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。
目前最大的赢家之一。
One of the biggest winners right now.
网络安全领域的巨头。
The big daddy of the cybersecurity space.
派拓网络在这一领域表现极为出色。
Palo Alto Networks is an outperformer in the space.
首席执行官尼凯什·阿罗拉。
CEO Nikesh Arora.
这可能让你感到意外，但人类写糟糕代码的历史已经非常悠久了。
This might come as news to you, but humans have been writing bad code for a very long time.
我在谷歌工作了十年，你知道，谷歌搜索让信息变得更加普及。
I spent ten years at Google, and, you know, Google search was democratizing information.
如果你把这个类比延伸一下，想想人工智能在做什么，人工智能正在让智能变得普及。
If you take that analogy and think about what AI is doing, AI is democratizing intelligence.
金钱是一种记录方式。
Money is a way to keep track.
是的。
Yeah.
这并不是目标。
It's not the goal.
您担任Palo Alto Networks的首席执行官已经八年了？
You've been the CEO of Palo Alto Networks for eight years?
再过几天就满八年了。
Coming up in eight years this week.
八年。
Eight years.
我记得您刚上任时，公司的市值大概是170亿美元。
And I think when you started, it was $17,000,000,000 market cap, if I remember correctly.
今天早上我查了一下，现在已经达到2380亿美元了。根据我们昨天聊的内容，既然您已经突破了千亿，接下来实现十倍增长的可能性就更大了。
And this morning, I checked, it's 238,000,000,000, which if you listen to what we said yesterday, now that you passed a 100, you're more likely to actually 10 x.
所以，第一个十倍增长其实要难得多。
So the first 10 x was actually much much harder.
所以，你正在朝着一万亿美元的目标前进。
So, you're on your way to a trillion dollars.
但愿如此，老天保佑。
From your mouth to God's ears.
我觉得你一定能做到。
Well, I think you are.
好的。
Okay.
那我们深入聊聊你的看法吧，因为你正好处在这样一个独特的位置，能看清这一切。
So, let's just double click into what you see because you are sort of in a really interesting position to see all of it.
你见证了人工智能的诞生，也经历了软件即服务的兴起与衰落，各种模式都跟你有关系。
You see the birth of AI, maybe you've the rise and fall of SaaS, all the models talk to you.
你曾经是其中之一。
You were one of
又崛起了。
The rise again.
对吧？
Right?
再次崛起。
The rise again.
你是最早一批获得访问 Mythos 权限的人之一。
You were one of the first and the few that got access to Mythos.
那我直接启动了。
So just let me just push the button.
开始吧，尼凯什。
Go, Nikesh, start.
首先，感谢你邀请我来这儿。
Well, first of all, thank you for having me here.
我觉得人工智能令人兴奋。
I think AI is exciting.
过去大概二十四个月里发生的一切，真的让人感到振奋。
I think it's exciting to see all the stuff that's gone down the last possibly twenty four months.
萨拉刚才说得很对，他们确实准确预判了所需算力的巨大规模。
I think Sarah just said it, they were right in anticipating the huge amount of compute that was going to be needed.
所以这些变化都在发生。
So all that stuff's going on.
但你也看得出来，我们之前简单提过一个观点，那就是AI正在让智能变得普及。
But you can see that, you know, there's this notion which we talked about briefly last time that AI is really democratizing intelligence.
这意味着，我市场部有250个人，产出的内容形式各不相同。
What that means is I have 250 people in marketing that produce varied forms of output.
现在我可以让他们90%的产出保持一致。
Now I can get 90% of the output to be consistent across those 250 people.
我还有5000名员工在直接跟客户沟通。
I have 5,000 people who talk to customers.
以前的问题是，这5000人做法不一，客户就会说：我想找Joe，因为他知道怎么解决问题，Jim却不行。
There's my my failure mode is when 5,000 people do different things where people say, I wanna talk to Joe because he knows how to solve the problem and Jim doesn't.
现在，这5000人跟客户互动时，行为几乎都能保持一致了。
So now you can get 5,000 people to act almost consistently in their interactions with people on the other side.
所以我觉得这会对企业的运营方式产生巨大的影响。
So I think it's going to have a phenomenal impact to how we run businesses, how we operate.
这会彻底改变整个格局。
It's going to change the entire landscape.
说到这个，你提到了 Mythos，我知道大卫一直深度参与其中。
Now, in that context, you've touched upon Mythos and I know David has been very involved with this.
Mythos 让我们看到，过去五十年人类写的那些糟糕代码，AI 都能评估出来，指出其中的安全漏洞。
Mythos has shown us that all the bad code that humans have written over the last fifty years can be assessed by AI and shown the vulnerabilities can be shown.
我们测试了六个星期，结果找到了原本需要五到七年才能发现的问题。
We tested for six weeks and in six weeks we found what have taken us five to seven years.
天啊。
Wow.
你再说一遍。
Say that one more time.
六个星期里，我们发现了那些 normally 需要五到七年才能发现的安全漏洞。
In six weeks we found vulnerabilities which would have normally taken us five to seven years to find.
所以这些漏洞出现在哪里呢？
And so we So Mythos These are vulnerabilities where?
什么？
Sorry?
是在你们自己的代码库里发现的吗？
These are vulnerabilities in your own code base?
我们的代码？哦，是你们自己的代码库？
In our own code Oh, in your own
代码库。
code base.
哦，天哪。
Oh, wow.
Mythos 真没吹牛。
So Mythos was not oversold.
它是实打实的。
It was legit.
AI评估代码漏洞的能力是真实存在的。
The capabilities of AI in being able to assess vulnerabilities in code are real.
不仅如此，如果你把它设为极致模式——也就是持续思考模式，它会不断尝试直到得出答案，实际上能将多个漏洞串联起来，我。
Not just that, if you put it on ultra mode, which is persistent thinking, so it keeps trying until it gets an answer, you can actually daisy chain vulnerabilities, I.
E。
E.
找到新的攻击路径来利用这些漏洞。
Finding a new attack path into your vulnerabilities.
我们公司以在代码测试方面处于顶尖水平而自豪，毕竟我们做的是网络安全业务。
Now we pride ourselves as a top percentile of companies that test our code because we're in cybersecurity business.
如果把这个能力放大到全球所有自研代码的公司，或者那上千万名开发者身上，AI能发现的东西，我们自己可能得花十年才能找出来。
If you take that and compound that across all the companies that exist in the world that write their own code or the 10,000,000 developers write code, this thing is going to find stuff which would have taken us ten years to find.
花了多少钱？
How much did it cost?
你们有统计过token的使用成本吗？
Like did you track the token cost?
是整整一亿美元，还是一千万美元？
Was it a $100,000,000, $10,000,000?
不是。
No.
不是。
No.
才几百万美元左右。
It was in the low millions.
不过，就像莎拉说的，成本曲线已经在下降了。
But again, the cost as Sarah said, the cost curve is going to come down already.
OpenAI 推出的新模型更便宜、更稳定，旧模型正被逐步取代。
OpenAI has got a model which is cheaper and more consistent, you know, and drop is come out of the other model.
你们就是被营销忽悠了。
You buy you buy the hype.
这可不是炒作。
It's not hype.
没错。
It's true.
