competitive advantages at the moment.
所以我们每个人都要跑好自己的赛道，但也得意识到，我们同属一个生态，需要一起带动大家共同前进。
So, you know, we have to all run our own races, but we all have to recognize we're part of an ecosystem that also needs to bring people along collectively.
你是不是把精力分散到太多项目上了？
Did you spread a little bit too then too many projects?
大家都在讨论那个新工具Sora，反而对企业的关注不够多。
People were talking about this new gadget, Sora, and and then maybe not enough focus on enterprise.
这么说的话，去年是不是有点偏了？
Is that a fair assessment of if there was a mistake in the last year?
就只是这样？
That was it?
不是的。
No.
我觉得世界总喜欢非黑即白地划分，比如你是不是一家消费公司，莎拉？
I I think that the world loves to go to binary isms, like, aren't you a consumer company Sarah?
你是不是一家企业服务公司？
Are you an enterprise company?
实际上，我们两者都是。
The reality is we're very much both.
我们不是非此即彼。
We're not one or the other.
目前我们的收入已经相当均衡，差不多是五五开。
Right now, our revenue is getting pretty balanced about $50.50.
我们对企业的重视程度非常高。
We are incredibly focused on the enterprise.
我大部分时间都在跟企业打交道，就说上周吧，我去了波士顿的赛默飞，跟纽约的一群银行客户碰了面，周五还跟旅行者保险通了电话，今天早上又跟一家科技公司开了视频会。
Like, I spend so much of my time with I mean, just even in the last week, I could tell you I've been to see Thermo Fisher in Boston, I was with a bunch of banks in New York, I was on the phone with Travelers on Friday, I spent this morning on the phone with a tech company.
不管哪个行业，现在大家都在积极拥抱人工智能。
It doesn't matter the vertical, people are really moving on AI right now.
我们的新营收负责人丹妮丝·德雷瑟，去年十二月就上任了。
Our new head of revenue, Denise Dresser, in seats since December.
她是个极具能量的人。
She is a force of nature.
所以总的来说，企业市场现在正全面发力。
And so I think the enterprise, broadly speaking, is really firing on all cylinders.
但我们也不想忽视消费者市场。
But we don't want to leave the consumer behind.
记住，OpenAI的使命是让通用人工智能造福全人类。
Remember, our mission at OpenAI is AGI for the benefit of humanity.
不是只造福那些付得起钱的人，也不是只造福企业用户，而是面向所有人。
Not for the benefit of humanity who can pay or for the benefit of humanity who live in an enterprise, but very broad based.
所以我们提供这么多免费服务，就是想让大家先体验一下。
It's why we offer so much free because we want people to get a taste.
一旦尝到了人工智能的甜头，人们愿意投入的程度会远远超出你的想象。
Once they get a taste of intelligence, the ability to come up a commitment curve is incredible.
我们的免费用户平均每天会问七个问题。
Our free users do about seven turns, seven questions a day.
我们的第一档付费用户，使用量是这个的两倍，大约十五次。
Our first paid tier do double that, about 15.
我们的高级付费用户，也就是20美元那个档位——希望你们都在用，或者更高档的，使用量是免费用户的三倍左右，而专业版更是达到十一倍。
Our our real paid tier with the plus, $20, hopefully you're all on it or higher, about three x, and pro about 11 x over a free user.
还记得你刚用翻盖手机的时候吗？那时候你可能还觉得，嗯，能打电话就不错了。
So remember when you got your flip phone and you're like, yeah, don't know, it does make some calls.
现在，同样的手机，想想它为你做了多少事。
Now, that same phone, think of all the things it does for you.
我们如今在智能领域走的，就是这条路。
That's the path we're on with intelligence right now.
不好意思。
Sorry.
你刚才说了一件非常有影响力的事。
You said something very influential.
我觉得大概十八个月前，对我们很多业内人士来说，你提出过一个很简单的经济权衡，就是吉瓦对现金。
I think it was about eighteen months ago for a lot of us in the industry where you framed a very simple economic trade off which was gigawatts to cash.
你当时说，一吉瓦的电力，差不多相当于OpenAI每年一百亿美元的收入。
And I think you said one gigawatt is roughly equivalent to about $10,000,000,000 a year of revenue to OpenAI.
所以你第一个观点是，一吉瓦等于你每年一百亿美元的收入。
So comment number one was this one gigawatt equals $10,000,000,000 a year of revenue for you.
但这不只是你一家，因为你可以把这个数字推算到Anthropic和其他公司，比如Gemini。
But it's not just you because you can probably extrapolate that to Anthropic and other folks, Gemini.
是的。
Yep.
但你当时真的走在最前面，早早拿到了电力、数据中心和能供电的土地资源。
But then you were really at the forefront of getting access to power and data centers and powered land.
当时看起来有点疯狂，但现在回头看，原来电力供应严重不足。
It seemed a little crazy, but now it looks like, hold on, there's a huge deficit of supply.
你能详细说说吗？就是这一切是怎么回事。
Can you just unpack all of that and explain Yeah.
一方面是我们现在所处的位置，另一方面是这些经济逻辑。
Both the spectrum of where we are and then Those economics.
具体的经济模型，以及这是否真的成立。
Specific economics and if that's So
首先，没错，目前算力是非常稀缺的资源。
first of all, yes, compute is a very scarce resource at the moment.
我的意思是，我们公司的情况就是，需求正沿着一条陡峭的上升曲线猛增，但可用的算力根本不够。
Mean, I what we see in our business, we're going up that kind of vertical wall of demand right now, and there's just not enough tokens available.
