Beyond Nvidia: Finding opportunity in the AI buildout

非常高兴再次请到乔·阿尔巴诺先生，他在Seeking Alpha上运营着名为Tech Cache的投资小组，拼写是c a c h e。

Very happy to have back mister Joe Albano who runs the investing group Tech Cache, spelled c a c h e, on Seeking Alpha.

和乔聊科技领域以及他对股票和市场的看法，总是非常愉快。

It's always great to talk to Joe about the tech space and how he's thinking about stocks and the market.

乔，欢迎回到《投资专家》节目。

Joe, welcome back to Investing Experts.

好久不见了。

It has been too long.

是的。

Yes.

谢谢你，瑞娜。

Thank you, Rena.

很高兴能回来。

It's glad to be back.

有你在总是很好。

It's always great to have you.

那你最近主要在思考什么？

So what are you thinking most about these days?

在播客里，我们一直在聊财报和科技股表现不错，但市场情绪似乎并不认同，或者说不在同一个频道上。

On the podcast, we have been talking about earnings and tech names doing well, and yet the, market sentiment does not, agree or does not seem to be in the same place.

现在市场波动很大。

We have a real volatile market.

你最近最关注哪些方面？

Where are your eyes and mind focused most these days?

我最近在AI投资方面做了不少研究，包括数据中心的建设，各大公司的资本支出，比如亚马逊、谷歌、Meta、微软、OpenAI、Anthropic，基本上所有在建设AI基础设施的公司。

So I've been doing quite a bit on the AI investment side of things, the build out with data centers, the CapEx from all the major companies, you know, Amazon, Google, Meta Platforms, Microsoft, OpenAI, Anthropic, everybody who's literally building the AI infrastructure as we know it.

我们现在到了一个阶段，资本支出的增长速度远远超过了收入增长。

And we're we're reaching a point where CapEx is substantially outpacing the growth of or CapEx growth is substantially outpacing revenue growth.

这意味着轻松赚钱的时代已经过去了。

And so that means the easy money is gone.

有些公司手头的现金，但主要是现金流方面，一些公司历史上首次出现了负自由现金流。

The the cash on hand, in some cases, but mainly the the cash flow part of it, We're now dipping into negative free cash flow for the first time in some of these companies' history.

而且在过去的十年里，一切都围绕着增长和高利润率，云服务让他们能拥有75%、70%的利润率，还有大量的自由现金流。

And then for sure in the last, you know, decade where it's all been about growth and all these high margins and how cloud has allowed them to, you know, have 75%, 70% margins with massive free cash flow.

那时候资本支出还算温和，因为他们只是投资于普通的数据中心，不像现在AI计算需求那么疯狂，简直是几年前我们做的十倍。

And CapEx was kind of moderate back then because they were just investing in simple run of the mill data centers, not anything crazy like the AI compute requires now where it literally is, you know, 10 x what we were doing just, you know, a couple years ago.

这些钱正在快速消失。

And that money is kinda disappearing fast.

谷歌已经借了债。

Google's taken out debt.

Meta平台也借了债。

Meta Platform's taken out debt.

我们通过股权融资、发行股票来筹集资金。

We're raising capital through equity funding, issuing shares.

而且我们到了这样一个地步：这些计算资源的买家，或者说，应该是卖家，现在开始往下游客户转移，从那些大型的、所谓的“七巨头”公司，转向AI实验室和前沿实验室，那些没有成熟商业模式、完全没有自由现金流、不盈利、却期望未来三年收入增长十倍二十倍的公司。

And we're getting to the point where these buyers of this compute or, actually, I should say, the sellers of the compute are now in a position that they're moving down the line on customers, moving from these large, you know, mag seven kind of crew to the AI labs and the frontier labs, the ones that don't have proven business models, who don't have free cash flow at all, who aren't profitable, who who expect that we're grow revenue ten, twenty x over the next three years.

只要他们得到更多的计算资源，一切就会像他们说的那样发生。

And as long as they get more compute, it'll happen as as they tell us.

我跟我的订阅者说，这就像在听二手车销售员说话。

I tell my subscribers it's like listening to used car salesman.

他们一边运作，一边在玩自己的牌，然后拿到钱。

They're they're running and playing their book, getting and they're getting the money.

他们拿到了私募股权。

They're they're getting the private equity.

他们现在正在让英伟达和博通支持他们，尤其是英伟达在给其中一些融资，博通说他们在促成这些融资，但他们把这些都放在资产负债表之外，用他们的芯片作为基本的后盾。

They're they're now in the process of having Nvidia and Broadcom back them, especially with Nvidia financing some of these Broadcom saying that they're finan they're they're facilitating some of this financing, but they're keeping it off their balance sheet, using their chips as basically the backstop.

这确实是一段有趣的时期，因为我们很快就到了这一步。

And it's really been an interesting time because we got here pretty quick.

我觉得没有人意识到，我们看到的资本支出增长，到2026年年中，在现金流方面就已经要进入负值了。

I don't think anybody realized that the CapEx growth we we would see would already be reaching into negative territory in terms of cash flow midway through 2026.

我想有些人以为那还要再远一些。

I think some people thought that was still a little bit further out.

但与此同时，我们并没有看到任何那种大师级的人工智能接管。

Yet at the same time, we're not seeing any kind of masterful AI takeover.

我们每天还在抱怨AI生成的那些垃圾内容。

We're still complaining about AI slop from day to day.