能力确实有那方面的问题。
It's that's The capabilities are
你心里清楚。
You know that.
对。
Yes.
这些能力是真实存在的。
The capabilities are true.
对。
Yes.
你记得IBM曾宣布投入50亿美元来修复开源项目吧。
I mean, you saw IBM announced a project for $5,000,000,000 to fix open source.
这才是最大的问题。
That's the biggest problem.
如果克劳德没有那些限制，直接公开发布，会发生什么？
What would have happened if Claude didn't have the restraint and they put it out in the public?
你觉得这会不会变成真正的攻击入口，让企业陷入混乱？
Do you think it would have been like a real attack vector and caused chaos think in corporations?
再过三个月，最多就是现在，这种技术就会被公开使用了
We're three months away, if not already there, from this being available in
在外界广泛使用。
the wild.
明白了。
Okay.
开源的。
Open source.
是的。
Yeah.
就差三个月了。
Just three months.
对。
Yeah.
是的。
Yeah.
我是说，我们一直认为，距离具备神话级能力大概还有六个月。
Because I mean, we've been saying that it's roughly six months away before Mythos level capabilities are
没错。
Yes.
能用了，是的。
Available Yes.
在中文模型里，你知道的，开源模型之类的。
In Chinese models, you know, open models, whatever.
但你说是三，三个月。
But you're saying three, three months.
嗯，4.8版本已经出来了，5.5也出来了。
Well, there's what is 4.8 is already out, 5.5 is already out.
它们的功能差不多。
They have similar capabilities.
你看，你不需要破解最难的代码就能得手。
And look, you don't need to crack the hardest code to crack.
只要找到一些已经公开存在的漏洞就行了。
Just need to find a few vulnerabilities in code that are out there.
拿一个运行着工业控制系统的老设备来说吧。
Just take in take an old industrial system which is running OT code on the edge.
这种漏洞其实挺容易发现的。
You can find that vulnerability reasonably easily.
现在我们和网络攻击者之间就是在赛跑，看谁先找到这些漏洞并打上补丁。
So so we're in a race right now between the cyber defenders finding these vulnerabilities and patching them before the cyber attackers do
的
the
对。
Yes.
同样的，你觉得我们在这场竞赛中表现得怎么样？
Same And how do you feel like we're doing in that race?
说实话，我们做得还不够好，但这对我们生意来说倒是好事，不过那是另一回事了。
So not as well as we should be doing, which is great for our business, but that's a different story.
每个公司都得去检查自己的代码，找出哪里有漏洞，然后赶紧修复。
So like every company has to go look at their code base and figure out where the vulnerabilities are and fix them.
如果你现在去问首席信息官，他们最大的困扰就是一堆供应商跑来要求：请修补我提供的设备，修补我给你的代码，因为我发现了漏洞，赶紧修。
So if you talk to CIOs today, their biggest problem is all the vendors are showing up saying, please patch my piece of boxes, the hardware that you have, please patch my code that you have because I found vulnerabilities, fix it.
而首席信息官们忙着找自己系统的漏洞去修补，再加上一个巨大的难题——开源代码，没人知道该怎么彻底解决。
While the CIOs are busy finding their own vulnerabilities to fix their own vulnerabilities, and then this huge thing called open source, which nobody knows quite how to solve.
所以，可以这么说吗？随着模型能力提升，大型企业的系统性业务风险也在上升？
So is it is it fair to say that it has model capabilities go up, systemic business risk of large enterprises also goes up?
在网络安全方面，没错。
On the cyber side, yes.
像我们这样的公司正在开发一些应对方案，让你不用非得把所有东西都打补丁。
There are antidotes being built by people like us and others where we're gonna provide some capability where you don't have to patch everything.
但你看，萨拉提到一个很有意思的观点，关于记忆和上下文的保留。
But look, Sarah said something very interesting around harnesses memory and context.
对吧？
Right?
我们在这儿没说的是，企业往往缺乏对日常所有活动的完整记忆和上下文信息。
The part we don't talk about here is organizations don't have memory and context of everything they do every day.
所以你需要在企业范围内存储更多数据，才能搞清楚什么才是正常行为，什么才是异常行为。
That's why you need to store a lot more data enterprise wide to learn what good looks like and what bad looks like.
没错。
Right.
网络安全领域也是一样的问题。
The same problem is in cybersecurity.
我们必须从网络安全的角度收集十倍的数据，才能理解如何应对AI攻击者。
We need to collect we need to collect 10 times the data in the enterprise from a cyber perspective to be able to understand how to defend ourselves against the AI attackers.
你觉得像SaaS这类传统公司，在这个新环境中还有没有立足之地？
Do you think that the traditional companies like the SaaS businesses that have existed in this world, what is their place?
随着这些知识变得越来越持久并被存储起来，SaaS会怎么样？
As all this knowledge becomes more persistent and stored, what happens to SaaS?
你看，就像比尔说的，SaaS其实是由不同部分组成的。
Well, you see SaaS is as Bill said, SaaS is different pieces.
对吧？
Right?
明白了。
Okay.
如果是做分析型SaaS的公司，那就完了。
If you're an analytical SaaS company, it's over.
真的完了。
It's over.
什么叫分析型SaaS公司？
What is an analytical SaaS company?
就是那种说，我会帮你收集大量数据并帮你分析的公司。
Somebody that says, I'm gonna collect a lot of data for you and analyze it for you.
我不需要你帮我分析数据。
I don't need you to analyze it for me.
我自己就能用数据跑模型、做分析。
I can run models against data and analyze them myself.
如果你想想看，其实每个SaaS公司都像是一个市场。
So if you think about, there's a lot of every SaaS company has a marketplace.
你可以去Salesforce的市场买东西。
You can buy Salesforce marketplace.
他们是怎么说的？
What did they say?
你有Salesforce的数据。
You have Salesforce data.
我是个市场里的应用。
I'm a marketplace app.
选我吧，我能帮你分析这些数据。
Take me and I'll help you analyze the data.
我不需要你。
I don't need you.
你也不需要那个。
You don't need that.
我直接跑个大模型分析数据就行。
I can just go run an LM against the data.
所以，那些被当成独立软件模块卖给我们的一切增量功能，其实根本没必要卖给我们——我更愿意直接用一个分析工具来处理数据。
So the entire incrementality that has been sold as incremental software modules to all of us doesn't need to be sold to us because I'd much rather have a lens running against
有意思。
Interesting.
你提到这个。
You bring this up.
我们之前有个SaaS产品，买了20个席位。
We had an instance with a SaaS product with 20 seats.
没人登录使用，但数据一直都在。
Nobody was logging in and using it, but the data was there.
对。
Yes.
所以我们创建了三个账户。
So we created like three accounts.
对。
Yes.
删掉了十七个，连上了 Slack，还连上了 Claude。
Got rid of 17, connected it to Slack, connected it to Claude.
对。
Yes.
现在大家都能用自然语言来操作，我们的账单降了九成。
And now everybody can interface it through natural language and we've reduced our bill by 90%.
还不只是这样。
Well, not just that.
那你每天具体要怎么做呢，Jason？
What are you going to do day, Jason?
你是想把不同产品的数据集中到一起，然后进行分析。
Is that you're gonna take data from different products, put them in one place, run the analytics against that.
