Fast Money 9/23/25

0:02Live from the Nasdaq market site in the heart of New York City's Times Square this is fast money Here's what's on tap tonight. Data-centered dreams. Oracle just announcing plans for a big AI build-out. The details on the deal and how the stock stacks up in the AI race. And the Jimmy Kimmel fallout. The late-night host isn't coming back to all ABC stations tonight. What it means for Disney and the future of broadcast television. Plus, Mike Rohn on the move after earnings. Oil stocks get energized in dining out on GLP-1s. What the boom in weight loss drugs mean for the industry, the restaurant industry.

0:33I'm Melissa Lee coming to you live from Studio B at the NASDAQ. on the desk tonight, Tim Seymour, Dan Nathan, Guy Adami, and Julie Beal. We start off with breaking news on Oracle's latest data center plans. The stock higher after hours following a 4 % drop during the regular session. Mackenzie Cigales has got all the details. Mac. Mel, OpenAI, SoftBank, and Oracle have just unveiled five more Stargate sites stretching across Texas, New Mexico, Ohio, and then an undisclosed destination in the Midwest. We're also getting a revised spend here that brings the size of the initiative to nearly 7 gigawatts and more than$400 billion of investment over the next three years.

1:10Now, these blockbuster deals have already fueled massive stock gains, but one of Oracle's new CEOs told me that he is not worried about the market overheating. We see a broad-based demand across a huge swath of the industry. So it's not just from any one individual place that we're seeing this demand coming from. And as this kind of continues through, I don't worry about a bubble because I see committed demand for it. And also, the other thing I would say is, you know, Oracle's not just in the infrastructure business. We're also in the application business, so we're also consumers of this technology.

1:44Now, those comments echo what I heard from OpenAI CEO Sam Altman, who told me that this scale of construction is the only way to keep up with AI's explosive growth. And this is what it takes to deliver AI. AI, unlike previous technological revolutions or previous versions of the Internet, There's so much infrastructure that's required, and this is a small sample of it. Sam also telling me that OpenAI's trajectory so far only underscores the need for even more compute. With what we see on the horizon for how much people are going to want to use this and what the quality of intelligence, the level of intelligence we can deliver with more compute, we're very confident at this point we're going to need much, much more compute to deliver on that.

2:28And this comes just one day after the NVIDIA OpenAI$100 billion partnership announcement. And yes, NVIDIA and Oracle both closed lower. I'm looking at those stocks. They are both moving higher after hours. So really answering that question that, yes, this latest deal is looking like it is reigniting that AI trade as investors chase anything tied to OpenAI. Mel? How are they going to power all this, Mac? That's a question I've been asking today. And Sam Altman has long been an advocate of nuclear power. He helped a startup called Oklo go public through a SPAC. And so that is one option that they're turning toward.

3:00But Sam and Sarah Fry are both telling me that more than close to 800 partners across North America have approached them about potential facilities to host them. So here in Texas, we're talking about renewable energy. We're talking about grid power. That is not enough because Mel, a key distinction here, that's seven gigawatts we're talking about that is additive to the NVIDIA announcement. So we're talking about 17 gigawatts in total. This is an unprecedented build out that is going to require the kind of power we haven't seen the likes of in a very long time in terms of U.S. infrastructure build.

3:31Yeah, that seems like a slight stumbling block to all this. Mac, thank you. Mackenzie Cigalas in Abilene, Texas. This is fascinating. I mean, as Mackenzie had mentioned, this comes on the heels of the NVIDIA OpenAI deal. And if you want to think about the circularity of yesterday's deal, then you loop in Oracle and OpenAI and Oracle teaming up, and it's sort of this big. Takes you back to that announcement back at the White House. And that was, you know, we saw those three stocks or those that could trade, excuse me, were out there moving. And, yes, there are linkages and there's overlap in terms of the ownership.

4:06And some of this, a lot of this goes back to Larry Ellison. And look, Oracle has done an amazing job transforming their business. And the question we ask constantly is, is it profitable business? And ultimately, but when you look at where they're going to be spending on data center relative to themselves, this CapEx is bigger than any of the hyperscalers who have announced so far. So no question the right partners. No question there is demand. The question is really what's it worth? Open AI is spending a lot of, you know, there's a lot of money being circulated. One has to wonder where it's going to come from.

