On this week’s edition of MI&S Datacenter Podcast, Moor Insights & Strategy co-hosts Matt, Will, and Paul analyze the week’s top datacenter and datacenter edge news. This week they talk about Stargate – is it real?
Watch the video here:
Listen to the audio here:
Stargate – Is It Real?
https://openai.com/index/announcing-the-stargate-project/
Disclaimer: This show is for information and entertainment purposes only. While we will discuss publicly traded companies on this show. The contents of this show should not be taken as investment advice.
Matt Kimball: Ah, we are live. Good morning, good afternoon, good evening, good night. Good whatever.
Paul Smith-Goodson: Good gracious.
Matt Kimball: Good gracious. Good gracious. Good all. Good morning, good afternoon, good evening. Welcome to the 36th episode, 36th episode of the Data Center Podcast for more insights and strategy. I am Matt Kimball, and I am joined by the super smart Paul Smith-Goodson, and super smart, smart Will Townsend. And we’re here to talk about all things data center. Before we get into the discussion, a couple of notices.
First is, I’m a little off my game today because I’m still kind of coming off the flu. So if I seem a little more air headed and dumb than usual, that would be why. So please forgive me. Thanks for laughing at that, validating me. And secondly, we are here to talk about cool technology trends. We are not here to give investment advice, we are not here to tell you where you shouldn’t, shouldn’t sock your money. We’re really just here to talk about tech. So please, do not take anything we say as advice or permission to go drain your 401(k), and drop it into one company or another. Unless Paul tells you that. And then, you should. No, I’m just kidding. Don’t take anything we say.
So, we have a special kind of episode, or edition today. As most of you all know, there was a big announcement from the newly sworn in President Trump and his friends around Stargate, the Stargate Project, or Project Stargate. And so we figured we’d make this a little bit of a truncated podcast, and just kind of talk through Stargate, what we know, what we don’t know, and what questions are still lingering for us. Will, Paul, does that sound right to you guys?
Will Townsend: That sounds spot on.
Paul Smith-Goodson: I mean, it’s the biggest project I think ever in terms of tech, so I think it’s worth some discussion.
Matt Kimball: Yeah, I think so too. Do you like how we made our banner scrolling down?
Will Townsend: I love it. It’s very multimedia, yeah.
Matt Kimball: Hey, I may have the flu, but my genius is still at work here. Anyway. All right. So Stargate, is it real? Paul, seeing as you’re the A guy, right? Can you talk first high level what you know about Stargate, what you don’t know, and we’ll just kind of leap in from there? Or how do you want to handle this?
Paul Smith-Goodson: Yeah, what I don’t know is probably more than what I do know, but it’s a huge monumental private sector project. And supposedly, it’s aimed at making the United States the undisputed global leader in AI. The project is funded by, the equity funders are OpenAI, Oracle, SoftBank, MGX, who’s an Abu Dhabi financial organization. And the technology being used in the project, as I understand it, is ARM and Microsoft, OpenAI, NVIDIA, Oracle. And I think that’s a lot more, but those are the main ones.
The lead partners in terms of financing is SoftBank, and OpenAI is the operations lead on the project. So, $500 billion is a lot of money. I don’t know how much they’ve got now, but I think it originally started off as a hundred-billion dollars project last year between Microsoft and OpenAI to build a supercomputer. And in the announcement things like that on TV, they said that it’s already started. And I think that that start was from the original project between Microsoft and OpenAI. It’s supposed to expand to 10 additional cities, and then maybe 10 more after that. We kind of have an idea of what the first one they’re building is in Abilene, because I think that’s the Microsoft project, but I don’t have any idea of what the other 10 will be. Do you want to add something to that?
Matt Kimball: Yeah. All right. So yeah, I think you kind of summed up everything we know, right? Which is not a lot, right? It seems like there are more unknowns than there are knowns to me. $500 billion over five years, to what you said, big names included, Oracle, SoftBank, OpenAI. NVIDIA was, as you said, Microsoft will list as partners. There are a lot of other contributors that I’d love to know who they will be contributing. But that’s, to what you said, that’s about all I know. Will, I don’t know if you had other details on this that you wanted to share.
Will Townsend: No, I mean, that’s my understanding as well. But what’s interesting is that, I think Paul was talking about this, Microsoft late last year, earlier this year before this big announcement, announced its $80 billion investment in building out AI data centers, right? And we talked about that on a prior podcast. But when you look at sort of the original partners, one observation is, the big hyperscalers are not listed, number one, which is sort of interesting. But Oracle’s in there. And I had a question for you, Matt, and the detail may not be there, but is this Oracle cloud infrastructure, is this all the traditional Oracle database and storage stuff that’s going on? What’s your understanding on the Oracle? Because when you look at the stock prices of all these companies that are participating that are public companies, they all saw pretty big boosts when this news broke on Tuesday.