我想把销售代表的数据、工作效率数据，还有SAP里的库存数据都汇总起来。
I want my data for my sales reps, my productivity data, my, you know, inventory data from SAP.
我希望所有数据都在一个地方，这样我就能分析出谁卖得最好。
I want it all one place so I can run analytics against it and say, who's selling a lot?
哪些地方的库存比较少？
Where do I have less inventory?
我们应该在销售业绩特别好的区域多备些库存。
Let's build inventory in the region where my salespeople are extremely productive.
要查这些数据，以前得分别跟三个不同的SaaS系统打交道。
To run that query, you'd have to have talk to three different SaaS products.
明天，你就能把所有数据都集中到一个地方了。
Tomorrow, you can pull all the data in one place.
所以这算是第一类应用。
So so that's sort of category one.
分析型SaaS已经过时了。
Analytical SaaS is dead.
分析功能本身没问题。
The analytics okay.
第一类分析工具已经没用了。
Category one analytics dead.
没错。
Yes.
从中期来看。
In the medium term.
今天和明天这些波动都无关紧要了。
You get all these bounces today and tomorrow that's these are marginally irrelevant.
基础软件被严重低估了。
Infrastructure software undervalued.
好的。
K.
什么是基础设施软件？
What is infrastructure software?
就是能帮你搭建数据库的东西。
Stuff that gives you databases.
你把数据都存到里面去。
You collect data into it.
让基础设施正常运行的那些工具。
Stuff that allows infrastructure to work.
比如数据库，像Databricks、Snowflake，还有MongoDB、Oracle这类东西。
Whether it's a, you know, database Databricks, Snowflake like Databricks, Snowflake, MongoDB Oracle.
Oracle。
Oracle.
这些都是你必不可少的。
All these things you need.
核心的存储基础设施，核心的数据。
Core storage infrastructure, core data.
我们需要的 enterprise 数据存储量将是现在的十倍。
We are gonna need 10 times the data stored enterprise than we have today.
没错。
Right.
未来三年内。
The next three years.
十倍。
10 times.
明白了。
Okay.
所以，任何能帮你收集和管理基础设施数据的工具，都是必需的。
So anything that helps you collect infrastructure data, manage it, you need.
我觉得中间这个类别，我们可以称之为‘工作系统’，或者有人管它叫‘记录系统’。
I think the category in the middle is called let's call it system of work or system, you know, a record people call them.
这些系统已经深深融入了企业的运作方式中。
Those are deeply embedded in the way businesses work.
我有六千名销售人员。
I have 6,000 salespeople.
他们很清楚这套流程是怎么运作的。
They know how this works.
接下来第一步，我们会去掉界面，让智能代理来完成这些工作。
What's gonna happen is step one, we will take away UI and let agents do the work.
企业软件和消费软件的界面，是我们技术人做过最糟糕的设计。
UI enterprise software and consumer software UI is the worst thing we did as technologists.
你之前提过几个这样的例子。
You had a couple of examples of this.
你跟我讲过这个故事。
You told me this story.
你愿意再讲一遍吗？就是关于那家公司的。
I don't know if you want to repeat it of this one company.
他们曾试图用许可证来要挟你。
They tried to hold you hostage on a license.
是的。
Yes.
那是分析型的SaaS。
That was analytical SaaS.
所以呢，
So that's so
你直接把AI用在上面，然后就解决了。
And you just pointed AI at it and you just Yes.
我们直接把它们取消了。
We just got rid of them.
这是另一个问题。
That's a different issue.
但你想想现在的情况。
But I mean, think about it today.
我们整天让产品经理设计界面，好让所有人通过界面去操作背后的数据。
We spend our lives having product managers design UI so all humans can interact with data behind the UI.
是的。
Yeah.
如果你相信智能代理能发挥作用，那我就直接告诉一个代理：从我的销售通话中提取关键信息，然后自动发到我用的Oracle或者Salesforce这类销售跟踪系统里。
If all if you believe agents are gonna work and I say, I just tell an agent, look, figure out from my sales call, figure out the key points and go post it into, you know, whatever sales tracking system I have with this Oracle or Salesforce.
对吧？
Right?
从概念上讲，智能代理本来就应该能做到这一点。
An agent conceptually should be able to it.
那我们为什么还要花上万亿美元去搭建这些智能代理的后台系统呢？
Should we're spending trillion dollars building these agentic back ends?
我们确实需要这些代理能完成这些任务。
We need these agents to be able to do it.
如果真能做到，用户界面就彻底没用了。
If that happens, UI goes away.
一旦界面消失了，我就能重新设计整个工作流程。
If UI goes away, I can rewire my system of work.
对吧？
Right?
想象一下，销售跑完客户电话，剩下的填表、走流程这些后台杂事全交给代理搞定，他就能直接收工。
I have sales guy shows up and says, I had the sales call, do all the paperwork and all the shit that needs to happen in the back end of the company and just I'm done.
没错。
Right.
一旦工作模式变了，那才是真效率，公司里五个人能干以前一个人的活，所有做工作流的SaaS软件都得重新设计。
If I can change the way work happens, which is where you will get true efficiency, where five people become one in a company, all these SaaS software that does system of work needs to be reengineered for
未来五年
the next five
里。
years.
而且这也是在被动发生，这真的很有意思。
And it's also happening passively, which is really interesting.
比如看看邮件。
It's looking at email.
它会自动抓取Zoom的会议记录和摘要。
It's automatically taking the Zoom transcript and summary.
所以销售系统的记录现在根本不需要你手动输入。
So the sales system of record is now like, you don't even need to input it.
我直接就有Zoom通话的笔记了。
It's like, I already have the Zoom call notes.
我手头还有演示文稿。
I have the deck.
演示文稿已经生成好了。
The deck was made.
销售用的演示文稿是AI做的。
The sales deck was made by AI.
以后我们都会盯着聊天窗口，直接说一句：我要这个。
It's just we're all gonna be looking at a chat window and just saying, here's what I want.
因为没人手动操作数据，你的审计追踪反而更清晰了。
Your audit trail becomes a lot better because humans are not touching your data.
整个流程始终由智能代理管理。
It's always being managed by agents.
我认为，未来五年内，整个工作体系和数据记录方式都将被彻底重塑。
So I think the whole system of work, system of record gets reinvented in the next five years.
是的。
Yeah.
根本不需要手动录入数据。
There's no data entry.
这个观点挺有意思。
That's an interesting point.
对。
Yeah.
我们稍微谈一下国家安全吧。
Let's talk about national security for a second.
我想把视角拉得更广一点。
I just want to maybe zoom out.
正如你所说，Mythos 的价值体现在你和企业身上。
So the one side of Mythos as you said is like the value that it has to you and to your and to enterprises.
Mythos 的对立面是，外国国家行为体可以借此在一国内部制造经济混乱。
The red team version of Mythos is where foreign state actors, you know, can essentially create economic havoc inside of a country.
是的。
Yes.
随着这些模型能力不断提升，你觉得当它们成熟时，应该采取什么措施？
As these models escalate in their capability, what do you think should happen when these models are ready?
说实话，每年都有几千起安全事件或攻击发生。
You know, the sad truth is in a year there's a few thousand breaches or attacks that happen.