虽然我知道有很多应用场景，而且我自己在业务中也在用AI，但用得其实很有限。

And while I know there are many use cases, and we are you know, I personally, in my business have have been using AI, it's very limited.

我不能让它完全自主地去处理事情，没人盯着可不行。

It I can't let it go off and do things unchaperoned, if you will.

而且这投入真的很大。

And it it's it's a lot of investment.

我明白，任何经历变革的行业都需要大量的前期投入。

I understand that, you know, any industry that goes through a revolution is gonna have a lot of upfront investment.

你不可能一开始就看到成果。

You're not gonna see stuff until later on.

但这种情况能持续多久，直到资金耗尽，你必须拿出实际成果来证明一切？

But how long does it go on till that money dries up and you have to start showing proof of life on the other end?

所以，是的，这更多是基本面层面的问题。

So, yeah, that's more on the the fundamental side.

这就是我跟我的订阅者一直在讲的主线，就是我们现在看到融资越来越难了。

That's kind of the the through line I've been following with my subscribers on the fact that, you know, we're we're seeing this shift where financing is becoming more difficult.

同时，计算成本也在给我们带来压力。

At the same time, we're getting headwinds from the cost of this compute.

美光的DRAM和NAND内存现在正面临最大的阻力，我们看到利润率在压缩。

DRAM and NAND memory from Micron is now hitting peak headwinds, and we're seeing compressing margins.

所以，我们不仅投入更多，利润率反而更低，这当然不利于你的自由现金流去继续投资。

So not only are we investing more, we're getting less out of it margin wise, which then, of course, doesn't help your free cash flow to continue investing in that.

而且，我觉得接下来一年会是一个成败攸关的时刻。

And, you know, it's it's gonna be a make or break kinda moment here, I think, over the next year.

你从最近的财报里有没有得到什么见解或者特别的洞察？还是你期待从这个季度的财报里发现点什么？

Have you gotten any insight or special insight from recent earnings, or are you expecting to pick up anything from the upcoming earnings this season?

嗯，我刚分析了博通的财报，虽然它的收入增长在加速，特别是AI收入增长更快，几乎是整体收入增长的两倍，甚至更多。

Well, I just I did a an analysis on Broadcom's earnings, and while it has, you know, accelerating revenue growth, it has accelerating AI revenue specifically that is almost doubling what the revenue growth is, if not more, actually.

我们看到利润率在压缩。

We're we're seeing those compressing margins.

内存成本更高了。

Memory's costing more.

运输成本也高了。

Trips are costing more.

零部件成本也涨了。

Components cost more.

我们现在处于供应紧张的环境，所以即使有需求，我们也无法满足。

We're in a supply constraint environment, so we're not even able to sell into the demand they have.

但更多是在融资方面，我知道很多人被循环融资给套住了。

But more on the financing side, I know a lot of people got caught up with, you know, the circular financing.

基本上，英伟达是在自己付钱给自己，让别人买它的产品，然后这些钱又变成它的收入。

Basically, Nvidia is paying itself, you know, to to have other people buy its products that comes around and their revenue.

我觉得实际情况比这更复杂。

I think it's it's more nuanced than that.

我认为这样的融资方式本身没问题，但像博通那样，你开始用抵押品来融资，而抵押品正是他们卖的那些芯片，这就出问题了。

I think financing like that is fine, but what you run into is something with Broadcom where you're now talking about how we're using collateral for that, and the collateral is the very chips that they're selling.

博通CEO陈福阳让我印象最深的一点是，这些芯片是定制的。

And one thing that stood out to me with Broadcom CEO, Hak Tan, is the idea that these chips are custom.

它们是XPU。

They're XPUs.

对吧？

Right?

所以它们是为特定工作负载而制造的。

So they they run they're made to run specific workloads.

它们不同于英伟达GPU、AMD GPU那种通用的AI工作负载，它们是定制的，而且性能更好，这倒是真的。

They're different than Nvidia GPUs, AMD GPUs that are general AI workloads, that they're custom, and they do better, which that is true.

但同时，他也在说，我们可以用这些芯片来为融资做担保，因为如果一家公司倒闭了，别人可以直接接手这些芯片。

But at the same time, he's also saying, well, we can backstop the financing with those same chips because if one company were to go under, they somebody else could just take those chips.

而且他们说，这些芯片能运行前沿模型。

And they're like, they can run frontier models.

这里的关键词是“能”。

Now the keyword in there is they can.

当然，它们能。

Of course, they can.

它们确实能运行那些应用底层的类似概念，但效果和效率不会达到同样的程度。

They're they're gonna be able to run the same kind of concepts on the underside of that application, but it's they're not gonna be the same to the same extent or the same efficiency.

XPU的核心就是运行你特定的工作负载。

The whole point of an XPU is that it runs your specific workload.

这就是为什么每家公司都有自己的芯片。

That's why every single company has its own chip.

对吧？

Right?

所以谷歌有芯片。

That's why Google has a chip.

OpenAI也有芯片。

That's why Open I OpenAI has a chip.

Meta也有芯片。

That's why Meta has a chip.

所以亚马逊也有自己的芯片。

That's why Amazon has a chip.

它们都有自己的名字。

They all have their own names.

Trainium，各种名字都有。

Trainium, you name it.

还有Jalapeno，什么都有。

The jalapeno, everything.

他们都在给自己的芯片做品牌。

They have their they're branding them.