我很庆幸能和格雷格和萨姆一起共事。
So I'm very grateful that I got to work alongside Greg and Sam.
我觉得我们在这事儿上算是押对了。
I think we're a press hand on this.
去年我们确实挨了不少批评，人家问为啥要买这么多算力，但现在回头看，真得感谢我们当时做了这个决定，因为到2026年，算力还是不够用。
And last year, we were definitely taking some, you know, arrows in the back about why are they out there buying all this compute, and I think, thank God we did, because in '26, we still won't have enough compute.
我们现在在算力发展这条路上到底走到哪儿了？
Where are we on the compute continuum?
各个环节都有瓶颈，而且这些瓶颈还会来回切换。
There's kind of choke points everywhere, and I think they will continue to move back and forth.
你们也都清楚，这事儿大家讨论得不少——首先是能源，然后是土地和电力，还有怎么让监管环境允许我们快速建设。
I mean, you all talk about this and know this as well as anyone here, whether it's energy first and foremost, land power, how we get regulatory environments such that we can build quickly.
到了机架和芯片本身，供应链里的产能到底够不够？
When you get into the racks and chips themselves, clearly, do we have enough in that supply chain?
目前内存供应特别紧张。
Memory spike is is on at the moment.
优秀人才的获取
Access to great talent.
我们的教育体系能输送足够的人才吗？
Do we have enough people coming through our education system?
我现在真的很担心这个问题。
I really worry about this right now.
我是斯坦福的校董，我亲眼看到，我们必须持续重视教育和科学。
I'm a trustee at Stanford, and, you know, I see just that, you know, we need to keep the focus on education and science.
还有信任。
And then trust.
我其实把信任也看作供应链的一部分。
I mean, I actually put that as part of the supply chain.
萨姆现在在密歇根州的塞琳。
Sam right now is in Celine, Michigan.
他再过两个小时就要剪彩了，所以我提前跟你透露一下，不过他们允许我在房间里说这事。
He's gonna be cutting the ribbon in about two hours, so you are getting a sneak preview, but they told me it was okay to say it in the room.
这将意味着在我们甲骨文园区内的一个十亿瓦的数据中心破土动工。
That will be, you know, sticking shovels in the ground on a one gigawatt data center, which is part of our Oracle complex.
在信任这一点上，我们绝不能让任何社区掉队。
Really important there on the trust side that we don't leave communities behind.
我在Nextdoor工作了七年，深入体会了什么是真正扎根本地，你不能自上而下告诉人们他们需要什么，因为他们嘴上说谢谢，心里却说不用了。
I spent seven years of my life working at Nextdoor, doing the hard work of what it means to be local, and you cannot tell people from top down what they need, because they will tell you thank you, but no thank you.
我会直接告诉你我需要什么。
I will tell you what I need.
所以在这样的数据中心项目里，我们花了很多时间跟社区沟通：第一，我们不会提高你们的电费，我们的基础设施和电力成本由我们自己承担。
And so in a data center like that, we're actually spending a lot of time in the community saying, number one, we're not going to raise your electricity bills, we're going to pay for our infrastructure and our power.
绝对不会让普通用户来买单。
It will not be the rate payer that has to pay.
第二，我们会带来两千五百个工会岗位，都是好工作，比如电工、暖通空调技术员等等。
Number two, we're going to bring jobs, 2,500 union jobs, good jobs, like electricians, HVAC, and so on.
我们会缴税，仅这个数据中心就会为密歇根州带来十亿美元的税收，此外，我们还会额外投资四千五百万美元，用于教育领域，支持编码学分项目，这正是你们这个周末讨论的内容。
We are going to pay our taxes, a billion dollars in taxes just for that data center into Michigan, and on top of that, we're going to invest $45,000,000 going into education for codex credits to do what you all talked about this weekend.
现在 anybody 不会用 Codex 就来上班，这简直就像我不会雇一个连 Excel 都不会用的财务人员，今天我几乎不会雇任何一个不懂 Codex 这类工具的财务人员。
It's like anyone who's not coming in facile to their new job, I have teenagers using codex, it would be like I would never hire a finance person who didn't know how to use Excel, and I pretty much probably wouldn't hire a finance person today that doesn't know how to use a tool like Codex.
所以当你谈到投资时，我们必须提前布局，走在需求前面。
So that you know, so when I think about investment, we're having to invest ahead of demand.
这意味着我们得同时找到足够的算力和相关资源，还得承担起相应的成本。
That means we need to both be able to find all of the compute and all the pieces and then pay for it.
这就回到了你之前问的关于上市的资本问题。
So that goes back to your capital question on IPO.
另一方面，从经济角度看，情况确实在持续改善。
And then on the other side on the economics, look, the economics do continue to get better.
多个方面都在向好发展。
They're getting better on multiple fronts.
我觉得我们正在更有效地向客户展示真正的价值。
I think we are doing a better job of actually showing true value to our customers.
而且我觉得，定价已经不再只是成本加成，而是更贴近实际创造的价值了。
And I think you get beyond kind of a cost plus type pricing into something that feels more akin to the value being created.
现在，令牌的稀缺性起到了作用，因为它让成本出现了一定程度的压缩
Now, scarcity of tokens helps because it's causing a bit of a compression
成本确实下降了。
in costs.
不用点名，但你知道，当前整个行业的算力供给和各方需求的格局是摆在那里明摆着的。
That in just like without specific names, where you know the landscape exists today in terms of all the power that's available and all the demand that exists across everybody.
没错。
Yep.
