My mission is simple: to make you money. I'm here to level the playing field for all investors. There's always a bull market somewhere, and I promise to help you find it. Mad Money starts now.
Hey, I'm Cramer. Welcome to a special San Francisco edition of Mad Money. Welcome to CR America. I'll be your friend. I'm just trying to make a little bit of money. My job is not just to entertain, but also to teach you. So call me at 1-800-743-CNBC. Tweet me at Jim Cramer.
At one point this morning, it looked like we were just going to crash. Oil had skyrocketed 4% because the savior Saudi pipeline, the one that helped offset the losses from the closure of the Strait of Hormuz, got shut down by a drone attack. There were no substantive talks to stop the war. Interest rates were shooting higher. The yield on the 10-year nearing 5%.
Iran's allies in Yemen seemed poised to shut down the Red Sea. Most important, the biggest theme of our era, artificial intelligence, looked like it was going to be on the ropes because of a self-induced slowdown that could snap shut the biggest spigot of cash in history. Nothing was going right.
But as is so often the case with this market, things changed. Oil somehow reversed hard after being up 4%, closed up 1.5%. The 10-year Treasury hits 5%. Then something happened that's been missing the whole time. Interest rates have been on a rampage. Buyers, actual buyers, came in and decided that 5% was a good yield for 10 years. That's right. Not everyone hates bonds at any price.
And as we got our arms around the forced AI slowdown that the big guns, OpenAI and Anthropic, talked about now seemed to favor, we decided, hey, maybe it wasn't the end of the AI world. In fact, we left the session convinced that not much in the data center would actually change. Hence why stocks rallied nicely off the lows. Dow finishing off just 152 points. S&P declined 0.48%. NASDAQ only down 0.56%. These were remarkable comebacks from these lows. I mean, it was a house of pleasure.
I don't blame anyone for being confused. First, it's the greatest — AI is the greatest thing in the world, right? I mean, the things that Dario Amodei, the exciting CEO of Anthropic, has been saying — that it could cure cancer, so many other diseases — made me feel great. Everybody was pitching it as the next industrial revolution.
Now though, we're hearing all the Skynet jokes about how AI could kill us all. They weren't really jokes. They're genuinely worried inside these places about the possibility of total human annihilation. Kind of a schizophrenic beast when you think about it, huh? So, which is it going to be? How about both?
I don't know about you, but when I read through the Amodei epistle urging the great AI slowdown, I began to wonder whether there could be something else at work here. Most important, I wanted to know whether there was a slowdown that he urged. It's kind of like to me like when you go to those Formula 1 races and they wave the yellow flag meaning you can't overtake. Doesn't that mean Anthropic has to win because it's making so much money and it can't be overtaken?
I like anyone who's thoughtful about the wholesale murder of innocents by a swarm of angry AI agents, but I am a little jaundiced that if the majors are advocating a slowdown that could lock them into a position where they can put down 10 billion and make 30 billion per gigawatt, which might be possible for Anthropic and OpenAI, well, they got the goods. That would put them well past all these hyperscalers we talk about all the time when we try to calculate profitability.
Although I don't believe these AI risks are all a hoax, as President Trump assures us, I'm naturally skeptical when these leading AI labs are pushing policies that could make them a heck of a lot more profitable and leave everyone else behind.
Now, there were real casualties today. Any company that actually makes or designs or constructs part of a data center got blown out in an attack that lasted most of the session. Market floor was littered with the stocks of companies that make fiber for the internet of these warehouses full of servers. But there were also some spectacular advances led by cybersecurity stocks like Palo Alto Networks. Yes. And like CrowdStrike.
George Kurtz will be with us in a minute. They led for a simple reason: when the heads of Anthropic and OpenAI are warning about AI risk, every other company knows that they need better cybersecurity. I don't know how you can get insurance without hiring one of those two companies.
I'm skeptical that AI agents will be able to wipe out the human race. Skeptical, but lots of business owners could certainly have their data held hostage by smart but rogue agents. Will we get an actual slowdown of AI expansion? We need to hear from Broadcom CEO Hock Tan later tonight. He's going to tell us.
Are any of the slap-down stocks worth buying? I would love to tell you yes, but then tomorrow oil could be back up. Rates might fly higher and we'll be on the verge of a Fed meeting where Fed Chief Powell might raise rates, much to the chagrin of the president who appointed him to cut rates. And maybe we'll hear about some new AI hack that makes the MGM and Caesars incident look like child's play.