这些事件大多源于非常基础的原因。
They happen for pretty rudimentary reasons.
并不是因为有人破解了什么高难度的系统。
It's not because somebody cracked the hard to crack thing.
高达 89% 的攻击，都是因为凭证被盗。
It happens because 89% the attacks happen because credentials get stolen.
用户名和密码。
Username and password.
就这。
That's it.
密码还是‘Password’。
Password is Password.
是的。
Yeah.
我猜也是。
I'm sure it is.
你用的是DollarSign吗？
Do you have DollarSign?
DollarSignPassword。
DollarSignPassword.
太棒了。
Fantastic.
干得漂亮。
Well done.
你看，你已经领先所有人了。
See, you're already ahead of everybody else.
百分之八十九的入侵事件都是因为一些简单的原因发生的。
So eighty nine percent of breaches happen because of simple things.
我觉得我们不需要更多模型去破解这些内容。
I don't think we need more models to go crack the stuff.
现在我们需要的是，这些模型能攻击关键基础设施，也就是我们从国家安全角度努力保护的东西。
Now we will need these models can attack critical infrastructure and things we try and protect from a national security perspective.
对。
Yes.
那里我们需要加强防御。
We need defenses there.
我不担心国家安全方面的防护，因为他们已经非常重视了。
I'm not worried about the national security part being protected because they're very on it.
他们才是对的人。
They are the right people.
他们把预算的10%都花在了信息安全上。
They spend 10% of their budgets on IT on security.
我担心的是全国各地那些小诊所，他们用的都是些简单的软件，比如牙科诊所或者内科诊所。
I'm worried about the small offices across the country where they're using some piece of packet software and you're running a dentist office or doctor's office.
还记得当初医疗系统被攻破的时候吗？所有医生诊所都停摆了。
Remember when change healthcare got breached, every physician's office shut down.
直接瘫痪，而且是勒索软件攻击。
Shut down and it's ransomware.
因为那次针对医疗系统的勒索软件攻击。
Because the ransomware to change healthcare.
那就是结算系统出的事。
That's what the clearing system.
那时候联合健康公司不得不给医生们提供数十亿美元的补贴，才能让他们继续运营。
That's when United Health Act actually have give billions of dollars of credits to the physicians to be able to run their businesses at that point in time.
这才是真正该担心的事。
That's what one should worry about.
重点不在于那些大目标会被攻破。
It's less about the big nuts will get cracked.
重点不是去攻击像PG&E这样的电力设施。
It's less about cracking some PG and E power generation facility.
而是会引发经济混乱。
It's more economic chaos.
对。
Yes.
那我们该怎么办？
And so what what what do we do?
我觉得没有一劳永逸的办法。
I don't think there's a sort of a silver bullet.
这需要时间。
I think this will take time.
我觉得这需要一段时间，等所有系统逐步升级、更新和修复。
I think this will basically take a while until every system gets upgraded, renewed, fixed over time.
我只是觉得它提升了整个行业的长期价值。
I just think it increased the terminal value of the industry.
对吧？
Right?
没错。
Right.
你认为会不会有这样的世界：这些模型变得太强大了，以至于你会支持国家加强对它们的控制、管理和使用方向的监管？
Do you think that there's a world in which these models become so good that you could see yourself advocating for more nationalism around how they're controlled and how they're managed and how they're where we point them?
还是说，应该有一些模型永远不能公开，只有国安局这类机构，或者像你这样的人才能接触？
Or do you think there should be a maybe a set of these models that never see the light of day that only the NSA and other folks could have access to or guys like you?
我对模型的发展方向，和之前OpenAI那边说的，看法有点不一样。
I have a slightly differentiated view about models and how they will evolve versus what we heard earlier from an OpenAI perspective.
我仍然相信，模型最终会变成一种基础服务。
I think I still believe models are going to become a utility layer.
你可以随时购买智能服务，而不是非得找个智商180的人来干这事。
You'll be able to buy intelligence on the fly or you can say, I don't need a 180 IQ person to go do this task.
给我一个智商120的，再配两个智商50的，就能搞定这个任务。
Give me a 120 IQ and I need two fifty IQ to do this task.
这事我愿意花10美元，那件事我只愿意花1美分。
I'll pay $10 for this and for this I'll pay 1¢.
我不觉得能用同一个最新模型来应对所有客户来电，比如跟客户说：‘抱歉，这个问题我也不太清楚怎么解决。’
So I don't know there's a one size fits all, give you the most up to date model to answer my customer calls saying, sorry, sir, I have no idea how to solve your problem.
所以我认为，从实用角度来说，模型之间会逐渐分化。
So I think models will get differentiated from a utilitarian perspective.
你看看现在市场上已经发生的变化，对吧？
So if you look at already what's happening in the market, right?
真正的利润点在应用层，而不是模型本身。
The profit pools are in applications, not in models.
莫尔和萨拉提到过Codex已经跑在前面了。
More and Sarah talked about Codex running away.
她可没说OpenAI在领跑。
She didn't say OpenAI is running away.
她说的是Codex在领跑。
Says codex is running away.
就像我肯定Dario会说，Claude的代码能力在领跑一样。
Say just the way I'm sure Dario says Claude code is running away.
所以你看到他们在抢占利润池。
So you're seeing that They're attacking profit pools.
他们之所以抢占利润池，是因为钱都往那儿去了。
They're attacking profit pools because that's where the money is going to come from.
利润池在于企业能直接使用的应用产品。
The profit pools are an application that companies can use.
利润池不在企业使用模型本身，因为大多数公司根本不知道怎么用模型。
The profit pools are not in model usage by companies because most companies have no idea how to use models.
从某个角度看，OpenAI和Anthropic就像新一代的微软Office，正在为各类企业打造所有办公和生产力软件。
Look at these companies in a way, OpenAI Anthropic, as the new Microsoft Office coming in and doing all applications, all productivity software for organizations.
不是。
No.
我觉得会出现一些应用公司，它们会在不同模型之间套利，帮你解决实际的业务问题。
I see there's going to be application companies which are going to arbitrage between models and solve your business problem.
所以你还是觉得它们不会进入应用层？这可是个大争议点。
So you still think they won't go to the application layer because this is a big debate.
你真的应该去和OpenAI合作，训练它们的系统，然后让它们来抢走你的生意吗？
Should you engage with OpenAI and train their systems to then take your business from you?
Anthropic一直在发布他们的法律模型、会计模型。
And Anthropic keeps releasing their legal model, their accounting model.
而且说实话，为了达成营收目标，它们可能真得像微软那样，把Office这样的产品直接搭在操作系统上面。
And it does feel like in order for them to hit their revenue numbers, they might need to do what Microsoft did, which is release the office product on top of the operating system.
如果我是家公司，我可不想自己从头写每一套软件。
See, if I'm a company, I don't want to write every piece of software myself.
我希望我的HR系统，能由某个应用公司提供，自带智能代理和AI功能。
I want my HR system software, which is agentic enabled and AI enabled to be delivered by some application company.
可能是家新的AI应用公司。
Could be a new AI application company.