如果按现在的速度和现有资源继续下去，明年会发生什么？
What's going to happen over the next year just at the current course and speed of what is available?
现有的数据中心、可用的令牌、整体基础设施，对所有人来说都够用吗？
Of the data centers that's available, of the tokens that's available, of the infrastructure that's available, for everybody.
我上周讲过这个故事，但再说一遍，像Anthropic这样的公司，有时候到了晚上十点半，系统就直接提示：‘Chamath，我们两点半再聊吧。’
Because, you know, I told this story last week, but you know, I'll use Anthropic and one of the frustrating things is at some point it just says, you know, 10:30, it's like, alright, Chamath, see you at 02:30.
是的。
Yeah.
这体验根本行不通。
And that's not a viable experience.
没错。
Right.
不过说句公道话，ChatJPT 我从来没遇到过这种情况。
And in fairness to ChatJPT, actually, I've never had that with
是的。
Yeah.
我们的令牌额度给得相当充裕。
We we're quite generous with our tokens.
而且我们是有意这么做的，就是为了让更多人能接触到。
And again, on purpose, we're trying to drive access so people understand.
即使你用的是免费版，用的也不是最新模型，但我们就是想让你亲手试试，感受一下。
Because if you're on that free tier, not actually getting the latest model, but we're trying to put it in your hands so you get a sense for it, by the way.
你想啊，小时候我在北爱尔兰一个小镇上，那儿正处在动荡时期，家里突然送来一套《大英百科全书》，那感觉就像乌云散开，阳光照进来一样。
Because, you know, if you're a kid doing homework, like I think about when I grew up in the Encyclopedia Britannica's showed up at the front door in Northern Ireland in a tiny little community in the middle of the troubles, it was like the clouds parted.
所以我们希望确保大家能有这种体验。
So we want to make sure that people get that feeling, by the way.
但到了2026年，如果你想买更多算力，祝你好运吧。
But the landscape right now, in '26, if you want to buy more compute, good luck to you.
说真的，你倒是告诉我，我上哪儿去找呢？
Like, tell me because I don't know where else to find it.
你知道的，埃隆那边倒是有点东西。
I mean, as you know Elon has something.
我本来想说，讽刺的是，埃隆反而是那个算力多到用不完的人，不过他能想办法把这部分卖出去，也挺厉害的。
What I was going to say, Elon ironically ended up being the one person that had too much compute in a way, but good job on figuring out how to sell that off.
到了2027年，情况也还是挺紧张的。
In '27, it's pretty limited as well, frankly.
不过现在有几件事正在发生变化。
Now, there's a couple of things shifting around.
我们说算力的时候，训练这部分目前基本都还在美国境内进行，主要是出于国家安全的考虑，确保这种关键资产留在美国本土。
When we talk about compute, there's training that mostly still all happens here in The United States for USG reasons, for making sure that a national asset in effect is happening on US soil.
推理部分，我们希望它能覆盖全球，尤其是在智能体时代，大家更需要实时响应。
For inference, we want that to be global, and I think particularly in an agentic world, you want much more kind of real time.
像Sora这样的视频应用，说实话，我们当时不得不做了一个艰难的决定，因为算力实在不够用。
Even for things like Sora and video, which by the way, yeah, we have you know, we had to make a really tough choice because we didn't have enough compute.
我们已经用得差不多了。
And we've used
这个我们用得很多。
this a lot.
现在视频确实占不少，但视频的潜力远没发挥完。
Right now, yeah, video does, but video is not over.
特别是当你开始思考，AI正把我们带向更丰富的多模态交互。
Like, in particular, when you start to think about where AI is taking us into more multimodality.
别忘了，上一代技术教会我们，说话全靠拇指打字。
So remember, we've all been taught by the last generation of technology to talk with our thumbs.
这简直成了一种病。
It's a disease.
你走在街上，人人都低着头，再也不肯抬眼看路了。
You walk around, everyone's looking down, they don't look up anymore.
晚上，青少年们坐在我沙发上，用拇指互相聊天。
Teenagers sit on my sofa at night and talk to each other with their thumbs.
我就问：你跟谁聊呢？
I'm like, who are you talking to?
我儿子会说：就是他。
And my son will be like, him.
我就说：行啊，那你聊吧。
I'm like, okay, talk.
多模态时代已经来了。
Multimodality is here.
希望你们这个周末都讨论过这个话题。
Hopefully, I think you all talked about it this weekend.
你们现在是在跟自己的工具对话。
You're talking to your tool.
我每天都会和编码工具对话，这变化得特别快，但要实现这种体验，需要更多实时计算能力——要是我跟查马斯聊天，那感觉可就怪了。
I talk to Codecs every day, and so that is changing rapidly, but that is gonna need much more kind of real time compute because it's an odd experience if I was talking to Chamath.
约翰尼，把这个冰球拿去吧。
Johnny, have this puck Totally.
这些耳机。
These earpieces.
那你能不能跟我们说说这个过程是怎样的？
So maybe tell us a little bit about that process.
你现在总算承认了。
You've admitted it now.
我要是跟你说这是个耳机，约翰尼立马就会跑来把我的儿子抢走。
If I if I tell you it's an earpiece, Johnny will come and steal my teenage son.
我或许真会把它给他，把儿子也一并给他，但你得相信，确实有这种可能。
I might give it to him, give him to But do believe that there
应该有一些
should be some
我们正在转向一种消费级的底层技术，现在还不能透露具体是什么，但今年年底我们会正式公布。
We're changing into a consumer substrate that I cannot tell you what it is, but by the end of this year, we will unveil it.
我亲眼见过。
I have seen it.