Bottom line, we wait. We see if the existential threat is off the table. And we remain grateful that despite a hideous setup this morning, the market didn't actually crash. In fact, it was barely down at all. Let's take calls. Let's go to Sunny in Massachusetts. Sunny?
Hey Jim, I just want to thank you, Regina, Jeff, and the entire crew for everything that you do and making me a better investor through the club. I have questions for you.
My first question is, when are you going to make some Cramer pasta sauce with all the vegetables that you make? And two, what are your thoughts on Tesla?
Well, I made 120 jars this year. So, you know, that's spoken for. I'm sorry. Tesla, I wish I could make that, you know, do as well as the sauce. It's at 358. I don't know. I mean, I think it got some approval in China. I thought it was going to bounce today. Right now, my only solution to it is that SpaceX has to buy it. Period. End of story. But maybe they will.
On Mad Money tonight, I'm kicking off a big week west with some of the most important voices in Silicon Valley. Broadcom's AI business is booming, but the stock isn't. I'm going straight to the source to get some answers on where we stand. Then, could self-driving cars be friend or foe to a company like Lyft? I'm sitting down with the CEO to get an update on the business.
Plus, this weekend, Anthropic's CEO issued a stark warning about the speed of AI development, as I've said. So, I'm checking in with CrowdStrike's CEO to find out if we have the capacity to defend against rogue AI agents. So, stay with Cramer.
Don't miss a second of Mad Money. Follow Jim Cramer on X. Have a question? Tweet Cramer #MadMentions. Send Jim an email to
[email protected] or give us a call at 1-800-743-CNBC. Miss something? Head to madmoney.cnbc.com.
Well, it's a tough time to come out to San Francisco, a day when the whole AI complex is getting hammered because Anthropic and OpenAI seem to be, I don't know, pulling in their horns. Actually, I think there's some real bargains in this group. Take Broadcom, the maker of chips, networking equipment, and tech infrastructure and software, with a stock that's down more than 10% over the past month.
At the beginning of this month, Broadcom reported a strong set of results, issued a spectacular multi-year forecast, but their guidance for the current quarter was, I don't know, viewed by some as in line. The stock got slammed. I think Broadcom's still on track to put up some incredible growth because they're at the heart of the AI ecosystem, because the company's never let us down in all the years my charitable trust has owned it.
And down here, roughly 150 points below its June high, the stock may be too cheap to ignore. Don't take it from me. Let's check in with the legendary Hock Tan, president and CEO of Broadcom, to learn more. Mr. Tan, welcome back to Mad Money.
All right, so Hock, you laid out a forecast that was so fantastic, $350 billion in the next two years, and you sounded very certain. Is there anything that's happened in this whole debate about the AI slowdown that would give you pause to that prediction?
No, not in the least. We see the demand for compute infrastructure for AI development, for AI frontier models and inference for the products that they feed to the world as continuing to be very strong and, I believe, very durable.
Okay. Now you are also an expert in finance and you've been able to get some really great companies to work with you to create a creative source of finance. Do you need, say, OpenAI or do you need Anthropic to do an IPO soon? I know OpenAI is saying they're going to put it off in order to have some, you know, perfect timing, because I know I'd be worried if they don't do an IPO or they have a hard time doing it that I might not get paid.
That's a fair question, and we have six customers, only six, but they are all creating frontier models. ILM as we know it, because my view is the real value to be created out of this phenomenon of generative AI and the compute that goes with it comes from the companies, enterprises who are creating their own frontier models. Of the six, two, OpenAI and Anthropic, are ones that are what I consider cash-flow poor, but they are valuation-rich because they have great technology.
And we're creating very good products for these guys to enable them to continue to improve their frontier model development. So what we have done, given their performance, and they have shown they have been very successful so far, particularly Anthropic, even OpenAI, we are creating a platform by bringing in strong financial partners, not your traditional finance partners, but partners who are willing to look at these companies and, in a way, like the credit and the potential these guys have, willing to extend them credit, which is enabling that, creating a platform.
And companies like Apollo, Blackstone, who love this sort of opportunity, step in because the rates are better than going to Treasury, for sure.
And they create some confidence, even if, say, OpenAI goes out and says, listen, we're not going to do an IPO. Those two companies, which by the way I know and are very hard-nosed, they are willing to lend.
Absolutely, they are. We're finding out that it's not difficult at all to find this kind of financial partners willing to come in and fund the infrastructure of these guys, even as the demand from these two guys is just huge over the next few years.
Perfect. Now let's talk about that last quarter. I cared a lot and we said it to club members. Look what he's saying — Hock is saying '27 and '28 are going to be amazing. He's always been money good on his projections. If that's the case, I think you could earn as much as $20 a share in 2028.