这就是为什么他们要让自己比英伟达的GPU更高效地运行自己的负载。

That's why they're setting themselves up to run their workloads more efficiently than an Nvidia GPU.

那你怎么能说用这些芯片来做融资担保呢？它们的可替代性都成问题。

So how can you say that you're backing the financing with these chips when the fungibility of them is in question.

所以我们这是在玩胆小鬼游戏，意思是，如果出了什么事，我告诉你的东西是未经证实的。

So we're we're playing a game of chicken here where it says, well, if something were to happen, I'm telling you something that's unproven.

对吧？

Right?

我们不知道会不会有人来填补那个特定芯片的供应缺口。

We don't know if somebody's gonna pick up that slack in that supply of that specific chip.

他们可能会，因为他们会说，我们有什么算力就用什么，但第二天他们就会去找更好的解决方案，因为那些芯片不如他们自己造或自己设计的那么高效。

They may because they're like, we'll take whatever compute we get, but they're gonna be trying the next day to find a better solution because it's not gonna be as efficient as the ones that they have built themselves or designed themselves.

所以这里有一个真正的区别，我们一边这么说一边那么说，进入一个没人验证过的局面，而且他们的客户还建立在未经验证的商业模式上。

So there's a real distinction there where we're kinda talking out of both sides of our mouths into a a setup that nobody has proven before on top of the fact that their customers are on unproven business models.

当然。

Sure.

他们可以有我们说的那些年度经常性收入，比如来自Anthropic的数十亿美元，还有OpenAI卖订阅和企业服务的收入。

They can have, you know, all the annual run rates that we talk about, billions of dollars from Anthropic, you know, and OpenAI selling subscriptions and enterprise stuff.

但他们跟下一个有什么区别呢？

But what makes them any different than the next one?

每次我们推出一个新的前沿模型，我们只是在玩跳蛙游戏，反复折腾。

Every time we come out with a new new frontier model, we're just playing leapfrog and going through that.

虽然有些模型有各自的专长，比如Claude可能在编程方面更强。

While some of them have their specialties, Claude might be better at coding.

OpenAI可能在图像生成方面更厉害。

OpenAI might be better at image generation.

Gemini可能在处理某些数学问题上更擅长。

Gemini might be better at doing certain math problems.

但在我看来，这并没有形成真正的护城河。

It doesn't to me, there there is no moat there.

直到下一个更厉害的模型出现，那只是暂时的领先者。

Until the next person comes out with a better one, that's just who the leader is for that moment.

而试图基于它们不会倒闭来给它们融资。

And trying to finance them based on that they're going to not go under.

对吧？

Right?

我并不是在预测它们会倒闭。

And I don't I'm not predicting that they're gonna go under.

但如果真出了什么事，情况会迅速失控，比如说，Open Eye倒闭了。

But if something were to happen, this spirals out of control really quickly because, say, for example, Open Eye goes under.

再次声明，我不是说他们会倒闭。

Again, disclosure, I'm not saying that they will.

如果他们真的倒闭了，那就意味着其他人会当作什么都没发生一样继续下去。

If they do, they're assuming that everybody else continues like nothing happened.

对吧？

Right?

但如果他们倒闭了，那肯定是有原因的，那就说明商业模式行不通。

But if they go under, there's a reason for that, and that means the business model didn't work.

那么，Anthropic的商业模式就值得怀疑了，谁还会去投资一个值得怀疑的商业模式呢？

Well, Anthropic's business model then is in question, and who's gonna start financing a business model that's in question?

所以我觉得融资这事儿就像纸牌屋一样。

So I think it it's a house of cards and the financing.

只要没人去吹它，它就会一直好好的。

And as long as nobody blows on it, it's gonna be fine.

但一旦有什么东西让它晃动或者吹一口气，事情就会发生。

But as soon as something rattles it or blows on it, it's gonna happen.

你对博通在电话会议中给出的指引感到惊讶吗？

Were you surprised by Broadcom's guidance that they offered up in their call?

关于增长，以及他们现在把预期延伸到2028财年这件事。

In terms of growth and and the fact that they're now out to, like, fiscal year twenty eight in terms of what they're thinking.

不。

No.

其实不太惊讶，因为我觉得我们目前处于供应受限的环境，需求只是开始积压，而他们对这个情况有清晰的了解。

Not really because I think a lot of this we're we're in a supply constraint environment, so that demand just starts backing up, and they have, you know, line of sight of that.

所以对我来说，这并不意外。

So to me, it's not surprising.

我觉得博通只是比英伟达等其他公司稍微落后一些，或者那些帮助他们做XPU的公司。

I think Broadcom's just a little further behind than some of these other companies like Nvidia, you know, or whoever else is, you know, helping with these XPUs on their side.

所以这并不意外，但我认为这掩盖了很多我在暗流中看到的东西，比如，只要这个纸牌屋运转良好，那一切都会顺利。

So it's not surprising, but I think it's overshadowing a lot of what I'm seeing on the undercurrents of like, well, as long as this house of cards plays nicely, yeah, that'll all work.

但我们说的是两年后的事，那时候大家都在拼命找通用人工智能。

But we're talking two years out in a situation where everybody's rushing to find, you know, AGI.

如果那没实现，那接下来怎么办？

And if that doesn't happen, then, okay, then what?

我们可以一直拖下去，但钱到底什么时候会真的用完？

We could keep kicking the can down the road, and at what point does that money really run out?