我亲自试过。
I've tried it.
我是个爱用手势说话的人。
I am a hand talker.
现在我就把双手放在腿上，不敢动。
Right now, I'm sitting on my hands.
Did
Did
an you iPad?
an you iPad?
Have an iPad shift.
Have an iPad shift.
什么时候啊。
When when yeah.
你用的时候，是不是像第一次拿到iPhone那种感觉？
When you used it, was it like having an iPhone for the first time?
约翰尼和团队最擅长的就是给设备注入人情味，我很难说清楚到底是什么，但你一看到就能感受到。
It's very what Johnny and team are really good at is bringing humanity to devices, and I don't really know how to explain that well, but when you see it you feel it.
某种意义上，感觉很自然？
It feels natural in some way?
特别自然，而且让人忍不住喜欢。
It feels very natural, but it feels very lovable.
真的吗？
Really?
我说不清那种感觉到底是什么，因为实在太...
And I can't really explain what that emotion is because In so much
某种程度上是因为
some way in terms of because
科技不会让你
technology not taking your
把手机掏出来，整个过程非常顺畅，这是那些玩过这项科技的人跟我说的。
phone out and it's it's seamless is what I've heard from people who've Technology played with
科技有时候会显得特别机械，但我们都明白，优秀的设计能让一切变得无感。
is very can be very mechanistic, but we all know great design just makes everything fade away.
对吧？
Right?
这就是当时所说的，简单其实最难。
It's what at the time, you know, the simple is hard.
我觉得这非常
I I think this is a very
乔布斯说的。
Job said.
回到之前那个问题，现在请换上CFO的角色，帮我们理解一下你们的资金配置模式。
This story, just going back to the earlier question, so putting on the CFO hat, help us understand the capital allocation model that you use.
过去一二十年，很多企业能取得远超平均水平的回报，是因为它们找到了独特的方式，以更高的资本回报率来配置资金。
Because a lot of businesses over the last decade, two decades that have kind of been these outsized returners have found some unique way to deploy capital at a higher ROC than anyone else.
然后你就会把所有资金都投向那些回报率更高的领域。
And then you end up plowing all your capital into that higher ROC bucket.
对你们来说，这个领域是什么？你们怎么看待这种投资组合策略，去更多地布局这类高回报项目？有没有一个机制，能让这种能力随着时间不断提升？
What is that for you guys, and how do you think about that portfolio approach to having more of these kind of big returner shots, and is there an engine where that gets better over time?
肯定有，因为最终，在这个时代诞生的持久且高价值的公司，不会是什么魔法般的存在。
There has to be, because in the end, the durable, high value companies created in this era, I don't think they're not going to be magical.
它们会像以往时代的优秀企业一样，实实在在地创造价值。
They're going to look like the great companies of prior eras.
它们会为客户带来真正的价值。
They're going to create customer value.
一切从客户出发，真正帮助客户做得不一样、更好、赚更多钱、更高效。
Starts with the customer and really helps the customer do something different, better, more revenue, more efficiency.
对吧？
Right?
赛默飞希望加快患者筛查的速度，以便更快获得FDA批准。
Thermo Fisher wants to be able to get patient screening done faster so they get FDA approval faster.
这真的非常关键。
That's really important.
比如，如果你患的是那种只剩几周生命的癌症，突破性进展早两周还是晚两周，可能直接关系到生死。
Like, if you have a form of cancer where you have weeks to live, the difference between a breakthrough in four weeks and two weeks can literally be life or death.
他们还有大约三万到三万八千名销售人员遍布各地——我可能记不太准数字，但你走进国内任何一个实验室，到处都能看到赛默飞的设备。
They also have I'm going to misquote this, but something like 30,000, 38,000 people in the field selling those amazing like, if you walk into any lab in the country, you'll just see Thermo Fisher plastered all over every device.
这些销售人员也希望提高工作效率。
Those people want to be more efficient going to work.
目前OpenAI内部增长最快的工具，其实是在我们的市场销售团队中使用的。
Like, the the fastest takeoff of codecs within OpenAI right now is actually in our go to market team.
开发团队也在用，但如果你看月度增长数据，几乎全部都来自市场销售团队。
Our devs are there, but, like, if you look at the pace of growth, kinda month over month, it's all in GTM.
所以他们希望市场销售团队的产出能更高。
So they want more productivity out of their GTM team.
当然，他们在金融等领域也在做些事情，这让我特别兴奋。
And of course, they're doing things in areas like finance, which I get really excited about.
所以，客户价值永远放在第一位。
So customer value first.
接下来，你得想办法实现很高的毛利。
From that, now you need to get to a great gross margin.
那怎么才能做到高毛利呢？
So how do you get to a great gross margin?
你要看收入成本，也就是运营开销。
You're looking at, like, the cost of revenue.
最大的成本就是计算资源。
The main input is compute.
好消息是，计算资源的成本正在大幅下降。
The good news on compute is that there is a massive deflationary curve on cost.
对吧？
Right?
从ChatGPT 5到5.4，我觉得成本下降了将近97%。
From ChatGPT five to five four, I think the deprecation cost was something like 97%.
这个下降曲线太惊人了。
It's like an amazing curve.
其实从4到5.4，成本也是降了97%。
Actually, I'm slightly from four to five four, it was 97%.
但这一切只用了短短两年时间。
But that happened in like two years.
这简直让人惊叹。
That's kind of wowing.
对吧？
Right?
确实很
That's
了不起。
incredible.
就连我们最新的模型，比如刚发布的5.5，我们现在也在努力把这种改进回馈给客户。
Even our newest model, if you look at five five that we just released, we're trying to now translate that back to the customer.