4:36But shelve that for a second. Let's talk about Oracle because they're being rewarded for this revenue and maybe justifiably so. But we still don't know what the earnings are going to be. And if you just do back of the envelope on what expected earnings are, I mean, you've got Oracle trading close to 40 times next year's numbers, which is historic for Oracle. And, you know, it's somewhat historic for the space in general, if you think about what's going on here now. They might earn their way into that, but that's a big leap to make from where it historically trades to where it's trading now. God, we sound a little pessimistic here because I'm just going to kind of jump on.

5:12Oh, wait a minute. So you asked the question. So it's the energy that you just asked about. It's the financing. OK, so fine. These hyperscalers, they got a lot of cash. They generate a lot of cash. They can keep doing 25, 30 percent of the revenue until things just kind of stall out a little bit. But where does a company like Oracle that is 100 billion dollars in debt and not a whole heck of a lot of cash and really is going to have negative free cash flow for the beginning here to try to build this out? And so you think about this, Oracle is actually going to need NVIDIA to come in and take a stake so they can buy their chips.

5:42Right. So when you think about the CapEx here for all of these build outs, about 50, 60 percent of it is NVIDIA GPUs. Right. And so there's obviously crazy demand for them. We haven't even talked about XAI. Right. And the demand and some of the things that they're talking about that they're building out. So who's going to finance all this? And so once you get away from the major hyperscalers, then you've got to tap debt markets and you've got to tap, you know, a whole sort of just not just public debt, but but private credit, that sort of thing. So at some point, this is going to become massively embedded into our economy if it hasn't already.

6:15I think there's some estimates that it's 50 percent of our GDP over the last year or so. So it will not take much of a hiccup to actually have reverberations throughout the economy, not just the tech economy. And this is one of the similarities. I think you can bring it back to 25 years ago. People thought it was very isolated to the Internet and the spend and the fiber that was being laid. But the moment there was not the sort of demand that the build was expecting, then you had a massive slowdown across the economy. Take that forward into 07, into 08. And I don't mean to draw too many lines together here, but the financialization of a whole host of things, the housing market, that led to this.

6:54So when you think about the interconnectivity, you think about the scale globally, but also all of these companies that are, listen, Oracle, they are attached at the hip with OpenAI. Microsoft to some degree is. So there's just a lot of interconnectivity here that I don't think the markets are pricing whatsoever. Although Microsoft's ahead of them in line, right? I mean, in other words, Microsoft, personally, has the first call on some of that capacity. And ultimately, this gets back to where who is benefiting now and where is there really not only a revenue accretion, but where there's actually profitability that changes in the face of AI.

7:27A lot of this is CapEx with the built and the promise on tomorrow. And it gets back to that trade, right? And we've clearly seen the rotation on the build and the promise for tomorrow over the last three weeks. This has not been about the early players. The first round of this was clearly meta, then Microsoft, and to some extent some of the other hyperscalers. Right. I mean, but to speak to the rotation that we've seen here, I mean, look at Microsoft. Microsoft since July. It's only up single digits here. I mean, like, Julie, where do you stand on some of the valuations being afforded to the likes of an Oracle, let's say?

7:59The desk here brings up good points in terms of growing into the revenue estimates and growing into that valuation. I think what's really, I agree with Dan, I think the difference with Oracle is now we're talking about debt. And you know that when you're in any kind of a CapEx cycle, once you've moved away and you're into debt, that starts to really change the economics and that starts to also really ramp the risk. And so clearly when you see an NVIDIA choosing to take this partnership link with OpenAI in order to juice up the demand, that tells you that they recognize that they're going to have to stimulate their own demand.

8:34And that's not necessarily a great sign in terms of the sustainability of the growth. And I think that that really reflects this interconnected nature of all of these companies. But what it really, really points to is that we have no clarity on what is the revenue potential of all of this investment. It is still completely an opaque concept that we have no real understanding of. Yes, enterprises are moving some of their workloads. They're experimenting with AI to see how it can improve their productivity. But they're not really seeing a lot of the fruits of that labor or investment. And at one point, do they start to look back and say, you know what, never mind.

