저자: HockTan/JimCramer진행방송—제목기준StockologyInsider채널 · 외국인
요약 보기
Cramer는 Broadcom 주가가 고점 대비 크게 빠져 저평가라고 주장하며 매수를 권유. Hock Tan은 AI 인프라 수요가 견조하고 2028년 AI 매출이 200~300억 달러, 주당 이익 30달러 이상이 가능하다고 밝힘. 대다수 고객인 OpenAI·Anthropic의 IPO 지연에도 사모 신용 파트너(Apollo, Blackstone 등)가 자금을 대줄 수 있다고 언급. 엔비디아 대비 맞춤형 XPU의 성능 우위도 주장. AI 규제 논의에 대해서는 안전장치는 필요하지만 개발 감속은 불필요하다는 입장.
다듬은 글
Well, it's a tough time for Cisco to come out this year. The day when the entire AI sector takes a hit, because Anthropic and OpenAI seem to be — I don't know — "furling the sails." I actually think there are some real bargains in this group. Take Broadcom, a manufacturer of chips , networking equipment, technical infrastructure, and software.
Their shares have fallen more than 10% in the last month. Earlier this month, Broadcom reported strong results. She released an impressive multi-year forecast . But their forecast for the current quarter was, I don't know, taken by some as expected, and the stock crashed. I believe Broadcom is still on track for incredible growth because they are at the center of the AI ecosystem.
Because the company has never let us down in all the years that my Chappells Trust has owned it. And at current levels, about 150 points below the June high , the stock may be too cheap to ignore. Don't just believe me. Let's chat with the legendary Hock Tan, President and CEO of Broadcom. Let's learn more.
Mr. Tan, welcome back to "Mad Money." Glad to be here, Jim. OK. So, Hoku, you've given your predictions. It was just fantastic. $350 billion for the next few years, and you sounded very confident. Has anything happened in this whole discussion about AI slowing down that would make you doubt this prediction?
No. Not at all. We see that the demand for computing infrastructure for developing AI or advanced AI models ... Yes. ... and the output for the products they offer the world remains very high and, I believe, very stable. Okay, you're also an expert in finance, and you've managed to attract great companies to create creative sources of funding.
Do you need, say, OpenAI or Anthropic to have an IPO anytime soon ? I know OpenAI is saying they're going to delay it to have, you know, complete certainty, because I would be worried that if they don't do an IPO or have difficulty with it , I might not get paid. That's a valid question, we have six clients, just six, but they all create cutting-edge models, LLMs as we know them.
Because in my opinion, the real value that will be created by this phenomenon of generative AI and the corresponding computations comes from companies, enterprises that create their own advanced models. Of the six companies, two, OpenAI and Anthropic, I consider to have weak cash flow, Yes. but high praise because they have great technology and we create very good products so they can continue to improve their cutting-edge models.
So , given their results, and they have already shown significant success, especially OpenAI, we are creating a platform by attracting powerful financial partners. These are not traditional financial partners, but those who are ready to look at these companies in a new way , evaluate the credit, opportunities and potential they have, and are ready to provide them with credit.
We're just helping them by creating a platform, and companies like Apollo or Blackstone, who love these kinds of opportunities, are getting in on the action, because the stakes are definitely higher than in Treasury bonds. And they create a certain confidence. You mean, if, say, OpenAI comes out and says, "Listen, we're not going to do an IPO." These two companies, which I know by the way and which are very pragmatic, are ready to lend.
Absolutely ready. We are convinced that it is not at all difficult to find such financial partners who are ready to come and finance the infrastructure for these companies, even considering that the demand from these two players will be huge over the next few years. Perfectly. Now let's talk about the last quarter.
I was very worried and told the club members: look what Hock Tan says—27 and 28 are going to be incredible. He always justified trust with his predictions. If that's the case, I think you could make up to $20 per share in 2028. I know you said in your speech that decisions on capital allocation will be made in December, but if you know you're going to do that, explain to people why you shouldn't invest on a large scale right now and why .