所以我们实际上把5.5 2x的价格上调了。
So we actually raised prices on five five two x.
但如果你看客户的实际成本，他们每处理一个token还是能省下大约20%到30%，因为效率提升太多了。
But if you look at what the cost of the customer is, they're probably still getting a break of about 20 to 30% cost reduction per token because it's just much more efficient per token.
这方面还有很多工作可以做。
So there's a lot to do in that envelope.
对的。
Yep.
做资本配置决策时，如果只按现在的成本来算，很可能把结果定价错了。
And and part of making an alloc a capital allocation decision is having to if you make it on today's cost profile, you actually might misprice the outcomes.
你得稍微往前看一点成本趋势。
You have to lean in a little on the cost profile.
当我们考虑未来的产品规划时，我现在的重点是：从28号开始，我能买到什么样的算力？
And then as we think about, like, the builds, yeah, you are having to make like, really, my focus today on compute is what's the compute I can buy for 28 onwards?
比如塞琳那座密歇根的数据中心，我觉得我们至少要到2027年底或2028年初才能用上那里的算力。
Like that Michigan data center in Celine, I don't think we will be getting compute out of it until probably end of twenty seven or early twenty eight.
所以你现在就得开始下注了。
So that's where you're starting to make your bets.
事实上，我现在最缺算力的地方，是已经得开始考虑2030年、2031年、2032年的事了。
And in fact, where I feel most short of compute right now is starting to look at thirty, thirty one, 32.
所以你得先构建一个商业模型。
So you're having to create a business model.
不过好消息是，每过一年，我们对这些项目的信心就更强一点。
Now, the good news is each year goes by, we get more confidence in the build.
实际运行效果远超预期，这让我们越来越有把握，市场也正越来越向我们靠拢。
We're seeing it massively outperform, and so that's giving us more and more confidence, and the market is coming towards us much more.
玛拉，你们是怎么预测未来好几年的算力需求的？
Mara, so how are you making the compute need forecast multiple years out
对。
Yep.
你们得考虑所有正在发生的架构和模型进步，也就是每单位能耗带来的价值或效用在提升。
Accounting for all of the architectural and model advancements that are happening where, call it, value or utility per unit of power is going up.
帮我们理解一下，你们是怎么估算这个的？毕竟现在技术发展太快，不确定性很高。
And help us understand how you kind of estimate that given that there's a lot of technology development going on that has a high variance to it.
是的。
Yeah.
对。
Yeah.
所以我们必须对算力本身做出多种假设。
So we we do have to make multiple assumptions both on the compute itself.
目前我们假设，每吉瓦的算力成本实际上在上升，因为电力、内存等各种成本都在涨。
So we assume right now that compute actually on a per gigawatt is getting more expensive because power is getting more expensive, memory is getting more expensive, and so on.
但另一方面，由于芯片技术的进步，我们获得的算力提升完全抵消了这些成本增长。
However, the the intelligence that we get on the other side out because of the deprecation on the chip side is is more than making up for that.
所以对客户来说，每单位算力的成本实际上会大幅下降。
So in terms of a per unit sold to a customer, it should actually get a lot less expensive
这方面有所提升。
improvement in that.
所以这仅仅是...
So that's just
这指的是芯片本身。
That's just the chip itself.
我们在模型方面不会高估，因为像五五这样的模型在能效上确实非常出色。
We don't try to overestimate on the model side because sometimes like five five is an incredibly good model on the efficiency side.
但如果你看前一代的五四模型，它是个规模很大的预训练模型。
But if you look at something like five four, the prior model, it was a really large pretrained model.
它的成本非常高。
It was very expensive.
实际部署起来也很困难。
It was actually hard to serve.
有时候我们确实需要这种大规模的预训练模型。
And sometimes we wanna do that really big pretrained moment
是的。
Yeah.
然后我们通过多次模型迭代来降低成本。
And then we take multiple model turns to be able to kind of drive down on the cost side.
我的意思是，在近期，比如2026年和2027年，我会从底层开始构建一个模型。
I mean, in the in the near term, like in '26 and '27, I clearly build a model that's bottoms up.
我知道自己的产品是什么，对定价也有把握，用户数量乘以单价，我能估算出会有多少惊喜点，也能看到收入曲线的形态——有多少人会订阅，广告收入也和每周活跃用户、每日活跃用户、消息量这些指标息息相关。
So I know what my products are, I have a sense of what the pricing will be, you know, consumer, p times q, how many wows do I think I have, I can see what the shape of the line is, how many of them will subscribe, advertising coming in is also still related to how many weekly actives, how many dailies, how many messages, and so on.
所以在2026和2027年，你其实可以做出相当准确的模型预测。
So you can you can do actually a pretty good model job in '26 and '27.
但即便如此，收入曲线的走势还是经常超出预期，往更高的方向走。
That said, the shape of the line keeps taking us by surprise to the upside.
到了更远的年份，你更多是在看已经采购的算力，然后反过来用一种算法推算：这么多算力大概能对应多少收入。
When you get into the outer years, you're actually looking more at the compute you've bought and almost just doing an algorithm the other way that's saying, this amount of compute should equate somewhat to this amount of revenue.
我并不确定最终收入会具体来自哪里。
I don't know for certain exactly where it will all come from.