我希望我的销售管理系统，是由下一代智能AI版的Salesforce这样的公司来打造的，不管是不是Salesforce本身。
I want my sales management system built by the new agentic AI Salesforce of the world, whether it's Salesforce or somebody else.
所以我想要的是现成的应用程序。
So I want applications.
萨拉说的是，利润其实都在应用层。
Now, what Sarah said is the profit pools are in the application layer.
这就是为什么他们想抢占应用层的位置。
That's why they want to be the application layer.
我觉得我们还在等这一批公司出现，它们会专门做应用层，因为成千上万家企业都需要同样的工具。
So I think we're still waiting for that layer of companies to be invented or created where applications will sit because 50,000 companies need the same application.
我干嘛要自己从头开发呢？
Why would I build it myself?
这效率太低了。
It's highly inefficient.
我明明不是那种非得自己动手重写整个销售系统的人，用不着直接去折腾OpenAI。
It's silly for me to use OpenAI directly and rewrite my entire sales system because I'm smart.
对吧？
Right?
我可没那么厉害。
I'm not.
我就是希望有人能替我搞定。
I want somebody to do it for me.
所以我觉得，专门做这一层的公司还远没成型。
So I think that layer of companies are still not fully formed.
我们还得再等等。
We're still going be waiting
你需要一个控制平台，一个框架，然后
for You want a control plane, a harness and then
没错。
That's right.
展开剩余字幕（还有 321 条）
他们会把控制平台和内存集成到这些应用层里。
They will build the harnesses and the memory into those application layers.
现在的问题是，这个应用层到底有多大？
Now the question is how big is the application layer?
是一个单独的应用吗？
Is it one application?
是一个能包办一切的大型企业软件，还是说它其实是专门化的？
Is it is one, you know, enterprise application that does everything or is it specialized You
你亲自做了，然后把那个软件供应商给换了。
did it and you kicked out this software vendor.
你这么做是因为他们的定价太霸道了。
You did it because they were being abusive in pricing.
我们还是在用另一家供应商。
We still use this different vendor.
那家是哪家？
What's that?
我们换了一家不同的供应商。
We swapped out for a different vendor.
我们只是拿回了更多控制权。
We just took more control.
太棒了。
Love it.
所以这本质上还是价格问题。
So it really is a pricing issue.
这也解释了为什么在某种程度上，SaaS的危机是有道理的。
And and that's why the SaaS apocalypse in some ways makes sense.
他们不再有定价权了，因为你完全可以说：我招十个开发人员来做，就能省下一千万。
They're not having pricing power because you could say, I'll just put 10 developers on this and I'll save $10,000,000.
对。
Yes.
我觉得这又回到了查马斯提到的那一点，关于是否应该对这些高性能模型进行监管。
I think the part back to what Chamath said about the regulation or whether you want to regulate these higher powered models.
问题是，当这些更强大的新型模型问世时，它们的价格会不一样，可能还得经过某种审核流程，才能明确它们到底有多强。
The question is at some point in time when these newer models, which are even more powerful get built, they will come at a different price point and they might have to go to a certain vetting process to understand what their capabilities are.
但我认为我们现在正处在一场全球竞赛中。
But I think we're in a global race.
我觉得把我们的模型推迟三到六个月，根本帮不上忙。
I don't think holding back our models for three to six months is gonna help us any.
别人迟早会把它们开源出来。
Somebody else is gonna put them out in open source.
我听说一家模型公司的CEO说，他们最新模型的所有权重数据居然能装进一根U盘里，当时我都惊了。
I I was I was shocked to hear when was talking to CEO of one of these model companies says, the entire weights of their most recent model can fit on USB stick.
你再说一遍？
Oh, say that again.
所有权重数据？
The entire weights
他们最新模型的全部权重数据，真的能装进一根U盘里。
Entire model weights of their newest model fits on a USB stick.
这就是他们的核心技术。
That's the IP.
这太惊人了。
That's incredible.
因为所有数据都能在二十四到四十八小时内完成提炼，模型就出来了。
Because all the data can be distilled in under twenty four to forty eight hours and model comes out.
我很好奇。
I'm curious.
这确实是他们的核心技术。
So that's the IP.
所以你的意思是，我们要把这东西锁六个月？
So are you telling me that, you know, we're gonna hold on to that for six months?
没错。
Right.
我们一直在讨论，要做出一个前沿模型到底有多难。
We we have a debate about how difficult it is to make a frontier model.
一些公司已经开始考虑利用自己的数据优势来开发前沿模型。
Some companies are starting to think about making frontier models using their data advantage to to build their own.
帕洛阿尔托那边你们有考虑过这个吗？
Have you thought about that at Palo Alto?
因为你们确实掌握着关于安全运作的专有知识。
Because it does seem like you have proprietary knowledge on how security works.
你们能不能自己开发一个大型语言模型，或者一个小型语言模型，来获得某种优势？
Could you build your own large language model or a VSML, a small language model that would give you some advantage?
这里有个关键点，
Here's the part Yeah.
但没人提过。
Nobody talks about.
对。
Yeah.
就是模型的误报率问题。
Is the false positive rates on the models.
四点八和五点五的误报率是多少？
What is the false positive rate on four point eight and five point five?
不清楚。
No idea.
你们从来不提这个，其实应该谈谈。
You guys don't talk about it, you should.
Mythos的误报率高达百分之三十。
The false positive rate on Mythos was 30%.
哦，真的吗？
Oh, wow.
对吧？
Right?
所以它以为发现了什么，但实际上并没有？
So it thought it found something but it hadn't?
对。
Yes.
所以问题在于，这玩意儿用于进攻挺厉害，但用于防御就太糟了。
So the problem is, it's great for attack, it's horrible for defense.
它有30%的概率误报，每次说发现了问题，你就赶紧去补漏洞。
It finds 30 times 30% of the time it finds something, I found a problem and you say, let's plug the hole.
等等。
Wait.
可根本就没什么漏洞。
There wasn't a hole there in the first place.
根本没导弹来袭。
No missile inbound.
对吧。
Right.
是的。
Yeah.
现在同样的问题也出现在企业环境中。
So now the same problem applies in enterprise.
如果你用的模型没有正确的约束和训练，误报率可能会高达百分之二十。
If you use a model without the right harnesses, the right training, you could be running into 20% false positive rates.
咱们拿这个模型去处理保险理赔吧，我也不知道行不行。
Let's use the model to pay, I don't know, insurance claims.
是的。
Yeah.
太好了。
Oh, great.
误报率百分之十、二十。
10%, 20% false positive.
我直接亏钱了。
I just lost money.
这些模型的奉承特性简直荒唐。
The sycophantic nature of these is ridiculous.
是的。
Yeah.
所以问题不在于谁想要最新的模型。
So so the problem is not who wants the newest model.
问题在于，你怎么把那个有10%到20%误报率的模型，优化到误报率只有百分之零点零一？
The problem is how do you take that model with 20% or 10% false positive and make it point 01% false positive?
在我的行业里，我要求的是零误报。
In my business, I want 0%.
但不能让漏报也跟着增加。
Without losing the false negative.
抱歉，你说什么？
Sorry?
不能让假阴性也上升。
Without losing the negative, the false negative.
这就像是说，咱们用上最新款的自动驾驶汽车吧。
But it's like saying, hey, let's take the new self driving car.