I know you said in your Munich talk, we'll make judgments on allocations and what to do with capital in December, but if you know that you're going to do that, tell people why it isn't a good idea to be really big right now buying.
Well, the way we have been conservative in judging the number for '28, we're putting out there that just on AI revenue alone, which is the custom AI accelerators we are providing for six customers, the networking that goes hand in hand — those are, we are a chip company, we sell those chips that go into their infrastructure — we believe we will hit $230 billion alone for those products in '28. That plus our existing non-AI business in software semiconductors, I indicated will exceed $30 a share in 2028.
So if that's the case, I mean, I'd be buying. You know, I wouldn't wait till December. I'd be buying a huge amount of stock in the open market.
That's exactly my view, but we will earn our way into a valuation, Jim.
Okay. Because I think sometimes people don't look at the numbers, they just make judgments. The judgment should be based on the numbers. There's something you taught me in many years that I've read your conference calls. Now, you were pretty tough in saying that Nvidia's product — which we know Nvidia is a great company —
Your Jericho chip you think is very competitive versus Grace Blackwell when it comes to inference. Can you tell people — bold supposition? Everyone feels that Nvidia is untouchable.
No, here's the interesting part of the whole thing, technology. If you were to co-design silicon, XPU we call it, or custom accelerator — every, the layout of the transistor — you co-design the chip with a frontier model developer, the algorithm guys, the data scientist who tells you the kind of workloads they need, architect it, and you create a chip that matches the kind of workloads they need, your chip will perform anytime much better than a general-purpose GPU. That's the underlying premise behind it all.
Excellent. Now, let's talk about what happened this weekend. We have Dario Amodei. He writes a thoughtful note, but at the same time, some could regard it as self-serving. If you get everybody to stop and you're in the lead, like a Formula 1 race, if you have the flag that says slow down, then you're going to win no matter what. Do you think there were some cogent points made, or is it cogent versus self-serving, or both?
Well, I'll address the cogent points. There are a couple of cogent points that are very real. One, very importantly, is AI, generative AI. The creation of those frontier models, and which is what generative AI is embodied in, will create huge value to us, to enterprises, to people, to consumers, because it makes us more intelligent, more productive. So there is a lot — it's like back to the Industrial Revolution. This is a phenomenon that is very analogous to having access to intelligence to be able to do our daily tasks, whether you're enterprise or consumer, with that tool.
But it is a tool. We use that tool, people use that tool, and like any tool, it's important to put governance, safeguards on how we use the tool. But it is still, at the end of the day, a tool that will make our society, humanity reach a better level of living, but still a tool. It's not a live animal that will run well by itself, not without us managing and controlling. So the governance is important.
We're not all going to die by 2030. That doesn't seem to be in sync with that scenario you just laid out.
No, it's a great tool. That's what it is. That's all it is.
Well, you know, look, I knew you'd make me — look, you're a rational person. Do you feel that some of the comments made in Dario's notes, or these people who do these frontier models, maybe they're a little too charged, a little too emotional? I don't know the reason, but I do know it's important.
And it's important though, and that I agree with Dario on, that we have safeguards put in place. And it's safeguards not just for the user, but safeguards for the companies like Broadcom and anyone else that chose us to deploy, as we have in our infrastructure, that we put in place.
We don't need a slowdown to do that.
No, I think the way I see it is, look, I don't know about training, but when you want to productize inference, I see it continuing to be very, very strong.
Let's leave it at that. That is Hock Tan, president and CEO of Broadcom, and the Charitable Trust has a position. Mad Money is back after the break. Hock, thank you so much.
Coming up, with leading AI labs calling for a slowdown in the technology's development, Cramer's investigating how or if it could work, with George Kurtz next.
Over the weekend, the co-founder and CEO of Anthropic published an essay where he pushed for a slowdown among the frontier AI labs in order to handle newfound security concerns. In response to that, the cybersecurity cohort just went crazy, because they are the companies that can prevent AI agents from randomly hacking into networks around the world.
Take CrowdStrike, which shot up nearly 14%, leading the S&P 500, something I feel great about, of course, because we own it big for the Charitable Trust. These guys have been adamant that AI represented a great opportunity for their business. And now Wall Street's finally gotten the memo. So, let's dig deeper with George Kurtz, the founder and CEO of CrowdStrike Holdings, to get a better read on the situation. Mr. Kurtz, welcome back to Mad Money.
Jim, always great to be here.
Oh, thanks, George. So, first, George, let's get right to it. Can you slow down the pace of the frontier?