一年前，我给投资者做了一个模型，预测代理式收入。我当时说，我们即将进入代理时代，开发者只要用自然语言就能构建应用，我们估计他们每月愿意付两千美元，现在回头看简直可笑，但当时没人相信。
Like a year ago, I built a model for investors that showed Agentic revenue, And the story was, we're going to have this thing, we're going to be in the Agentic era, we're going to hand it to a developer with natural language they're going to be able to build, and we think they will pay upwards of maybe $2,000 a month for it, which is kind of laughable in hindsight, but nobody believed.
他们都说：‘我不知道她在讲什么。’
They were like, I don't even know what she's talking about.
这根本不可能发生。
There's no way that will happen.
每月两千美元？
$2,000 a month?
还记得当初大家为ChatGPT Pro每月200美元就疯掉的时候吗？
Remember when people were losing their minds over ChatGPT Pro being at $200?
天哪，这太离谱了。
Like, oh my god.
根本没人会为这个买单。
One will ever pay for that.
那为什么是1220亿美元？
So why a 122,000,000,000?
这会推到2031年、2032年吗？
Does it take you to 2,031, 2,032?
你在做这种预测时，是怎么算出资本需求的？
Like, how do you get the calculus on the capital needs as you do that modeling?
也许我们可以更具体一点。
You know, maybe even more specific.
我看到的估算显示，建一个吉瓦级别的AI算力中心，光是土地、电力、建筑和芯片这些，加起来就要大约500亿美元。
So the estimates I've seen is that to stand up one gigawatt of AI compute costs about $50,000,000,000 in land, cap power, shell, chips, everything.
总共差不多500亿美元。
All in around 50,000,000,000.
建一个新的数据中心，你得一次性垫付全部这笔钱吗？
Do you have to front all of that money when you create a new data center?
还是说你只出一部分？
Or how much of it do you do?
其中有多少能靠贷款解决？
How much of it can you get debt for?
一千亿美金，是只能建两吉瓦的算力，还是能建五吉瓦？
Does a 100,000,000,000 raise only get you two gigawatts or does it get you five?
也就是说，这笔钱到底能换来多少？
Like, what does It's it get
这个问题问得特别好。
it's a great question.
如果我们看看我们的算力战略，就会发现世界变化之快简直令人咋舌。
So if you look at our compute strategy, and it's crazy how fast the world has changed.
就在两年前，我们实际上只有一家客户。
So just two years ago, we were literally one.
我们当时只跟一家云服务提供商合作，就是微软的 Azure。
We had one CSP we worked with, Microsoft Azure.
芯片方面，我们也只用英伟达的。
We we sat on one chip, NVIDIA.
产品也只有一个，就是 ChatGPT，定价每月 20 美元。
We had one product, ChatGPT, one price point, $20 a month.
展开剩余字幕（还有 108 条）
所以我常拿魔方来做个比喻，那时候我们就像只有一个底层的方块。
So I often use a Rubik's cube as kind of my metaphor, so we were like one cube in the bottom.
今天，如果你看我们的策略，首先就是跟多个云服务商合作，因为云服务商帮我们把资本支出转成了运营支出。
Today, if you look at our strategy, it's been to go, first of all, multiple CSPs, because what CSPs do for us in effect is they shift CapEx into OpEx.
也就是说，你随着收入增长才付费，真正用到数据中心的时候才花钱。
So you pay as you get the revenue, so as you're actually utilizing the data centers.
换句话说，我们某种程度上借力于他们建数据中心和融资的能力。
So in effect, we are riding somewhat on their ability to build and have CapEx and financing.
现在，我们的系统已经覆盖了所有主要云服务商：甲骨文、Core Weave、微软、谷歌云、AWS，还有不少新兴的小型服务商。
So today, we sit on top of every CSP, Oracle, Core Weave, Microsoft, GCP, AWS, and a bunch of small neoscalers.
在芯片方面，我们也采取了多芯片策略，就是为了确保我们始终站在技术前沿。
On the chip side, we've also gone for a program of being multi chip because we wanna make sure you're always on the frontier.
我觉得如果只依赖单一芯片，迟早会掉队，因为技术上总会有跨越式的发展。
I think if you're only on one chip, there's just inherently a moment where you can't be on the frontier because there's some leapfrogging that happens.
所以现在，NVIDIA 依然是我们最重要的合作伙伴。
So today, NVIDIA remains our absolute priority partner.
他们掌握了前沿芯片技术。
They have the frontier chip.
我们今年秋天最大的一次采购将转向维拉·鲁宾的系统。
Our next big trading run-in the fall will be done on Vera Rubins.
我们对此非常期待，现在正规划接下来的费曼系列产品。
We're really excited about that, and now we're plotting kind of the Feynman series that's coming.
但我们现在也从AMD获得了在途的芯片。
But we also now have chips in the pipeline from AMD.
Cerebras的芯片已经上线了。
Cerebras is already online.
它是一款超低延迟的芯片，特别适合需要实时编码的开发者，此外我们还在和博通合作开发自己的芯片。
It's been an incredible low latency chip, great for devs, for example, that want real time coding, and there's our own chip that we're working on with Broadcom.
除此之外，我们还通过其他方式实现了多元化布局。
And then beyond that, there's other ways we've diversified.
现在，再想想那个魔方。
So now think about that Rubik cube.
现在情况变得多维多了，这让我们能有效利用信用评级高的云服务商，加快节奏，把支出从资本性支出转为运营性支出。
It's become much more multidimensional, and it allows us to effectively utilize investment grade CSPs in order to be able to go fast and push it back to be more OpEx, not CapEx.
现在，我们开始转向更定制化的建设模式。
Now, we are starting to shift gears into more of a built to suit type environment.
我们刚宣布了和软银能源在德克萨斯州合作建设的数据中心。
Announced a data center we're building with SoftBank Energy down in Texas.