奔驰要用Opus 4.8，你坐进去，车就能自己开。
Mercedes is going to use Opus 4.8 and you can just sit in the car and it's going to drive you.
我不会让我的孩子坐进一辆误报率高达10%的车里。
I'm not putting my kids in that car with a 10% false positive rate.
你也不会吧？
Are you?
没错。
Right.
在模型之后，还有很多工作要做，才能让这套系统在实际业务中真正派上用场。
So there's a lot of work that happens post the model, which needs to happen to make this thing useful and effective in the business context.
我稍微换个话题。
Let me slightly pivot for a second.
你曾经在谷歌担任首席业务官很长时间。
You were for a very long time the chief business officer at Google.
你还当过软银的总裁。
You were the president of SoftBank.
现在你是派拓网络的首席执行官。
Now you're the CEO of Palo Alto Networks.
那咱们来当一回纸上CEO吧。
So let's play armchair CEO.
纸上CEO。
Armchair CEO.
纸上CEO。
Armchair CEO.
我还在为大卫·弗里德伯格硬要区分创始CEO和非创始CEO的事感到不爽。
I'm still I'm still bristling from David Friedberg trying to create a distinction between founder CEOs and non founder CEOs.
就是说一下。
Just saying.
就是提醒你一下，大卫。
Just saying, David.
顺便说一句，假阳性的问题。
By the way, false positives.
什么？
Sorry?
假阴性也是。
False negatives too.
说说你认为哪些地方该保留，哪些该改，还有哪些地方你喜欢这些公司的地方。
Give us what you would keep, what you would change, and what you like about the following companies.
这会被录下来发布出去，让大家听到
This is gonna get recorded and put out there to say
你的看法。
your thoughts.
你是最懂商业的人之一。
You're one of the smartest business people.
我并不是
I don't
喜欢这样去看待
like that to look at
全盘接受，别
the all in don't
你准备好了吗？
Are you ready?
你准备好了？
Are ready?
当然了。
Yeah.
没问题。
Sure.
好的。
Okay.
说说你对优步的看法：哪些地方好，哪些需要改，喜欢什么，不喜欢什么。
What you keep, what you change, what you like, what you don't like, Uber.
在这个世界里，我
In a world of I'm
都腻了，老兄。
bored of it, dude.
我不能谈我是不是
I can't talk about my Are you
优步的董事会成员？
on the board of Uber?
我就是优步董事会的成员。
I'm the board of Uber.
我不是
I'm not
打算谈这个事
gonna talk about
我还不知道呢。
didn't know that.
不好意思。
Sorry.
你去问达拉吧。
Talk to Dara.
他是首席执行官。
He's the CEO.
他这个人很不错。
He's a great guy.
好的。
Okay.
Waymo。
Waymo.
设法去争取
Try to get
把我给炒了。
me fired.
Waymo。
Waymo.
我到底喜欢Waymo什么？
What do I like about Waymo?
这些车都能正常运行。
The cars work.
太惊人了。
It's amazing.
它们应该更快地在更多城市推广开来。
They should have more in many more cities around the world faster.
我就会跟塔希拉这么说。
I just I would say that to Tahira.
我觉得她谷歌
I think she Google
把这事儿放大来看。
writ large.
我觉得谷歌被低估了。
I think Google's underrated.
我觉得它会成为我们这一代第一家市值达到一万亿美元的公司。
I think it's gonna be the first $10,000,000,000,000 company in our lifetime.
我觉得他们已经具备了让这件事成功的全部资源。
I think they have all the assets that are are needed to make this successful.
人们往往低估了这一点。
I think people underestimate.
就算你是个优秀的公司，也得有个能说服客户接受并购买这些模型的销售团队。
You can be a model company, you still need to have a Salesforce that convinces customers to go out there and embrace these models and buy them.
你想想看，三大云服务商拥有的销售人员数量是最多的。
And if you think about it, three hyperscalers have the biggest number of salespeople out there.
所以他们理应
So they should be
他们有点被低估，部分原因就在于集团结构太复杂，不容易看明白。
Part of why they're a little bit undervalued is just the conglomerate nature is hard to understand.
我也不太确定。
I don't know.
你们对这些事儿可真懂。
You guys are smart of that stuff.
我只是个被雇来的CEO。
I'm just a hired hand CEO.
我没这么说。
I didn't say that.
是弗里德伯格说的。
Friedberg said that.
咱们说清楚点。
Let's just be clear.
我知道。
I know.
我知道。
I know.
我当时是在分析SaaS行业复苏的逻辑。
I was I was providing a thesis on recovery out of the SaaS pocalypse.
明白吗？
Okay?
好的。
Okay.
明白了。
Got it.
我再明确一下。
Just to be clear.
很高兴能合作。
Nice to work together.
这里有个
There's a
这种分类是可以细分的。
there's a way to segment that basket.
你不在这个类别里。
Got And you're not in that basket.
我觉得你之前是在区分，创始人和CEO有权承担更多风险，也允许承担更多风险。
I think you were making a distinction about how people who are founders, CEOs have have the right to take more risk and are allowed to take more risk.
我刚才就是这个意思。
I was saying that.
我觉得你恰恰提出了一个独特的不同观点。
And I think and I and I think you you provide a unique counterpoint to that.
而且这样的人其实不多。
And and there's not a lot
你的模型里几乎没有误判。
false positive in your model.
我觉得杰夫·韦纳也是这样。
I think the same was true of Jeff Weiner.
我觉得还有几位特别出色的CEO，但他们就像《黑客帝国》里的尼奥，属于罕见的特例。
And I think that there's a few other really great CEOs, but they are like neo in the matrix type anomalies.
我觉得你就是这种人，有一种非常罕见的性格特质：愿意承担风险，主动承担起原本不属于自己的责任，并把它变成自己的东西。
And I think you're one of those people, and there's a very rare kind of personality profile of someone that's willing to take risk and take ownership of something that wasn't theirs in the first place, and they make it theirs.
这是一种极其独特的品质，实际上比成为可扩展的创始人还要罕见。
And it's a it's a extraordinarily unique trait, far more unique actually than being a scalable founder.
这真是个了不起的扑救。
It's an incredible save.
你被原谅了。
You're forgiven.
是的。
Yeah.
扑救得真好。
Good save.
这个扑救很有说服力。
Credible save.
哇。
Wow.
这实在太惊人了。
That was incredible.
但他其实更不舒服，他现在的情况是这样的
But he's more sick He's like, actually
我其实
I'm actually
我才是最棒的。
getting I'm I'm actually best.
是的。
Yeah.
咱们回到《沙发CEO》这个话题吧。
Let's go let's go back to Armchair CEO.
我挺喜欢这个的。
I'm liking this.
他拥有OpenAI。
He has OpenAI.
更频繁了。
More often.
人工智能。
AI.
他们的销售速度更快。
They do sell faster.
OpenAI。
OpenAI.
他们本该卖得更快。
They should sell faster.
对吧？
Right?
他们本该卖得更快。
They should sell faster.
我的意思是，我说过了。
I mean, I mean, said it.
你刚才不是在Sarah在这儿的时候说过吗？Anthropic的年经常性收入增长速度比OpenAI快多了？
Didn't you just say it when Sarah was here that Anthropic seems to have improved their ARR much faster than OpenAI?