9:10We'll just wait for this technology to mature. For more reaction, let's bring in DA Davidson, Managing Director and Director of Research, Gil Loria. Gil, great to have you with us. Thanks for having me. It seems like every partnership gets applauded by the markets. I mean, NVIDIA basically hands over a check or chips to open AI and that gets rewarded. Oracle gets rewarded. At what point should we actually scrutinize what this means for each company? Right now. And you've had a great conversation so far. Let me put it together. There's real companies here, Microsoft, Amazon, Google. They have real customers.

9:48Their cost of capital is probably 5 % since they have cash on hand and very low borrowing costs. They're probably getting 20 % return on that. That's great. They have all the customers. They have very low cost of capital. When we start talking about these incremental players, Oracle, CoreWeave, et cetera, we're talking about companies that have to borrow at 12%, maybe 10%, and then are getting maybe mid-single-digit returns. So that's like buying treasuries on margin. That is value destructive, and that's where we have to focus. And what's happened is we started with a core AI trade, which was NVIDIA and Microsoft, who have done well and captured most of the value.

10:30and the trade then expanded to these marginal players that are now trading at higher multiples. In what world should Oracle trade at a higher multiple than Microsoft when their economics are flipped? Microsoft is creating value. Oracle is going to be destroying value. In what world should Oracle trade at a higher multiple than NVIDIA? NVIDIA is going to win anyway. Everybody's going to buy NVIDIA chips. It doesn't matter who it is. Yes, they're overstimulating the market by making the deals they've made the last couple of days. But even without that, they still have the organic demand from Microsoft, Amazon, Google, Meta, Elon, et cetera, China, when they're able to sell there.

11:11It doesn't make any sense for these marginal players to trade at a higher multiple than the real companies building AI. I'm curious then, when you get inbounds about Oracle, let's say, and Oracle increases their capbacks, they have debt, but at every step of the way, the stock goes higher. What do you tell your clients and what is their reaction? Do they just not want to hear it? Do they just think, you know what, this is a trade that's in front of us and we understand that it doesn't intellectually make sense, but it is a trade. Well, I point them to two things. One is if you go back to the transcripts from Oracle for the last few quarters, you'll see that it's not just the last deal from OpenAI that increased their backlog.

11:53It's actually been several quarters where it's really OpenAI that's been driving all of this. And then I point them to the fact that OpenAI, I think you pointed out earlier in your conversation, Microsoft still has right of first refusal on all their compute. And then they committed another 25 to CoreWeave, another 100 billion here. They're committing to building Stargate in the UK, Stargate in the Gulf. They're designing their own chips. They're building their own hardware. And by the way, they're going to lose more than$10 billion this year. Having that is the only thing that's added value to Oracle is very risky.

12:28That's not a customer that can pay all their obligations. They're double, triple booking, maybe quadruple booking capacity. They will not be able to live to those obligations. So if you're adding$400 billion of market cap to Oracle based on that, I think we should revisit the math. So, Gil, what are the capital markets implications here? I mean, Oracle's up two and a half times from April. I mean, should they try to raise some equity here? Because you talked about a reliance on capital sources that just may not have the capital there. Would you be going to the capital markets? Will any of these folks be going to the capital markets?

13:02And I mean the public ones. Well, the amount of capital they need to raise is very substantial. They may and they may. Companies like Oracle and Cori may need to go to the market for equity. But as you were pointing out earlier, what they really need to do is raise tens or hundreds of billions of dollars worth of debt. They're quickly going to take over the entire below investment grade category, which is an expensive category. And again, their returns are in the single digits. So if they're borrowing high single digits and getting returns in the single digits, that's value destructive. This really should continue to move forward through the hyperscalers that have all the customers that can sell them other stuff, which is why they can get higher margins, which is why they can get returns.

13:47And again, their cost of capital is negligible. Gil, the multiplier effect for the stocks is clear, but the return on investment is something that people have been writing about. MIT just had a piece out about ROI for all the spend in AI. I mean, when is people going to start to critically look at that? And if there is a downturn, start to sort of walk away from it. I think we're all very well aware of the ROI. So, again, that's why we focus on NVIDIA is getting great ROI. Microsoft, Amazon, Google getting great ROI. Very few other companies are. But what we also know is that these tools are great.