Well, I think the pacing of the frontier is going to be up to the frontier. And if they want to voluntarily slow down, they want to put extra safety guardrails in, they want to update their testing, that's fantastic. What we have to remember, Jim, is the genie is out of the bottle. There are plenty of models that are already out there, both frontier as well as open-weight models, that can already be dangerous. So, it's incumbent on the security industry to be able to help protect at least the models that are out there while the frontier models determine what pace they're actually going to evolve at.
Now, I felt when you had your Falcon conference that you actually have prevention remedies. You actually can stop this stuff. How much time did Dario Amodei and Sam Altman take talking with you? You are the leader of the cybersecurity field.
Well, they're both partners and customers of ours, Jim, and they were both at our Falcon event, and we are working closely with them. And I think what's important to realize is there's two elements to this, even beyond just CrowdStrike, but security in general, right? There is what the frontier models are doing, which is basically training in guardrails, right, putting guardrails around what these models can do. That happens in the post-training phase and essentially is training out all the bad behavior that could happen.
Well, when a model gets released, these things can deviate and potentially you can work around the guardrails. So what the security industry is focused on is runtime security, and this is what we do, right, which is when these agents are running, the agents are nothing more than software, they're programs that run and they call other programs. Unlike traditional software, they're not deterministic, meaning they can randomly infer something and take a different path.
So similar to what we've done in the human world, where a human can take random events, and we built a company around that, we can look at what these programs do. We can put our own guardrails around them at runtime. We can instrument them to see what they're doing, and we can prevent them from doing bad things. So it's a combination of what the frontier models want to do from a safety perspective and what the security companies like CrowdStrike can do to prevent these agents from going rogue.
But is it not true, George, that unless you hire, say, CrowdStrike or maybe Palo Alto, you could go out of business if something happens? I mean, these guys are obviously — what you're saying is that the agents are far more sophisticated than the so-called older bad guys. These agents are more capable. If you don't hire CrowdStrike, if you don't hire Palo Alto, aren't you just legally — I don't know, I don't want to say deserving of what's going to happen, but you know, you have to take insurance, and the insurance is cybersecurity.
Well, I think, Jim, what you're really pointing out is a bit of a church and state, right? So, if you're creating the model, you want a different company to be able to protect you from those agents and what those agents are doing. And one of the things that I really want to double-click on with you is that the security industry is moving from purely security, which has been keep the bad guys out for as long as I've been in the industry, 30-plus years, keep the bad guys out, into safety, which is how do we allow agents to operate effectively.
It sounds the same. It may sound the same, but it's very nuanced. And the safety is really a higher calling in security for the security industry. And I think that's where we all can play a valuable role to take advantage of this powerful technology for the good of humanity.
Well, George, we heard from a former one of these people who works at one of these companies talking about how we're all going to die by 2030. I know that it's entirely possible that we could all be extorted by 2030. I mean, is it possible that these machines that are software are able to just create, say, a virus or create some sort of, let's say, atomic weapon, weapon of mass destruction, that would be where it would be, I don't know, directed by some sort of robot to kill us?
Look, Jim, I'm not a doomerism type guy. I'm the glass is half full, and I think there are plenty of technologies, specifically around cybersecurity, that can help these agents from going rogue. And you know, when someone's trying to come up with these predictions of 10% and 2030, if that was so accurate, you know, they should be working on playing the stock market. I mean, I don't know how you can come up with these sort of predictions.
What I do know is that the agents are dangerous. The agents need to be controlled. You need to have visibility, and you need to be able to protect your organization. All that is true, but at the end of the day, what the agents are doing, they've only been trained to do, which is what companies like CrowdStrike have been focused on stopping, right? All the different breach indicators, the indicators of attack, how these attack chains work, those are all known. They just happen at superhuman speed with agents and an agent swarm.
So the key, Jim, is that you need equivalent or better AI defenses to combat the AI agents, and that's at the end of the day what we're talking about. You can use this technology safely. You just need the right AI defenses to be able to allow you to harness the power of AI.
Okay. Well, where are we in terms of thinking about what the Chinese may be doing versus us? If we slow down, China goes faster. Does that matter?
Well, I do think it matters. I mean, if you look at sort of the geopolitical landscape, right, from my perspective, one man's view, like we want to continue to lead that we have. And to be fair, I think it's a narrow lead. If you look at the open-weight models, many of them are Chinese. They're incredible models, incredible capabilities, and we always want to be on the frontier. And this is what I think made this country great, is being at the edge of technology.