我的意思是，
I mean,
这只是一个统计数据。
that's just the statistic.
他们完全押注在企业市场上了。
They kind of went all in on enterprise
而且我觉得特别集中在编程领域。
and I think coding specifically.
我觉得现在大家讨论的焦点就是谁能抢下利润最高的那块市场。
I think I think that's like the the conversation right now is it's a race to take over the profit pools.
你每年得投入几十亿甚至上百亿美元，才能达到那一个吉瓦的水平——十亿美金，那建一个这样的系统到底要花多少钱？
You are going to need tens and tens of billions of dollars every year to get where that one gigawatt is 10,000,000,000 What are what are the most What does it cost to build you?
最令人兴奋的
Are the most exciting
成本是五十亿。
costs 50.
所以，这正是关键所在。
So, this is a Exactly.
好的，抱歉。
Great Sorry.
那接下来呢？
So, what
那么最值得关注的利润领域有哪些？
are the most exciting profit pools then?
我们先说编程。
So, we got coding.
编程已经成了最突出的应用，规模早就超乎想象了。
That's been the breakout application over the It's past massive.
就像你提到的，还有基础设施，比如新型数据库。
You've got infrastructure like you said, the new databases.
我认为网络安全肯定算一个，因为威胁不断变化，补丁更新也快得多。
I think cybersecurity is clearly one of them because of the threats and patching cycle is so much more dynamic.
确实有一点不同，你看这些模型正在努力成为提升网络安全的助力，这很好，毕竟我们所有人都得靠它们来测试。
There's a slight difference in yes, as you can see, these models are trying to be the enablers of better cybersecurity, which is good because all of us need to use them to test.
你可能会看到，Anthropic 已经把他们的网络安全模型全面开放了，任何人都能用。
And you're probably going to see, I mean, if you saw Anthropic has already made their cyber capable model available generally so that everyone can use it.
OpenAI 也有类似的产品，我猜谷歌肯定也有，他们很清楚，首席安全官们正想用这些工具来测试代码。
And OpenAI has got one, I'm sure Google has one too, but they understand this is a place where CSOs or chief security officers want to use it to test the code.
这又是一个不错的盈利点。
So it's a good another profit pool.
我觉得，针对应用软件公司的冲击还没真正到来。
I think we haven't seen the onslaught against the application software companies yet.
应用软件市场有成千上百亿美元的规模，正等着被重新定义，就像我们之前说的。
I mean, there's tens and tens of billions of dollars in application software, which is waiting to get reinvented as we talked about.
最终你会看到这些人开始想：如果我把这四千亿、一万亿美元的市场总量给颠覆了呢？
I think eventually you'll see these people saying, what if I took this $40.50, $100,000,000,000 TAM down?
对。
Yeah.
我可以打造一套全新的底层架构，完全基于 AI，做出极致的差异化，让客户主动切换过来。
I can build a whole brand new backbone rejecting AI and it'd be so differentiated that it'll cause customers to move.
现在孵化器里，这已经成了一套标准做法。
We are seeing it as a playbook in the accelerators now.
初创公司，尤其是刚成立的第一年，都主动来找我们推销这个点子。
The year zero and year one companies, people are coming to us with the pitch.
这是一款每人每年一千美元、每月五百美元的SaaS软件。
This is a thousand dollar a seat per year, $500 a month seat SaaS software.
我们能做得更便宜。
We can do it for less.
我们会根据实际使用量来收费。
We're going to charge them based on consumption.
我们能把成本砍掉八到九成，比你打算做的还要狠。
We're going to take 8090% of the cost out as to what you're about to do with eighty, ninety.
最快能赚到钱的两个方向。
Two fastest places to make revenue.
没错。
Yeah.
在企业市场，替换型客户成功经理最有效。
In enterprise are replacement TAMs.
如果你是在替换现有产品，客户已经有预算了，这事儿就简单多了。
If you replace something, I already have a budget, it's easy.
把差的东西换成更好的，不仅能拿回钱，还能赚更多。
I take something bad or replace to something better, I get money.
所以，替换型客户成功经理简直太理想了。
So replacement TAMs are beautiful.
要是能取代整个行业，重新分配利润池，那就更棒了。
If you can replace an industry, replace the profit pool is great.
第二个赚钱快的地方是消费者市场。
The second place is consumer revenue.
在消费者端，每用户收五美元要容易得多。
It's a lot easier to get $5 per user on a consumer side.
就像Netflix那样。
Netflix.
所以我说的就是这个意思，你看看就知道了。
So that's where I mean, look at it.
我觉得我们每个人每月花在订阅服务上的钱，可能比过去任何时候都多。
I think we collectively probably pay more on subscriptions per month than we ever did historically.
是的。
Yeah.
你以前还觉得有线电视账单够贵了呢。
And you thought your cable bill was high.
对。
Yeah.
如果你要猜的话，你觉得未来你会建造更多还是更少的硬件？
Do you think that you're gonna end up building more or less hardware in the future if you had to guess?
即使在今天，硬件仍然是处理低延迟、高吞吐量数据最经济的方式。
Hardware even today is the cheapest way to manage low latency, high throughput bits.
你还是需要数据中心。
You still need a data center.
是的。
Yeah.
数据中心是什么？
What's a data center?
其实就是处理高吞吐、低延迟的数据流。
It's just managing high throughput, low latency bits.
对。
Yeah.
所以你看，金融行业最不愿意上云，就是因为延迟会增加。
That's why if you look financial services is the most reluctant industry to go to the cloud because you increase latency.
延迟一高，利润就下降。
If you increase latency, you reduce profit.
所以你看那些最大的金融机构，比如高盛、摩根大通、摩根士丹利……
So if you look at every of your largest financial services company that is Goldman or JPMorgan, Morgan Stanley
在第八街。
Eight Street.
或者这些公司，是的。
Or these guys Yeah.
他们在做硬件。
They're doing hardware.
对。
Yeah.
让他们把业务迁到云上试试看。
Try to get them to run their business in the cloud.
他们做不到，因为延迟会增加。
They can't because they will have higher latency.
那样就会亏钱。
They will lose money.
没错。
Right.
所以硬件还是得继续生产。
So hardware is still be made.
我是说，我记得以前买Silver Lake的时候，听说戴尔都快不行了。
I mean, I remember when I used to buy Silver Lake and I would heard Dell was done.
那时候没人想碰硬件。
Nobody wanted hardware.
我觉得戴尔的市值可能又会回到三千四百亿美元左右。
I think Dell's might speed back to like a $3,400,000,000,000 market cap.
所以硬件还是会一直存在。
So hardware is still going to be around.
我们依然需要它。
We're to need it.
它是最快的数据处理方式。
It's the fastest way to move.
人工智能会不会改变硬件的开发周期？
Are are hardware development cycles changing because of AI?
比如，你有没有看到生成式设计正在取代过去那些手动、耗时很长的芯片设计流程？
Like, are you seeing a lot of like generative design stuff moving in silica that historically was manual and long cycle?
是的
Yeah.
但真正拖后腿的从来都不是设计环节。
But the long pole in the tent is never designed.
对吧？
Right?
是的
Yeah.
