Nvidia just approved a $150 billion stock buyback. Here’s why it matters

And they come this morning courtesy of Nvidia, which is making two announcements this morning. One probably much more important for the stock and the other one more important for the AI ecosystem, of which it is an integral part. >> Potentially the state of the world. >> Yeah, for Yeah, no big deal. Um so, the first [laughter] for the stock and the reason that it is up uh this morning is that the company is doing an additional share buyback authorization of an additional $150 billion. I've seen um by some measures it is the largest buyback authorization ever announced um because it means that the total remaining authorized amount that they can buy back is $235 billion and it expects to do like a just for people who are watching like companies do these share buyback authorizations. They don't always actually buy back all of the amount that's authorized. They say they're planning to buy back $235 billion worth of stock through the end of their fiscal 2028. So, that's helping the stock. And then just quickly on the other thing that they announced, um they are announcing this new AI security system, which they are calling the open agent safety platform. Um it's in cooperation with a bunch of different companies, notably not OpenAI, but Anthropic is currently partnered with them. And it's sort of a response to all of the headlines that we've been getting lately about all the hacking that has been happening. But it definitely feels like as I was saying to you, Jake, like this is definitely a you know, self-protective measure, shall we say. >> 100%. 100%. Um just to fill in what what this is. >> Yes, please. I've been talking too much. You talk some now. >> [laughter] >> So, uh Jensen Huang, the CEO of Nvidia, went on uh CNBC this morning and I thought actually managed to give a pretty succinct description of what is a pretty technical thing. And the way he describes it, I'm just going to read his words. He says that in order to make sure that an agentic system is safe, an agentic AI is where we've seen the problems. Agentic being that it makes its own decisions. It kind of drives the car itself. You have to make sure the sandbox, the force field that's around that agent keeps the agent within the minimal rights it has and that it needs to do its job. And so it can't do anything it's supposed to do. The idea here is take the agent and kind of wall it off, build a shell around it. >> but there was already supposed to be a shell in many of these cases >> now kind of stepping on top of that. >> Yeah, an additional shell, an outer wall if you will. If the sandbox is the moat, this is the wall beyond the moat or maybe the moat this is the moat and it was already in the wall. >> is the We're getting too >> [laughter] >> Our metaphor's getting too stretched, but they're also it's important to note this is also part of this is open source. The idea here per Jensen is that other companies can kind of build on top of this. Take the code, build their own systems to adapt to their own agentic systems. He was asked actually, why is OpenAI not listed? Did they decline to participate? And he kind of sidestepped whether they declined to participate, but said look, they're welcome to take what we're giving them, to use this platform. We would love to see them take this. Sam Sam Altman at OpenAI is a really smart guy. We would love to see them push this forward. >> So you mentioned OpenAI and we talked about just these stories over the weekend about the different breaches and the hacking that they've been I mean I've been seeing a lot of people talk about are these really hacks, right? Hacking. Is it really a hack to try to access unauthorized part of a site that is still fully open, right? If it's looking for data. According to some of these instances of where the OpenAI software was looking for information, couldn't get it directly from a site, so it said how do I get into this site cuz it's blocking me, that sort of thing. I'm curious about just back to the Nvidia this sort of open shell thing that they're calling it. It's kind of like did did Jensen say that this would have stopped some of these OpenAI sort of intrusions? >> I think he said it would have helped the hugging face. >> have It would have helped or stopped the hugging hugging face. >> a good thing then, right? That's that's like sort of what you you to hear going to calm the waters a little bit. But just real quick, I want to go back to the to the to the buyback you're talking about. You know, what? 238 billion >> 235 total with this 150 addition. >> Right. And then you're talking about a $4 trillion company, you know, that's still a sizable buyback. Right. And it kind of goes back to the whole thing. What are you buying with an Nvidia, right? Just just as an investment case. When it's really talk about this. Like isn't this a growth company you want to keep investing into the company, keep buying like make new sort of chips, make new sort of GPUs? Why are you doing a buyback? >> Well, he says you can do This is what I think is interesting about his statement that was in this announcement. >> notorious buyback hater, by the way. Notorious. >> Oh, okay. Well, we'll get it. So, he says, "Our cash generation gives us the