Matt Kimball: Go figure, right? So, Oracle’s interesting, because they have OCI, they have a life sciences division as well. I was actually at a startup that got acquired by Oracle, full disclosure, that did clinical trial management and safety reporting. And they have the traditional database, which kind of feeds into all of this. They’re a data platform at the end of the day. So they have multiple, multiple horses in this game. And one of the things that I noticed to what you said is, and it’s a question that comes to mind is this. At the end of five years, let’s say, which it’s not going to be five years, but at the end of this initial phase, you have 11 data centers, right? And if they all fall under what they’re doing in Abilene, each one’s 500,000 square feet. A whole lot of power going in. These are powered by Abilene is the A 100, but I’m sure they’ll be advanced as they move on.
So you’re effectively building a cloud, an AI cloud. So, does this become a competitor to AWS and OCI? Or does it compete more with the core weaves, and the Lambda labs of the world, and the vultures of the world? This is the stuff that it would be really, Paul and I were talking before you came on, well, this is the stuff that would be really interesting to know upfront. What does the end state look like? If you are Sam Altman and Larry Ellison, and we’ll use them because they’re more the front end of what this is, not the SoftBank side necessarily, but if you’re those two, what does success look like? And what does a finished product or service look like? That’s the part that has not been articulated by anybody. Paul, I don’t know if you had thoughts on that, but that’s where I get a little bit… Turned around is probably the best word. Confused, certainly.
Paul Smith-Goodson: What’s confusing me is originally announced back in March of ’24, that basically the announcement was this initial project in Abilene, which is probably the supercomputer. And then that announcement in March also talked about expanding it to other cities and building other parts of this. So it sounds like the whole plan was laid out back in March, but now we’ve got these other people coming in. So that’s what’s confusing to me.
Matt Kimball: Well, it feels like this is, it felt like what was being done in March, you could draw a parallel, or draw a comparison between that and what Elon Musk is doing with this data center here in Austin, right? Super big, supercomputer designed and optimized and tillered specifically for AI, right? Data center level supercomputer. You could see that. But if this is a public/private funded venture, but it’s going to be its own company, who are you serving? I think Larry Ellison mentioned Sloan Kettering and Stanford Medical as types of customers that could be using this instead of doing on-prem. Is that who you’re servicing? Or is this for everybody? What exactly is it? And by the way, I can see where this is really interesting to someone like my dad who’s a retired firefighter, I mentioned this to Paul too, to my dad who’s a retired firefighter and doesn’t spend his day thinking about tech, right? You hear $500 billion, you hear AI, you hear global leadership, and you go, “Good, good, good.”
Will Townsend: And you hear a hundred thousand jobs.
Matt Kimball: New jobs, high paying jobs. Yeah. And it’s like great, great, great, right? But we already outspend, when you look at investment in AI projects, the US already outspends the rest of the world combined. I think we’re at 70 billion from a government flowing into projects. And the rest of the world is like 30 billion, with China being 20 of that 30 billion. So we’re already greatly outspending them. But yeah, I can’t tie all this together in my head.
Paul Smith-Goodson: Yeah. Well, the question is also, is it worth $500 billion to do this? And there’s one instance that it would be worth $500 billion alone, is if you were building AGI, and one of our global adversaries at AGI and was able to crack military plans with AGI, then it would certainly be worth $500 billion to protect that. And China obviously is on the path for AGI. They’re not quite where we are, but that’s an issue.
Matt Kimball: No, but you know what? I will say this, Paul. When the discussion a couple of weeks ago around Biden signing that final executive order around restricting the sale of chips, sale of chips into outside of those, what was it, 18 countries that could sell easily? I started digging into that a little bit, because China is obviously the target, right?
Paul Smith-Goodson: Yeah.
Matt Kimball: Kind of giving some credence to this a little bit, it’s always about the money, it is about the dominance. And when you look outside of the US, China has done a very strong job of propagating AI foundational models and frameworks around the world. What’s it called? Ali has its own foundational model that apparently is phenomenal at multilingual support, and a couple of other functions, that has really taken hold in South America and some of the emerging economies, right? So I can see, seeing that as a country and go, “Wait a minute, we need to do something.” So to that side, I can see, kind of tying back to what you said, Paul, maybe that’s a little bit of the motivation that goes behind this, I’m not sure. What were you going to say, Will? Sorry.
Will Townsend: Yeah, I was just going to say, and sorry, the latency is kind of poor on my end, so you guys are freezing a little bit. But this is sort of analogous to the Silk Road, right? I mean, a lot of what Chinese infrastructure providers like Huawei and ZTE have done to over invest, and less economically advantaged regions like the African continent and that sort of thing, to establish a foothold, right? And so it almost seems like China is doing with language models and that sort of thing is, there’s a similar parallel to that, right? Would you agree, Matt?