14:23You get people all the time telling you about how they're using the AI tools. We use them ourselves. We know we're going to bring them to work. We're going to bring these tools to work. The MIT study was about companies struggling to implement AI. It's because it's hard. It'll take them a few years to do that. The upstarts will take a few years to mature. But it's not going to take that long for us to use AI because we're going to bring AI to work. Just like we brought the iPhones to work. It's not the IT department that gave us iPhones and moved us to the age of mobile. We brought iPhones to work and that made us more productive.

14:58It's going to be the same with AI. And again, the tools are getting so good that we do need this. The need for AI, the tools around AI are great. I'm actually here at a conference for Verkata, a video monitoring company. They're implementing AI. All of a sudden, you have automatic thread detection and translation, and the AI can even warn people to leave. We're going to see this everywhere. It'll take time. In the meantime, our usage of AI is what's going to drive the need for inference, which is real. Gil, great to speak with you. We haven't gotten to the power issue, but we'll do that next time.

15:37Gil Luria of DA Davidson. One of the bullet points, one of the points that Gil was making also is that he is worried that NVIDIA is becoming the investor of last resort to the AI world with the deal that it made yesterday. And so I'm wondering, you know, how you view that in light of all the billions of dollars swishing around here. Well, it's been going on now for a while. I mean, not that there's anything illegal about it, but vendor financing is something we talk about. And when, you know, you invest in the companies that are subsequently buying your goods, I mean, that's concerning. It's not problematic.

16:12And I think right now the market's clearly giving them a pass. But a couple months from now, it might have a much different picture. Well, think about this. So OpenAI, that's the one. I mean, first of all, let's be really clear. Oracle is going to be the patsy in this whole thing. So when we look back in a couple of years, it's going to be on Oracle's footsteps here. And I feel bad for them right now because they're literally chasing this thing in a big, big way. But look at OpenAI. It's being valued at a half a trillion dollars. That's half of what Oracle is. The company has already told you that they are basically going to lose tens of billions of dollars for the next, I don't know, five or so years.

16:42So they're going to continue to actually fund this thing in any way, shape or form. And it's probably going to be, I think, I mean, it's going to be a problem for some of their biggest customers eventually. But then OpenAI, they have 700, 800 million, you know, weekly active users. They have 12, 13 billion dollars in revenue. I mean, as long as the private markets, whether it's equity, whether it's debt, are going to continue to fund that. That's not a bad thing for them because there's 8 billion people on the planet. And let's just say in five years, half of them are going to be using OpenAI technology.

17:12So at the end of the day, I just think there's going to be obviously winners and losers. The infrastructure guys who are there, the hyperscales that you guys just mentioned, are going to be able to get some return on this investment because they are going to be the compute for the future. Fine, but it's not going to trade at the multiple, I think, that is being afforded to them right now. And just back from a market perspective, I mean, think of the headlines we've had really since, well, NVIDIA reported, but including that report. But really think about the NVIDIA-based headlines that have been related to these types of partnerships.

17:41NVIDIA has underperformed the semiconductor space. NVIDIA has actually been flatlining from, you know, kind of early August. I'm not saying, in fact, I think NVIDIA, of all the names we've just discussed, I would agree that the investment thesis around NVIDIA is the best. The valuation is the best for the growth. But in terms of what the market is doing with all these headlines, they've helped everybody else but NVIDIA. Yeah, but before we leave this conversation, I do want to get back to Microsoft just briefly because of the noted underperformance relative to the other MAG7 stocks. And in this game, Gil mentioned it, Oracle and Microsoft's economics are flipped.

18:14So why is Microsoft just sort of, you know, stuck in the mud and Oracle's soaring? I think part of it's a valuation thing as well. I mean, Microsoft was reasonable at a certain point and got, I think, expensive, again, historically expensive and clearly expensive to the broader market when it had that huge run up a couple months ago. I think right now it's just a valuation reset. Meanwhile, major indices retreating from record highs today as Fed Chief Jerome Powell raised concerns over market valuations at a special luncheon in Rhode Island. CNBC's Eamon Javers has got the highlights here. Hi, Eamon.