Well, we were conservative in estimating the performance for the 28th year. We claim that only on AI revenues, which are the custom AI accelerators that we provide to our six clients and the networking solutions that go with them. We are a chip manufacturing company. We sell these chips, which become part of the infrastructure.
We believe that on these products alone we will reach 20–30 billion dollars in 2028. This, along with our current non-AI software and semiconductor businesses, as I mentioned, will exceed $30 per share in 2028. Oho. Not $20, but $30 per share. So, if that's the case, I mean, I would buy, you know, I would n't wait until December.
I would buy up a huge amount of shares on the open market. That's exactly how I look at it. But we will live up to our rating, Jim. Okay, because I think sometimes people don't look at the numbers. They just draw conclusions. No, conclusions should be based on numbers. There is something you have taught me over the years as I have read your conference calls.
You spoke quite harshly about Nvidia's product, even though we know that Nvidia is a great company. Yes, that's right. You find your Habanero chip very competitive compared to Grace Blackwell when it comes to the conclusion. Can you explain such a bold idea, knowing that everyone considers Nvidia untouchable?
No, here's the most interesting part of all this—the technology. If you're co- designing a silicon XPU , as we call our client accelerator. At every stage, from layout to transistors, you design the chip together with the advanced modeler, algorithmists, data scientists who tell you what workloads they need, build the architecture, and create a chip that meets those needs, it will always perform much better than a general purpose GPU.
This is the main premise of all this. Perfectly. Now let's talk about what happened this weekend. Dario Amodei wrote a meaningful note, but at the same time, some may perceive it as arrogance. If you make everyone stop and you are ahead, like in a Formula 1 race, if you have a flag that says " slow down," then you will win under any circumstances.
Do you think there were strong arguments, or was it more a matter of benefit, or both ? Well , I'll stop at the strong arguments. There are some very real and weighty points. First, and this is very important: AI, generative AI, creating the very advanced models that embody generative AI will create enormous value.
OK. for us, for companies, for people, for consumers, because it makes us smarter and more productive. So, in many ways, it's like going back to the Industrial Revolution. This is a phenomenon that is very similar to us having access to intelligence to be able to perform our daily tasks, whether you are a business or a consumer, with this tool.
But it's a tool. True. We use this tool. People use this tool. And, like any tool, yes, it is important to put controls and safeguards in place regarding how we use it. But, in the end, it is still a tool to force our society, humanity, to achieve a better standard of living, but a tool nonetheless.
This is not a living being that will run wherever it wants, without our guidance and control. So management is important. We won't all be dead by 2030. This doesn't seem consistent with the scenario you just laid out. No, it's a great tool. This is what he is. That's all he is. Well, you know, look, I knew you'd make me...I mean, you're a rational person.
Do you think some of the comments in Dario's notes, perhaps from people involved in these cutting-edge technologies, are a little too intense, a little too emotional? I don't know the reason, but I know it's important, and I agree with Dario that we need the safety nets in place. Any safeguards are not just for the user, but also safeguards for companies like Broadcom and anyone else who chooses to deploy them the way we do in our infrastructure.
That we are introducing safeguards. do it. We don't need to slow down to do this. No, I think the way I see it...Look, I don't know about learning, but when you want to turn an inference into a product, I see it continuing to be very, very powerful. Let's leave it at that. This is Hock Tan, president and CEO of Broadcom, and Halftime has a position.
Mad Money returns after a break. Hawk, thank you very much. Thank you, Jim. Certainly. I was happy to help. Next: Leading AI labs call for a slowdown in technology development, Cramer explores how it might work, or if it's even possible, with George Kurtz. Next. Don't miss a second of Mad Money. Follow @JimCramer on X.
Have questions? Tweet Kramer #MadMentions. Email Jim at [email protected] or call us at 1-800-743-CNBC. Did you miss something? Go to madmoney.cnbc.com .
Comments 0