So, you know, I don't think anyone is calling for the halt of developing AI, but when we think about how we put this together and how we sort of train it and put guardrails around it, we still want to be able to use the technology and keep our competitive advantage. And again, I think what I've said, you know, I put out a post last night, is that CrowdStrike is ready to come to the table, take what we know about these attacks and work with these labs and be part of the solution. We're happy to...
do that. And I think greater safety in the lab and greater safety in production in runtime is ultimately the best course of action.
Okay. Now, you do race cars as a as a sideline. I was thinking of you when I read this piece by by Dario because I know if you're in the lead and they do they wave the yellow flag, no one's allowed to pass anybody. And I was thinking that might be a cool thing to do if I'm in the lead and know I'm going to win and not wave the yellow flag. No one's ever going to catch. I mean, yeah, that had to be thought by. You had to think that that's a possibility.
Well, look, people are going to have a lot of views and and you've seen and heard all the views that are out there. Is it regulatory capture? Is it a way to slow things down? You know, I I'm not sure. I mean, I guess history will ultimately judge that. But I think what's important is if we put too much regulation around this, then it's going to stifle innovation in my opinion. Um I think if anyone voluntarily wants to come up with their own uh best practice, wants to be the exemplar in this uh community of of how it works and the safety around creating agents and creating models, fantastic. I think then if you look at best practices, the the market should be able to determine, hey, I want the the safest agents, I want the best models, etc. And I think ultimately that that's going to get a a better outcome than just hoping that there's a lot of regulation around this. Um, I'm not sure that's the the exact solution, but you know, we'll let others kind of figure out what that looks like as it unfolds.
All right. Then last question. Any of the companies out there that are listening to you right now, would you be able to say to them, as I think like you said at Falcon, what we have at CrowdStrike right now will prevent the kind of of misalignment that everyone seems to be so worried about. Our product can do that. Is that true, George?
It's true because once you train the model, Jim, and once you put it into production, it can drift and the uh there's plenty of ways to work around guardrails. Basically, the training of the model, particularly in post-training, trains out all the bad behaviors. But when there's a situation where the model and the harness don't know what to do, they have to infer something and they may draw a conclusion and take an action that you hadn't intended. Which is why you need the second half of the equation. Jim, this is very important: runtime protection and control. You need visibility into what the agent is doing. That's after it leaves the labs and you're using it. You need to put the guardrails around it when you're actually using it in your own environment. And you can see what the agents are doing and you can actually put guardrails around it. If you look at the Hugging Face incident as an example, I if you just look at what the agent was doing and you tracked it, now there's a lot of activity. If you looked across that entire timeline, you would see that it was actually doing bad things that it wasn't intended to. And that's what companies need to do to be able to look at and inspect at what these agents are doing so they can use them safely, which means they're harnessing the power of AI for good.
Well, there's some common sense in a debate that seems to be fraught right now. I want to thank you, George Kurtz, CEO of CrowdStrike, and everyone should read uh your your post on X. Uh it's more uh thorough than what we could cover in an interview, and I want to thank you for coming on the show tonight.
Coming up, Kramer's picking up the CEO of Lyft for a chat as Mad Money continues from San Francisco next.
All right, what are you making of the recent action in the stock of Lyft, the number two ride sharing platform in the United States? Its stock's been under a lot of pressure this year, even as the company keeps putting up solid numbers. We've seen a pretty impressive turnaround here. Yet, that's not being reflected in the stock. A year ago, Lyft acquired Free Now, beefing up their European business, and they started doing self-driving deals last week. They made Waymo available in Nashville. They got some great bike initiatives. Can the stock get back to its groove? Well, let's check in with David Risher, the CEO of Lyft, to find out what's going on. Mr. Risher, welcome back to Mad Money.
It is great to be here. Thank you, Jim.
All right. David, I I I was thinking it's so tough to continue to take on the beast. Okay. But you have you've been poking it and now you're in many cases beating it in a lot of places. And yet your stock's not flying. So let me ask you something. Are you still excited about the job?
I love my job. I literally jump out of bed every morning and that includes Saturdays when my wife is like, "David, settle down." But I'm like, "Hey, why isn't it Monday yet?" So here's why. Look, at the basic level, we are giving people a billion rides every year. That is a lot of people getting out and living their best lives, going to work, going to parties, going to concerts, going to Eagle games, whatever it is, right? So, that's a big deal. And that feels really, really good to be part of people's lives. And we're going through this incredible transition of self-driving cars that are safe, reliable, people really like them. So, how often do you get to be in an industry that is going through such an incredible transformation towards even better economics and even better service at the scale we are?