18:46Hey there, Melissa. Well, a lot of analysts suggested that Jay Powell didn't say anything today that he hasn't really said before. But in the Q &A part of the conversation, Powell laid out his view of what's happening in the labor market, and he called it a low-fire, low-hire environment. Here's how he explained it. We all see the data that it's just gotten tough for people just entering the labor force to be hired. But remember, the overall national aggregate hiring rate is at very, very low level. The layoff rate is also at a very, very low level. You're in a low-fire, low-hire economy. And it feels like companies are just, they've kind of stopped hiring or slowed down their hiring because they want to see how this all shakes out.

19:31He also talked a little bit obliquely about all the political pressure that's been on him as the Trump administration and President Trump himself has been bashing Powell and demanding that he lower interest rates. Powell not specifically referencing President Trump here, suggesting that in the post-COVID era, post-financial crisis, institutions broadly have lost a lot of confidence. Here's how he described the moment. often things are seen through a through a lens of is it good for this party or bad for or bad for this party or this politician we're just not looking at things that way we're looking at what's the best thing for the people that we serve in the medium term what's the best policy and no one many people don't believe us because they think yeah come on oh god you're really political but truth is mostly people who are calling us political it's just a cheap shot so powell they're not saying that Trump is guilty of a cheap shot here, but Trump is the person who's calling him political, right?

20:28So you can sort of read between the lines there. Meanwhile, the president still continues to call Powell too late, Powell. And Treasury Secretary Besson told a few of us yesterday that he is continuing to interview candidates for Powell's job. Up to 11 candidates now, he said, by the end of next week, he'll have interviewed 10 of them. No word on who's in the lead, guys. Back over to you. All right. Eamon, thank you. Eamon Javers. I thought what was also interesting out of this speech in Q &A that Powell did was that he basically said forecasters is a very tough job. No one can forecast the economy beyond three months.

21:00That's sort of the economy that we are in right now. And so that sort of underscores a predicament that the Fed is in in terms of reading the data and charting a path. Yeah, I mean, he used the term, this is a complicated situation. And again, referenced inflation, also referenced markets that were fairly highly priced, which I think is fascinating because they are fairly highly priced. But there was I don't think there was anything that should make you should make an investor miss or interpret last week's Fed meeting any differently. But I do think some of these comments point out that the Fed is not run away cutting rates.

21:35Julie. Yeah, I agree. I think that, you know, he's really trying to build in enough flexibility that if we have a suddenly weaker employment report or that inflation really spikes, you know, he has the flexibility to move the way he wants to move. Chair Powell is one of the ones who's paid the most attention to how immigration is impacting the labor force and how that actually could create more structural inflation. And so I continue to pay attention the most to that segment of his comments. Coming up, high energy moves in the oil patch, the bouncing crude and the stocks coming along for the ride.

22:08But first, some after hours action with shares of Micron on the move. The numbers and details from that quarter next. Don't go anywhere. Fast Money is back in two.

22:23Welcome back to Fast Money. Let's get to an earnings alert on Micron. Shares are up more than 2 percent right now after hours of memory chip company posting better than expected revenues and earnings in this latest quarter. The stock has been on fire, to say the least, up about 40 percent in just this month. CNBC's Christina Parts Nevelis has got the latest from the earnings call. Christina. You're commenting that it's 2 percent higher, but it's actually coming down from just when the earnings came out. And maybe this is a sell the news type of event because the stock has doubled just in the last year and just climbed dramatically.

22:52And it was managed to beat EPS estimates, even though they pre-announced back in August, which is incredible because for the last 11 quarters, Micron has beat estimates. Now to the call. The most interesting thing is a high bandwidth memory, which is really important. HBM4 for the more advanced AI accelerators that are coming out next year, primarily for NVIDIA. And there was concerns that, you know, they'd be losing some share to Samsung and concerns about pricing. And so on the call, he did say this is the CEO that they were going to sell out 2026 HBM supply in coming months. So I saw that as maybe a bullish comment.