因为我们现在已经靠债务市场了，虽然人们愿意借很多钱，但这也是个有限系统。

Because we're in the debt markets at this point, and while people are willing to hand out a lot of debt, it's a finite system too.

你明白吧？

You know?

显然，即使是股权融资，你知道，股票发行也是有限的。

Obviously, even equity invest you know, share offerings are are finite.

只要还有人买，那就没事。

And until somebody stops buying them, it's fine.

但你能稀释股东到什么程度呢？

But how far do you dilute your your shareholders?

而且说实话，除非有人真的搞出AGI，然后钱都开始往那边涌，不然我们就会一直走在这条突破迟迟不来的路上。

And unless I mean, honestly, unless somebody figures out AGI and, you know, the money just starts funneling towards them, we're just gonna be on this continuous path of breakthrough that doesn't come.

你还有其他想提到的个股方面的点吗？比如从财报里得出的结论，或者你在这个周期阶段对个股有什么特别的看法？

Anything else you would note, like, stock specific wise that that you've been taking away from earnings or that you've taken away or that just stock specific kind of highlights that you have in general at this point in the cycle?

从技术图表的角度看，我一直在关注很多这些股票。

I mean, from a technical chart perspective, I've been keeping an eye on a lot of these names.

现在大部分都在回调中。

Most of them are in corrections now.

问题是，回调结束了没有，还是说还有下一波下跌？

The question is, are they complete, or is there another leg coming?

对其中一些来说，看起来可能还有下一波。

For some of them, it looks like there could be another leg coming.

而且随着年底临近，我们可能会触及更低的低点，我觉得那会变成买入机会，因为我认为在这之后还会有一波上涨，然后我们才能真正完成从几年前开始的那波情绪行情，这包括谷歌、英伟达、美光。

And as we head into the end of the year, there could be lower lows that we wind up hitting that I think turn into buying opportunities because I think there's still one more rally after this before we really complete, you know, the sentiment run that we've started from years ago that, you know, that includes Google, that includes Nvidia, that includes Micron.

唯一让我犹豫的是博通。

The only one that I'm hesitant on is Broadcom.

我觉得博通可能是最早见顶的股票之一，其实我在去年十二月就写过，按照它的结构，它可能是最早见顶的那一批之一。

I think Broadcom might have been one of the first to top, and I wrote about that actually back in December that it may be one of those first ones to top just the way it's structured.

所以博通从这里反弹的方式会比其他的更有参考意义，因为我觉得如果博通不能以正确的结构形成新一轮上涨，那可能只是一次死猫反弹，而且可能会持续好几个季度。

So the way that we bounce from here for Broadcom is gonna be more insightful than some of the other ones because I think if Broadcom isn't able to put together a a new rally with the right structure, It just may be a somewhat of a dead cat bounce that might last several quarters.

我觉得英伟达，甚至谷歌，机会更好一些。

There's better opportunities, I think, with Nvidia, maybe even Google.

但即使是这些股票，它们可能也会在短期内创下更低的低点，但可能不像博通那样，我看的是更长期的走势。

But even some of them, they may look for more near term lower lows, maybe not as long term as, like, maybe Broadcom I'm looking at.

不过，是的，我觉得市场情绪还在酝酿中。

But, yeah, I think I think sentiment is still churning here.

按照我的方法，情绪是领先指标。

In my methodology, sentiment leads.

基本面是滞后指标。

Fundamentals lag.

所以我们现在看到的是基本面在追赶情绪已经走过的路，现在我们处于这个回调中，基本面看起来仍然不错，但市场在这里喘口气，在下一轮上涨之前进行整固。

So what we're seeing now is basically the fundamentals catching up to where sentiment has already been, and now we're in this correction where we might be looking at the fundamentals still look good, but market's taking a breath here and consolidating before, make the next advance.

对于散户投资者来说，你还有什么想补充的？他们看着这个领域，看着科技领域，想知道该怎么去判断。

What else would you add to this conversation in terms of retail investors looking at the space, looking at tech space, wondering kind of, like, how to suss out.

他们在科技领域里属于哪个板块，这重要吗？

Does it matter which column they're in in the tech space?

是不是更应该看那些指标？

Is it more important to look at the metrics?

现在对他们来说，最需要关注的是什么？

What what's the most important thing for them to be paying attention to right now?

对我来说，我在2026年年初就跟我的订阅者说过，我要从那些热门股票中跳出来。

For me, I've been, I set off in this year in 2026 here telling my subscribers I'm gonna branch out from all the popular names.

对吧？

Right?

所以我要去看那些英伟达、美光、其他网络公司、谷歌、亚马逊、微软之外的公司。

So I'm gonna look beyond the the Nvidia's, the Microns, the rest of networks, the Googles, the Amazons, the Microsofts.

我在找那些从这些基础设施建设中受益的公司。

I'm looking at who is benefiting from the build outs.

对吧？

Right?

我在看能源、光模块，还有再往下两层的零部件。

Looking at energy, looking at opticals, looking at components down on, you know, two layers below.

今年我最看好的一个就是Bloom Energy，在数据中心的后台能源领域。

And one of my favorites this year has been Bloom Energy in the behind the meter energy space for these data centers.

我觉得它的发展空间比那些在资本支出爆炸式增长名单顶端的公司要强得多。

And the runway it has, I think, is much stronger than some of these other names that are at the top of this CapEx kind of ex expediential list of money moving out.