这标志着我们迈出了超越传统云服务商的第一步。
That's the beginning of something that's beyond a CSP.
那里需要投入更多的资本性支出。
There's a little bit more CapEx required there.
最后，随着世界不断向前发展，别忘了我们只用了两年就做到了这一切。
And then finally, I think as the world progresses, remember we've done all that just in two years.
我喜欢用魔方来打比方，因为——别让ChatGPT替你解释——魔方能组合出大约一千万亿种不同的形态。
The reason I like a Rubik's cube is, again, please chat GBT this, but I think a Rubik's cube has something like a quintillion different forms it can come up with.
没错。
Yeah.
这就给我们带来了很大的灵活性。
And so it just gives us a lot of optionality.
还记得我之前说的吗？我的目标就是最大化灵活性。在还没达到投资级信用、无法以低成本获得债务融资的时候，和合作伙伴一起运作就显得尤为重要。
Remember what I said, my job is maximum optionality, and in a moment where I'm not yet an investment grade type of entity where I can go get lower cost debt financing, being able to work with partners to do that is really important.
先生，您觉得五年后，整个技术栈会不会完全融合在一起？
Sir, do you think that in five years from now, the stack is just merged together?
我什么意思呢？
What do I mean?
在过去传统的市场里，英伟达只卖芯片，仅此而已。
In traditional or historical markets, you'd have NVIDIA sell the chips, but that's all they do.
微软呢，就只做云服务，别的什么都不碰。
And then you'd have, you know, Microsoft just run a cloud, that's all they would do.
而消费者应用公司，也只专注做自己的应用。
And then you would have a consumer app, that's all you would do.
但现在，我们看到每个人都在做所有的事情。
But now we see everybody doing everything.
你们自己在研发芯片，也开发模型，未来说不定也会决定自己做一个新型云服务。
You know, you guys have silicon that you're spinning, you have models that you make, you may or may not eventually decide that you need to be some form of a neo cloud yourself.
看看英伟达，他们的芯片实力超强，还推出了自己的开源模型，现在也越来越像一个主要客户了。
If you look at Nvidia, they have incredible silicon, but they also have their own open source models, they're increasingly becoming an off taker.
谷歌首先是家云公司，但现在也有了自己的芯片和模型。
Google is a cloud company first, but they also have a chip, now they have models.
所以这一切都在融合。
So it's all merging.
如果这种趋势继续下去，竞争格局是变得更简单了，还是更容易了？
If that continues to happen, does that make the competitive landscape simpler or easier?
你想啊，大家都想抢占离客户最近的那个层级，因为那里通常能拿到生态系统里最大的利润。
I mean, think where everyone is trying to make sure they reside is the layer that is closest to the customer where usually you take the the largest portion of the profits of the ecosystem.
对吧？
Right?
没人愿意被边缘化。
No one wants to find themselves abstracted.
池子。
Pools.
远离。
Away.
是的。
Yeah.
当然。
Absolutely.
所以这就是为什么，当我现在思考我们的定位时，还是会回到我一开始说的：我们想成为那个AI智能层，因为一年前人们还在谈大语言模型的同质化。
And so that's why today when I think about our positioning, comes back to where I started, why we want to be that AI intelligence layer is because a year ago people talked about the commoditization of the LLMs.
但说实话，情况恰恰相反，因为当你开始构建代理层时——我们都开始用‘整合’这个词，但真正带来上下文和记忆的，正是这个整合层。
And frankly, it's gone the opposite because as you start building an agentic layer, and we've all started to use this word harness, but the harness is what brings the context, the memory.
对吧？
Right?
在我的代码库中，我有个超大的记忆文件，它知道我就是我。
I have in my codex, I have a whole ginormous memory file where it knows that I'm me.
它知道我是OpenAI的CFO，知道我平时怎么写东西、怎么表达，还知道我关心什么。
It knows I'm the CFO of OpenAI, it knows how I like to write things, how I like to say things, it knows what I'm interested in.
它甚至知道我是个妈妈，有个青春期的孩子。
It actually also knows that I'm a mom, teenagers.
它就这么记着这些事，让我用起模型来感觉更得心应手。
I mean, it just carries all this memory, and that makes the model more powerful for me.
现在想想，如果这种记忆和上下文被带到企业环境中，会怎么样？
Now think about what happens when that memory in that context is brought into an actual enterprise environment.
这已经不只是关于存在那里的数据了，我总想起当年在华尔街的时候，那时候满世界的数据都在告诉你，公司发布财报后股价应该怎么走。
So now it's not just even about the data that resides there, but I always think about the the intuition of, like, back when I worked on Wall Street, right, there was all the data in the world that told you what a stock should do post an earnings call.
但你等一下，你给交易员打个电话，他就会说，不对，这股票不会涨，Sarah。
But, give me one second, then you called your trader, and the trader would be like, yeah, stock's not going up, Sarah.
我就会问，你什么意思？
And I'm like, what are you talking about?
所有数据都显示股价该这么走，可他却说，我知道，但我清楚那个基金现在压力很大，得赶紧抛售持仓，这周股价肯定得跌。
Like, all the numbers say it did this, did this, this, and he's like, yeah, I know, but I know this fund is under pressure, and they need to sell down their book, and that is going to kill the stock for the next week.
对吧？
Right?
这才是企业的直觉。
That is the intuition of an enterprise.
我总是拿这个当例子，因为我以前在金融圈工作，但生活中每个领域都有这样的直觉。
Like, it's the best example I always think of because I came out of a financing world, but there's this intuition in every walk of life.