真正拖后腿的是生产环节。
Long pole in the tent is production.
现在你根本没法把机器造出来，因为每个硬件部件都缺货。
Today, you can't get a box produced because every every piece of hardware componentry is back ordered.
没错。
Right.
什么东西都贵，全世界的工厂都在排队，因为我们正忙着为每个数据中心生产这些基于GPU的芯片卡。
Everything is expensive and every factory in the world is back ordered because we're trying to build all these GPUs based, you know, chip cards for every data center of the world.
你觉得美国有能力填补这个供应链缺口吗？
You think The US is equipped to fill that supply chain need?
我们能在本国做到吗？
Can we do that here?
想象一下十年后。
Do think ten years.
十年。
Years.
只要有坚定的、自上而下的全力投入。
With a commit with a with a firm top down commitment.
好消息是，我觉得硬件行业正迎来前所未有的大好时机。
Well, I mean, the good news is that I think the hardware industries is seeing a bonanza of a lifetime.
一般来说，遇到这种千载难逢的机会，企业敢砸下几千亿、上万亿美元。
And generally, when you see a bonanza of a lifetime, you can go commit $10.20, $50,100,000,000,000.
我见过一位CEO在电视上宣布，要投入一千亿美元来扩建内存产能。
And I've seen a CEO on television commit to go a $100,000,000,000 plan to go build more memory.
这倒是好事。
So that's good.
这意味着他们有资金真正投入建设，为未来打造这些设施。
That means they have the money to go put the money in the ground literally to go build these things for the future.
所以我觉得这让我们更有把握了，
So I think that gets us more certain that the
我觉得税收优惠在这方面起了很大作用。
I think the tax incentive has a big has a lot to do with that.
资本支出可以加速折旧。
The accelerated depreciation on the the CapEx.
第一年就能全额抵扣。
You get a 100% write off in the first year.
对吧？
Right?
根据
Under the
最后一个问题，我们快收尾了。
under the Just a final question as we wrap up.
过去八年，你们既有机增长得非常迅猛，也做了不少收购。
You, over the last eight years, you've grown organically very aggressively, but you've also been pretty acquisitive.
你们会出手尝试，而且大多数都成功了。
You'll, you know, you'll take shots and they've generally worked.
你们在市场上已经积累了很大的信任和空间。
So, you have a ton of permission in the market.
当你们听到比尔·哈克曼说，市场上有些公司被过度打压，只有少数被捧上神坛，这些其实正是你们可以挑选的目标——但有些收购可能需要你们往网络安全核心领域之外走得更远，有人会这么认为。
When you hear what Bill Hackman said about how there's this kind of over beaten companies, there's a few that get celebrated, that's a right pool for you to pick from, but some of that would require you to go maybe a little horizontally far afield, some would say.
你们怎么保持纪律？或者，你们会不会在某个时候考虑那些并不完全在网络安全核心范围内的机会？
How do you maintain the discipline or do you see yourself at some point considering things that are not nearly so much right down the middle of of cyber?
我跟你说，大概一年半以前，我们还专门收购产品型公司，然后把它们塞进我们的市场推广体系里。
So I tell you what, until about an year and a half ago, we used to buy product companies and throw them into our go to market engine.
我们可以重做它们的后台，让它们更好地融入我们的市场推广系统。
We could rewire their back end so they can work better with our go to market engine.
对我来说，如果我向客户卖了1000万美元，下次再去的时候，如果能再卖2000万，那就是我摊销市场投入成本最高效的方式。
So for me, if I'm selling $10,000,000 to a customer, next time I go to it later, if I can sell them 20, it's the most efficient way for me to amortize my go to market spend.
对吧？
Right?
所以这就是我们当时的模式。
So that was the model.
我们就是这么做的。
We played that.
这套打法我们用到了一千五百亿美元以上。
We ran that playbook to north of a 150,000,000,000.
后来我们发现，身份认证领域正迎来一个转折点。
Then we go to a point where it says, oh, we see an inflection arriving in identity.
无论从智能代理还是安全的角度来看，这都变得至关重要。
It's going to be important from an agentic perspective, security perspective.
于是我们收购了一家估值250亿美元的公司，三个月前刚完成交易。
So you bought a $25,000,000,000 company, which we closed three months ago.
现在，出现了一个完全不同的机会。
Now, it's actually a very different opportunity has presented itself.
这个新机会大致是这样的。
And the different opportunity sort of goes like this.
如果你能最擅长利用人工智能，打造全球最高效的企业的运营模式，你的利润率将远超行业平均水平。
If you can be the best at leveraging AI to run the most efficient enterprise business in the world, Your operating margin can be far in excess of the industry.
只要你能破解这个关键点，
And if you can if you can crack that code
你是说毛利和净利吗？
Gross and net, you're saying.
毛利能到九十几，净利也能到四五十。
Gross in the nineties, net in the forties, fifties.
没错。
Yeah.
只要你能破解这个关键点，那买什么就都不重要了。
If you can crack that code, then it doesn't matter what you buy.
是的
Yeah.
所以我觉得现在的问题在于执行层面。
So I think the problem right now is the execution problem.
大多数规模较小的公司根本负担不起优化运营、提升效率的代价。
Most subscale companies cannot afford to go optimize their company and run it better.
如果我们能把公司运营得比谁都好，利润率也更高，市场自然就会认可。
So if we can run our company much better than everybody else and have a higher operating margin, then the street will say fine.
如果你能把一个20%利润率的业务，把它变成你的第一个突破点
If you take something at a 20% margin, make it a Your first m
那真的很难。
and a was really tough.
对吧？
No?
当时他们挺怀疑的，但你后来直接用事实打他们的脸。
Like, they were pretty skeptical and then you kind of shoved it in their face.
他们一开始挺怀疑的，因为来的人连网络安全都不懂，也不懂企业市场，却偏偏在谷歌工作过——而谷歌出来的人到底能不能成功，情况一直参差不齐。
They're pretty skeptical when they found a guy who didn't know cybersecurity, didn't know enterprise show up, who worked at Google and they're you know, the track record of people leaving Google and being successful out of Google is still varied.
所以你的意思是，现在机会敞开了，你……
So basically, you're saying the menu is open and you
我觉得接下来六到十二个月，得看看人工智能到底怎么落地，我们才能真正把它用好，融入企业运营。
I think we need the next six to twelve months to figure out how this AI settles down and how can we use that effectively in enterprises.
很多人总盼着用更少的人来管公司，但我恰恰持相反看法。
I think if you think about it, know, the the the people keep hoping that less people we need to run companies, I actually have a counter view.
我觉得，因为人工智能推动了全面转型，帕洛阿尔托的技术团队规模，反而会比以往任何时候都大。
I think we're gonna have more people at Palo Alto on the technology side than we've ever had before because AI is causing everything to ask for a transformation.
现在我手下的技术人员，比如果没有人工智能的时候还要多。
So I have more technical people today than I would have had if AI didn't exist.
女士们、先生们，帕洛阿尔托网络公司的首席执行官，尼肯什·阿罗拉。
Ladies and gentlemen, CEO of Palo Alto Networks, Nikesh Arora.
谢谢大家。
Thank you, guys.
谢谢您。
Thank you, sir.
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。