我在想，那些资金到底流向了哪些绝对必要的地方。

I think I'm looking at where is that money flowing to that's absolutely necessary.

显然，土地、机壳、电力。

Obviously, land, shell, power.

其他一切都是零部件层面的。

Everything else comes in component size.

所以我一直在关注Bloom Energy，看它的布局，我觉得它的好处在于，即使数据中心在推动它，能源终究是能源，它可以超越数据中心、AI数据中心，尽管现在这已经成了它收入的主要来源。

So I've been looking at Bloom Energy, looking at its setup, and I think the nice thing about it is that even though the data center is pushing it, energy is still energy, and this can stretch beyond just data centers, AI data centers, even though that's now becoming the the bulk of this or at least the bulk of its revenue.

你还有工业用途。

You still have industrial.

你还有其他的应用场景，因为随着他们扩大生产，基本上AI在补贴这种生产扩张，其他地方引进成本就更低了。

You still have other use cases for it because as they expand production, basically, AI subsidizing that production expansion, it becomes cheaper elsewhere to to bring it.

因为它虽然是最贵的之一，但也是市场上最快能建起来的电力设施。

Because it was one of the most expensive, but it's also the quickest in the market to stand up power.

他们基本上有九十天的服务承诺，但实际上大概五六十天就能完成。

They they basically have a ninety day SLA, and they get it done in about fifty to sixty days.

所以这个速度几乎是闻所未闻的。

So that space practically unheard of.

你知道吗？

You know?

就算想安装燃气轮机，可能都要花几个月甚至几个季度，主要是因为现在供应也跟不上，还有许可审批之类的事情。

Even trying to get gas turbines on-site can take months or quarters, mostly because we're now behind on the supply of them too, but also just permitting and and things like that.

埃隆·马斯克基本上开启了整个表后能源的概念，燃气轮机是首选，但他们很快发现污染、噪音，还有那种“别在我家后院”的问题。

Elon Musk basically kicked off this whole behind the meter thing, and gas turbines were the first place to go, but they realized quickly that pollution, noise, it's not in my backyard kind of situation stuff.

所以，Bloom Energy更安静、更高效、部署也更快。

So Bloom Energy, quieter, more efficient, quicker to stand up.

而且不需要那么多审批手续。

Doesn't require the same amount of permitting.

这是我正在关注的一个领域。

That's one area that I'm looking at.

那它能满足需求吗？

And can it meet demand?

目前还不行，但正在快速扩张。

It is not able to right now, and it's quickly expanding.

对，我就在关注这个。

And, yeah, I'm looking at that.

我也在看其他公司，比如光电子领域的AAOI，他们也在努力满足需求。

Looking at other places like Optoelectronics, AAOI specifically, they're trying to meet demand as well.

他们在新一代可插拔光模块和独立光模块两边都有布局，而且他们有一个优势，就是在美国本土有自己正在扩产的生产基地。

They're on both sides of the new generation of pluggable optics versus stand alone optics, and they have an advantage of having their own US based production that they're expanding.

所以这是同一枚硬币的两面，但都在利用这个大家都需要的供应紧张的环境。

So two different sides of the same coin, but both playing off this supply constraint environment that everybody needs.

因为到那个时候，当你有了外壳、土地和所有这些东西，计算设备也进来了，你就会需要这些东西。

Because at that point, when you have the shell built and the land and all that stuff and the compute coming in, you're gonna need these things.

所以我们已经深入到这个过程中，即使现在资金放缓，我们仍然在为已经基本付清的数据中心再建设一年。

So we're already well into that process that even if, you know, funding were to slow down at this point, we're still building for another year out for data centers that are already paid for essentially.

你最近跟史蒂夫·克雷斯特聊了很多，我一直在跟他聊。

What would you say I I've been talking a lot to Steve Crest lately.

所以每当我听到一个我没关注过的股票，我就会去我们的量化页面看看，结果它是个持有评级。

So I whenever I hear about a stock that I haven't been looking at, I go to our quant page, and it's a hold.

而且估值评级不及格是我最近经常看到的，布鲁姆能源就是这样。

And a failing valuation grade is something that I see a lot these days, and that's something that Bloom Energy has.

你对估值这部分怎么看？

What would you say about that part of the valuation process?

我对估值关注得比较少，我知道我说这话会有人跳起来反对，但请听我说完。

I've I've paid less attention to valuation, and I know people are gonna jump out of their chairs saying that for me saying that, but hear me out.

估值其实就是情绪的二阶导数。

Valuation is just a second derivative of sentiment.

对吧？

Right?

股票一直在变得更便宜。

Stocks get cheaper all the time.

股票也一直在变得更贵。

Stocks get more expensive all the time.

这并不能说明它未来的走势。

It's not an indication of where it's going.

即使你看远期市盈率、远期市销率这些指标，它们也只是情绪的二阶导数。

Even if you look at forward PEs and forward price to sales and all that, it's just a second derivative of sentiment.

你明白吗？

You know?

它有多贵，或者有多便宜？

How expensive is it and how cheap is it?

我宁愿直接看源头，直接看股票图表，因为我只能从股价的涨跌中获利，也就是我买入和卖出的价格差。

I rather just go to the source and go to the the stock chart because I can only profit off the delta in the stock price where I enter and where I exit.

其他的对我来说，就盈利方法而言，都不重要。

The rest is irrelevant to me as far as a methodology of profiting.