我觉得现在的模型正越来越能连接到你公司内部的记忆、上下文和直觉，这正是让CEO和高管们兴奋的地方——他们终于看清了，这不仅能实实在在提升收入，还能提高效率。
And that's where I think the models are now getting very connected to the memory and context and intuition of your company, and that's what gets CEOs and c suite really excited because they're like, okay, now I really see how this is going to add value to drive my revenue line, my top line, but also, you know, I can think about it as an efficiency play as well.
所以回到你刚才的问题，我觉得大家最关心的是，如何尽可能贴近这种价值。
And so back to what you're asking, I think what people want to make sure is they stay as close to that value as possible.
同时也要有足够的灵活性，随时调整方向。
And be flexible enough to pivot as you need.
没错。
Yeah.
但是
But
不好意思，杰森。
Sorry, Jason.
别客气，完全没问题。
It's quite alright.
整个过程非常棒，你对细节的处理太到位了。
Been wonderful and you've been so great with the details.
最后再问一个细节问题，快速一下。
One final detail question, rapid fire.
我们这一代最出色的三家消费类企业，分别是iPhone、Meta和谷歌的广告网络。
Three great greatest consumer businesses of our lifetime, iPhone, Meta, advertising network and Google's advertising network.
这三家里有两家是靠广告盈利的，就连苹果也有一点广告收入。
Two of those three are ad based and even Apple has a sprinkling of ad ads.
我还没怎么听你聊过广告的事。
Haven't heard you talk about ads much.
有人说在体验版和免费版里已经看到一些广告了。
People tell me they're seeing some ads in the experiment and the free version.
你对广告版本的承诺是什么？
What is your commitment to the ad version?
你们在超级碗期间被Anthropic给耍了一把。
You guys got a little trolled by Anthropic during the Super Bowl.
哦，你们是要放广告的。
Oh, you're going to have ads.
但广告真的是让这款产品全球免费的解决方案吗？
But is ads the solution to making this free for the world?
是的。
Yeah.
首先，关于广告，我们想坚持自己的原则。
So first of all, on the ad front, you know, we want to stick by our principles.
我们希望确保你始终得到的是模型最优质的结果，而不是受赞助的内容。
We want to make sure that you know you're always getting the best result based on the model, not by something that was sponsored.
这一点必须坚持。
So that has to hold true.
第二点是，我们始终会提供一个无广告的免费版本，给那些不想看广告的用户。
And I think the second thing is that we'll always provide a free a tier, sorry, an ad free tier for people that just don't want ads.
这样的话，你就明白了。
With that You're said okay.
如果菲吉说得特别到位的话，
If you took if you took Fiji says this really well.
要是谷歌和Meta生个孩子，那肯定就是ChatGPT。
If, you know, Google and Meta had a baby, it would be ChatGPT.
因为你看谷歌搜索，顺便说一句，我们至少占了11%的搜索市场份额。
Because what you have in Google search, and by the way, we know we have at least 11% of the search market.
实际占比更高，因为当你在谷歌上搜索，页面刷新一次就算一次。
It's a lot more because actually, when you do a Google search and the page refreshes, that counts as one.
但在ChatGPT里，哪怕你连续问五十个问题，整个对话也只算一次。
In ChatGPT, when you do a whole conversation where you might ask 50 questions, that also only counts as one.
所以实际上，我们的用户意图要高得多。
So in reality, we have a much higher portion, very high intent.
这对广告主来说太棒了，因为我直接告诉你我在做什么。
That is great for advertisers because I'm effectively telling you what I'm doing.
对吧？
Right?
我想买一双特别酷的鞋，就放在舞台上。
I want really cool shoes to sit on the stage.
我等于在告诉你我想买什么。
I'm telling you what I want to go buy.
在Meta那边，他们靠的是类似‘你这样的人’这种兴趣标签，来定位人群画像。
In Meta's case, right, they use this, like, people like you sort of intent so they have the demographic.
但我们更胜一筹，因为我们有记忆。
We have more than that because we have memory.
对吧？
Right?
我刚说了，它知道我是谁。
I just told you it knows who I am.
所以，想象一下把记忆和上下文跟用户意图结合在一起，你会得到一个非常强大的广告平台，能够让你大规模地触达全球市场，因为现在你有能力为此付费。
So imagine putting memory and context next to intent, you should have a very potent ad platform, which gives you an ability to offer up massive access to the world writ large because now you can pay for it.
我想起你之前问过的一个问题，弗里德伯格：如果现在只看每个令牌的收益，为了眼前利益，我会把所有令牌都给API。
And I think back to a question you asked, Friedberg, like if you look at the revenue per token right now, if I was optimizing only for today, I would give every token to the API.
对。
Right.
全部给API。
Every token to the API.
比给消费者的多出一个数量级。
Order of magnitude more than to the consumer.
但我跟你说过，我们在玩自己的游戏。
However, I told you we're playing our own game.
我们有一个战略，相信AI基础设施就像电力一样是一种公共事业，未来你必须有能力服务整个世界。
We have a strategy where we believe there's an AI infrastructure layer, a utility like electricity, and in a future state, you'll want to be able to serve the world writ large.
无论是消费者、小企业、大型企业，还是政府。
Consumers, small businesses, large enterprises, governments.
这就是我们的策略。
That's our strategy.
女士们、先生们，有请OpenAI的首席财务官莎拉·弗莱尔。
Ladies and gentlemen, the CFO of OpenAI, Sarah Friar.
讲得非常好。
Well done.
太精彩了。
Fabulous.
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。