capacity to invest in the technologies that advance this transformation and return capital to shareholders." So, he's saying like, "Why not You know, as the meme says, why not both?" >> I can't we do both? Why not both? >> This is but this is sort of remind me of when Apple went from high growth to like, "We're just going to do buybacks. We're just going to do dividends." >> Right. >> No, it's been totally successful, but >> But also Nvidia's growth is still like, if you look at its revenue and earnings growth, it's still outpaces its margins. Also, our just a stock continue to be pretty strong. >> Yeah, and they're saying we don't know what to do with our money. That's what they're saying. >> Well, don't forget last week >> don't know what to do with it. >> we got Jensen saying that he believes Nvidia is the world's first growth value company. >> [laughter] >> Why not both? >> Sure. >> Why not both? >> Sure. >> No one wants to hear that the V word, okay? Nobody wants to hear that word. >> No one in Silicon Valley wants to hear that word, Ernie. >> Do not say the word that shall not be named. >> Well, if you Okay, so I'll I'll take I'll I'll argue the other side of this for just a moment. If I'm an Nvidia investor and I'm looking at the fundamentals of the company and I'm saying, "Why hasn't this thing done better, right? Like why why are we looking at What is the PE now?" The PE >> 30-ish. >> Yeah. Although, you know, there's a lot of people who say like that that is still low that it is not valued as highly as it could stock. The shares are up about 21% this year. >> Yeah. So, why not give it a little juice? Why not return I've been a patient shareholder through the ups and downs of this. Why not give me back some of the capital that I >> No, I mean that makes that makes sense to you if you want >> Why not keep your investors happy? >> to stabilize stock price so it stays above 20% for year-to-date, right? >> Sure. >> I mean that's that's fine. I mean that's But again, you're what are you what are you at? You're at near 52-week 52-week high. You're buying a stock at I mean you can't really time these things. I get it, but I'm just saying it just makes sense to >> Not saying that they're buying the stock at these levels. They're going to be doing it through >> yeah. They might find over time. You might have a little pullback and they decide, okay, we we'll we'll buy it. So, yeah, I totally understand that. Um I'm just saying that we're talking about the most valuable company in the world that's also a freaking high-growth tech company that's authorizing 230-plus billion in buybacks that just to me is mind-boggling. >> look at the also look at the backdrop, right? Like the reports that you mentioned over the weekend, there was one report from Axios um and our old friend Matt Levine who reported there could have been tens of thousands of ma like whether you call it a hack or not, a model acting under its own authority, right? Not because it's given explicit directions to go out and gather information in places it's not supposed to, right? And I think I saw something in one of the one of the articles about all of this is that like the issue with this is if you're coding if you're giving the agent instructions because the agent doesn't sorry, anthropo- anthropomorphizing the agent doesn't think like a human it does things that doesn't even occur to us to tell it not to do. So, like how do you program these things to not do to you know, to not do things that we wouldn't that wouldn't even occur to us? Do you wait until it goes out and does those things and then you say, well, wait a minute, don't do that. And in that case, like you can have this open this new Nvidia security system but if these companies are going to continue to train these models are they going to continue to want to not wall them off until they need to >> Also anything that addresses the problem afterward is coming after the harm is already been done. >> I think Julia I think you nailed it right there. >> What's that? >> They don't necessarily want to keep these things behind the wall because they want to see what they can do. >> I mean they've said they're going to pause and they're getting you know >> Or they are not giving them super explicit don't exploit this area. >> Right. >> It's like get just get the job done. And they're like we want to see what will happen if we freak out the agent right? Like Like I guess they call frustrate the agent right? >> Yeah and then it's doing these things. It's like it's like a kid right? Like you don't give them clear instructions. They're going to find a a a way to do it but also what do you mean you didn't say I could do it like this? You didn't say I could do it like that. >> Right. >> they they'll always find a little way to like just >> And also similar to a child it you know Can I do it? Can I do it? Can I do it? Can I do it? Can I do it? But like endlessly you know it will try right? These things don't run [laughter] out of patience. >> And then you catch it messing up and it's like you catch it >> They don't get cranky. Like they just do it. >> No and you catch them messing up and it's like you're right I did that. >> [laughter] >> Right exactly. >> Sorry I didn't think you noticed yet. What do you mean you didn't notice?