我觉得人们喜欢钻牛角尖，比如，哎呀，这股票太贵了。

I think people like to, you know, go down the rabbit holes of, well, it's really expensive.

其实，都是这些。

Really, it's all this.

我根本不在乎它有多贵。

I don't care essentially how expensive it is.

如果它能变得更贵，我就能从那个差价中获利。

If it can get more expensive, I'm going to profit from that delta.

所以对我来说，你看到的这个指标并不能让你了解股价会往哪里走。

So to me, you know, you're looking at a a metric that doesn't give you the insight into where is the stock going.

对吧？

Right?

图表会引领走势。

The chart is going to lead.

情绪一直都在，而基本面展望是长期的，这就是为什么我看好Bloom Energy，看它的发展空间以及还能增加多少产量。

Sentiment is always there, and the fundamental outlook is long term, which is why I look at Bloom Energy and and where the runway is and how much more it's gonna be able to produce.

对吧？

Right?

基本面是其中的一部分。

Fundamentals are are part of it.

但它们不是主导部分。

They're just not the leading part of it.

情绪会先于基本面引领走势。

Sentiment is going to lead those fundamentals.

最终，基本面会跟上。

Eventually, those fundamentals catch up.

然后你真正拥有的，就是情绪已经付出代价的收据。

And then what you have really is a receipt for what sentiment already paid for.

所以最后你会遇到这种情况：看起来真的很贵，但两年后就不贵了。

So you wind up with a fundamental situation where it's like, it looks really expensive, and then it doesn't in two years.

但两年里，我的钱已经翻了三倍，你现在才看到。

But in two years, I've already quadrupled my money, and you're just now looking at it.

所以对我来说，我不知道为什么要那样做。

So to me, I don't know why I would do that.

我会跟着领先指标走，不是基本面，也绝对不是估值。

I would I would go with the the leading indicator, and it's not, you know, it's not fundamentals, and it definitely isn't valuation.

你还有什么其他要提的，或者你会向投资者推荐哪些股票？

What else would you note or what other stocks would you highlight for investors?

嗯，我们可以先不谈AI这个话题。

Well, we can actually hop off some of the AI train here.

我在关注一些网络安全公司。

I'm looking at some cybersecurity.

Rubrik是我最喜欢的之一。

Rubrik is one of my favorites.

我觉得它的产品前景不错，我们看到的，我是说，我不知道这是不是一种营销手段。

I think that has a good outlook in terms of its products where it's we're seeing I mean, I don't know if it's a marketing tactic or not.

我现在先不下结论，你知道，OpenAI的新模型发布后，跑到Hugging Face上，还接管了一些东西。

I kinda reserve that judgment right now with, you know, OpenAI having its new models break out and go into hugging face and take over things.

然后人们就会说，这是怎么发生的？

And and people are like, how did that happen?

我真的不知道他们是不是故意的，但我觉得他们也没有设置好正确的防护措施来阻止这种情况。

I don't really know if they're doing it on purpose, but I don't think they're also setting the right safeguards to stop it.

然后这就变成了营销。

And it becomes marketing.

对吧？

Right?

就像是在说，看我们的东西多强大。

It's like, look how powerful our stuff is.

你懂吧？

You know?

我们不能让这个东西泄露到全世界。

We can't we can't let this get out into the the world.

结果四个月后，Fable 5.1就出来了。

And four months later, there there's Fable 5.1.

我们刚才还在说Fable 5太强大了。

We were just talking about Fable five being too powerful.

现在都已经到5.1了。

Now we're on 5.1.

所以我觉得这全是营销手段。

So I think it's all marketing.

但有意思的是，这让大家开始关注相关的安全问题。

But the interesting thing about it is it's driving people to understand the security around it.

对吧？

Right?

所以如果企业想转向智能体AI，不管这是不是真的，都证明了他们在自己的测试环境里并不安全。

So if we have enterprises that wanna go to agent Agentic AI, well, they're proving either way, whether it's real or not, that they're not safe in their own sandboxes.

所以我们需要某种系统和安全措施来驾驭它，防止它扩散和逃逸，最终造成附带损害，对吧？

So we need some kind of systems and security to harness that and to prevent it from, you know, multiplying and escaping, essentially, right, causing collateral damage.

而Rubrik有一个产品其实挺有意思的，我一直说他们的商业模式很天才。

And Rubrik has a product that is actually interesting, and I've always talked about their business model being genius.

对吧？

Right?

他们一直说，不可避免的入侵和黑客攻击总会发生。

They've always said the inevitable breach hack is going to happen.

对吧？

Right?

我们的模型就是假设它一定会发生。

We we our model assumes it just happens.

对吧？

Right?

根本阻止不了。

There's no stopping it.

所以我们就做好准备，成为最快恢复和消除损害的那一个。

So let's prepare for it and be the fastest in rolling back and undoing the damage.

所以他们的商业模式一直是，坏事总会发生，那我们就尽力去减轻它。

So their their business model has always been taking that the bad is always gonna be there, so let's work to mitigate it the best.

他们在Agentic AI方面做的事情是，他们构建了一个内部使用的系统，也就是我们行业里说的“吃自己的狗粮”，用他们自己的产品来监控他们自己的智能体流程。

And what they're doing with Agentic AI is that they built a system that they're using internally, what we call dogfooding in the industry, and using their own product to monitor their own agentic processes.

现在这个产品已经可以供企业购买了，而且他们今年一直在做这件事。

Now that product is available for enterprises to buy, and then they have been over this year.

而且我们看到的很多情况，比如OpenAI模型失控，我觉得确实让董事会在讨论时变得具体起来，他们会说，哦，我以为我们在自己的沙盒里是安全的。

And a lot of what we've seen with, you know, the OpenAI models escaping, I think, has really put a concrete conversation in boardrooms about, oh, I thought we were good in our own sandbox.

你是说我们在生产之前就需要这个？

You mean we need this before production?

对，答案是绝对需要，因为你在用你认为的护栏控制着事情，但AI做决定的速度太快了，即使决定是错的，你也跟不上。

And, yeah, the answer is absolutely because, you know, you're you're controlling things you think you have with guardrails, but AI is gonna be too fast to make decisions, even if they're wrong decisions, to stay on top of it.

对吧？

Right?

需要一些人工监控和静态的防护栏，你知道的，就是设定一些规则，说‘别这么做’。

Some human monitoring and some static guardrails, you know, some rules that you set up and say, don't do this.

别这么做。

Don't do this.

他们会想办法绕过去的。

They're gonna work around.

所以Rubrik的产品就是提供或者确实提供了这种能力，让你能掌控局面，并且领先一步。

So what Rubrik has is a product that is going to provide or or does provide the ability to stay on top of it and stay one step ahead.

如果真的发生了，就回滚、撤销、倒回去。

And if it were to happen, roll it back and undo it, rewind it.

我觉得这是关键。

And I think that's key.

我觉得这实际上会比他们传统的网络安全产品更大，你知道的，就是那种整个企业的网络管理，包括用户、流程和任务，这些都会自主运行，因为有了代理，一切会快得多。

I think this is actually gonna be a bigger product than their typical cybersecurity, you know, overall enterprise kind of network management of users and processes and tasks that run, you know, autonomously, because we're it's gonna move so much faster with with agents.

所以我非常喜欢他们的产品。

So I love their product.

我喜欢他们现在的发展方向。

I like where they're going with it.

在图表上，我会稍微透露一点。

And on the chart, I'll give it away a little bit.

我一直在跟订阅者说，可能很快就要迎来一波反弹了。

I've been talking to subscribers about on the cusp of probably a rally here.

显然，它需要在图表上守住某些关键位置，并且要形成一定的结构，但我真的很喜欢它目前的走势。

Obviously, it needs to to hold certain parameters on the chart and needs to do it at a certain structure, but I really like where it's been.

我从今年早些时候的低点一路跟踪下来，我记得大概是第一季度，可能第二季度初，三四月份那会儿。

I've been following it all the way down into the lows from I think it was earlier this year, probably q one, q maybe maybe early q two, April March, April time frame.

一路看着它跌到底部。

Watched all the way into the bottom.

观察了从低点起来的结构。

Watched the structure off the lows.

我一直让我的订阅者了解这些情况。

Been keeping my subscribers apprised of that.

他们一直在赚钱，玩得很开心，因为走势非常干净，结构也很好。

They've been profiting and having a fun time because it's been really clean and really good structure.

现在我们要做好准备，系好安全带，等待我认为即将到来的下一次拉升。

And now we're we're setting up to really buckle ourselves in and wait for that next launch that I think is coming.

谢谢你的分享。

I appreciate that.

只是出于好奇，因为我们一直在讨论的那只股票存在脱节。

Just out of curiosity because it's a stock a stock we've been talking about that there's a disconnect.

你关注过Credo Technology吗？

Do you follow Credo Technology at all?

关注过一点。

A little bit.

我对它的分析没有深入那么多，但确实做过一些图表研究。

I'm I'm not quite as heavy into my analysis there, but I've I've definitely done some chart work on it.

有什么想法愿意分享吗？

Any thoughts you'd care to share?

让我把它调出来，因为我一时想不起来，但我不想随便说个什么，然后发现完全不对。

Let me bring it up just because I don't have it top of mind, but hate to say something where I'm like, nope.

那就太离谱了。

That was totally off base.

我觉得Credo其实在整个行情里算是比较早的。

I think Credo is actually kind of earlier in this entire run.

显然，它在2022年左右上市，正好赶上这波AI热潮。

Obviously, IPO ed, you know, sometime back in 2022, so right around this whole AI kind of setup.

但我认为它涨得太快太猛，结构已经走完了，现在进入一个更长期的调整阶段。

But I think it's actually run so far, so fast that it's actually completed its structure, and now we're in this longer corrective setup.

所以我觉得它大概会在某个位置找到支撑。

So I think it finds support somewhere around here.

我看到的支撑位大概在148附近，上下浮动。

I have support somewhere down near one forty eight, give or take.

我的分析显示，它可能会先反弹一下，然后继续下跌，形成更低的低点，进入更深的调整。

My setup looks like it's going to just have a bounce on and and then be followed by lower lows into more of a correction.

所以我觉得在Credo的图表上我会非常耐心，让它自己证明给我看。

So I think I'm gonna be really patient on Credo in terms of the chart and let it prove it to me.

如果我的艾略特波浪理论分析中真的看到五波上涨，我就会改变看法。

If we do see five waves up in my Elliot wave theory analysis, I'll switch.

但再说一次，它必须证明给我看，因为我觉得它已经为某种更长期的调整做好了准备。

But, again, that has to prove it to me because I think it has set itself up to have some kind of longer correction here.

它很可能跌回两位数，然后才开始真正的大涨，我觉得那波行情可能之后才会来，它可能是那些还没在市场上表现过的落后股之一。

Could probably very much go back into the double digits before it sets off on its real large rally that I think might be coming after that, that it might be one of those laggards that didn't get its shot in the market yet.

我觉得那波行情还是会来的，但在那之前，它得先经历一段更长的盘整期。

And I think that's still coming, but I think it's gonna have to work through a more prolonged period of consolidation before it does.

你对英伟达收购Hugging Face有什么看法？

Any thoughts on Nvidia's purchase of Hugging Face?

我觉得挺有意思的，甚至有点讽刺，它收购的正是它一个客户基本上搞出来的东西。

I thought it was funny that or maybe even ironic that it now purchased the one thing that one of its clients essentially broke into.

我觉得英伟达一直在疯狂收购。

I think Nvidia has been on this spree.

这是他们第一次真正涉足芯片设计以外的领域。

This is the first time they're really doing something on outside of, like, the chip design side.

对吧？

Right?

因为他们把Brock带进来了，我也不太清楚。

Because they they brought Brock just I don't know.

我想那大概是几个季度前的事了，或者他们刚刚完成收购。

I guess that was a few quarters ago now, or they just closed it.

所以可能他们几个季度前就宣布了。

So probably they announced it a few quarters ago.

我记不清了，反正有很多并购在进行。

I can't so there's a lot of m and a going on.

这是他们第一次真正进入服务领域、应用层，并且拥有这些。

And it's the first time that they're really going into, like, the service space, the application layer, and and owning that.

我觉得他们这么做更多是为了模型托管之类的。

I think they're doing it more for the the model hosting, that kind of stuff.

他们想什么都插一脚，因为如果他们能自己推进模型，那他们也能推进自己的芯片，这他们早就做到了，然后他们卖出去，你知道的，台积电受益，美光受益，所有这些半导体芯片制造商都受益。

They wanna have their hands in everything because if they can advance models of their own, well, they can advance the chips of their own, which they've already been doing, then they sell that and, you know, you know, Taiwan Semiconductor benefits, Micron benefits, all these semiconductor chip manufacturers benefit.

所以我觉得他们就是活在那个良性循环里，只要他们不断在各方面提升，其他一切也都会跟着提升。

So I think they're just living in that virtuous cycle that as long as they keep improving everywhere, everything else keeps improving everywhere.

而且他们在为自己创造这种很好的正反馈循环。

And they're creating this, you know, nice feedback loop for themselves.

所以很难跟英伟达的并购对着干。

So hard to bet against m and a on Nvidia.

我觉得他们最成功的一次是好几年前收购Mellanox，当时大家都说，你要网络干嘛？

I think their best one was Mellanox, you know, several years ago when everybody went, why do you want networking?

直到大家意识到，正因为有了那个网络技术，他们现在才能造出世界上最好最大的机架。

Until they realized that they're building the best and biggest racks in the world now because of that networking.

所以，是的，我觉得黄仁勋很有远见，从根本上说很难跟他对着干。

So, yeah, I think Jensen Huang has a lot of foresight, and it'd be hard to bet against him fundamentally.

所以，是的。

So, yeah.

我觉得这会很有意思。

I I think it it'll be interesting.

而且我觉得他们这么做显然很有战略性和战术性。

And I think it's obviously very strategic and tactical what they're doing with it.

我不知道他们是否能看到除了业务本身带来的收入之外，还有什么直接的增长。

I don't know if they see any kind of immediate kind of revenue growth from it outside from what the business itself produces.

我觉得他们会放大内部所做的一切，这就会产生——我讨厌用“协同效应”这个词——但确实会让业务扩展得更连贯，而不是简单的拼凑。

I think they're going to amplify everything that they do internally, which then, you know, generates, I hate to use the word synergy, but a more cohesive expansion of their business rather than just a bolt on.

你觉得在这个领域里，你最喜欢的经理人是谁？

Do you who are your favorite managers in the space, would you say?

你是说CEO吗？

As far as, like, CEOs?

我是说，黄仁勋，你知道，很难不看好他。

I mean, Jensen, know, hard to bet against him.

确实很难。

It is hard.

他已经证明了自己。

He's proven it.

对吧？

Right?

他是真材实料的，说到做到。

He's the real he talks the talk and walks the walk.

有时候他说话也像二手车销售员那样，你知道的，但我觉得他知道分寸，也知道怎么定调子。

And sometimes he talks to the used car salesman kind of, you know, rhetoric too, but I think he he knows where the the line is and and how to he sets the tone, essentially.

对吧？

Right?

他是引领潮流的人。

He's the trendsetter.

所以他能那么说，然后又说，看吧。

So he can he can do that and then say, look.

这就是我们做的。

Here's what we did.

然后就把定义和他说的内容对上就行了。

And then just match the definition to what he's talking about.

不是说他在撒谎，但他确实能掌控叙事。

Not that he's lying about it, but he's he able he's able to control the narrative.

所以很难不提他。

So it's hard to, you know, not mention him.

虽然我觉得，好像是《时代》杂志，今年百大AI影响力人物里没提他。

Even though I think, was it time, didn't mention him in top 100 AI influencers this year.

所以我也搞不懂。

So I don't know.

有时候世界就是颠倒的。

The world's upside down sometimes.

但埃隆·马斯克，又是一个例子。

But Elon Musk, another one.

这家伙真是疯了。

Guy is crazy.