Matt Kimball: Yeah.
Will Townsend: And then just sort of some additional thoughts on this. Certainly this is all linked to Trump wanting to reverse some of the policies that the Biden administration made, not only on limiting the exports of AI technology, but also when you look at the guardrails that the Biden administration were trying to put in place around safety and security, and model validation and requiring companies to allow the federal government to validate this, which is just insane, because they’re the last people on earth that understand technology. But this is a pretty complicated and political football, right?
Matt Kimball: Yeah. Were you going to say something, Paul? Sorry.
Paul Smith-Goodson: No, I was just going to say, I said it started off being a hundred-billion dollar private project, and then I think President Trump saw the advantage of making it larger and more powerful. And that’s how we got here.
Matt Kimball: Yeah. You’re absolutely right, Will, you hit the nail on the head a little bit to a degree too, which is there’s a geopolitical aspect to this that can’t be discounted, right?
Will Townsend: Yeah.
Matt Kimball: If nothing else, this is kind of a big statement that Trump is trying to make as he comes into office. And from an economic, a hundred thousand well-paying jobs over the next five years.
Will Townsend: Everyone loves that.
Matt Kimball: Yeah, everyone loves that, right? What I’m kind of curious about is, Sam Altman’s up there, Larry Ellison is up there and doing Larry things, right?
Will Townsend: Larry, he’s reversed the aging process, man, let me tell you. How old is he?
Matt Kimball: I don’t know, man. It’s, I think he’s got some crazy rituals going on, Voodoo or something. I don’t understand how he’s doing that. Yeah, but where was Elon Musk in all of this? He came on pretty quickly after.
Paul Smith-Goodson: He’s a fallout between Musk and Altman.
Matt Kimball: Yeah. Well, he let it be known. Does him heading up Doge kind of have any impact on all of this?
Will Townsend: Certainly could.
Matt Kimball: Yeah. So let me ask you guys this. Do you see what they are doing as protectionist, or national protection, kind of?
Paul Smith-Goodson: Let me ask you, what is Sam Altman’s number one goal? Three letters: AGI.
Matt Kimball: Oh, yeah, yeah, yeah. Right. Artificial…
Paul Smith-Goodson: General intelligence, yeah.
Matt Kimball: Artificial general intelligence, yeah.
Paul Smith-Goodson: That’s his number one goal.
Matt Kimball: Yeah. But you can see what his vested interest in all of this is, right?
Paul Smith-Goodson: Yeah.
Matt Kimball: I still don’t necessarily see Larry’s full vested interest in this, other than maybe this gives him an opportunity to carve out more opportunity for gold in the end.
Paul Smith-Goodson: Data feeds AI. I mean, you can’t live without it.
Matt Kimball: Sure, data feeds AI, but I look at what is going on and I think, does this cannibalize existing business lines that he might have? Does this cannibalize OCI? Does this cannibalize Oracle database? Because a lot of the data that’s going to feed AI might originate in a database, but it’s not sitting there forever, when you kind of look at these. Where exactly is he going, what angle on all of this? And the last thing is, when do you bring in the broader ecosystem? It doesn’t seem, I don’t know, it felt like there needed to be more people than just, more companies than just Oracle, OpenAI and SoftBank, with mention of Microsoft and NVIDIA in passing. Just seems like you need a bigger consortium, if you will, or team.
Will Townsend: Yeah, I agree. To answer the question, is it defensive? Is it not? When you look at where AI is headed, and I’ll just call it modern AI, I mean, Paul knows better than anyone, AI and machine learning have been around for decades. But when we look at modern AI, we look at gen AI, we look at gen tech and AGI and all of that. Clearly the United States is the leader. We are driving that ecosystem. I spent time with Cisco recently in Palo Alto, they hosted an AI summit. They brought in a who’s who of companies mainly from Silicon Valley. And I was super impressed just with the depth and breadth of what US companies are doing to drive it. And when you hear there were lots of concerns in the EU when the Biden administration suggested these export restrictions, because they wanted unfettered access to US technology, right?
So, I think Trump sees this as low-hanging fruit, really jumpstart. I mean, this announcement came out on the second day. At the same time he cleared a lot of cybersecurity advisory boards within the Department of Homeland Security. So we’ve all expected this, right? I mean, he’s bombastic, he’s going to come in, he’s going to reset things and that sort of thing. And so this is just part of a greater agenda. He sees the GDP impact that modern AI is going to have on this economy and how it can catapult the country to the next level. And if there’s any notion of a recession, completely shake that off.
So, I think he sees an opportunity for the government to help facilitate this. It’ll be interesting to see if the federal government puts funding behind this, because they have done that for connectivity for underserved areas through bead funding and that sort of thing. I wouldn’t be surprised if we see the federal government kick in dollars in that there’s a lot of research that might be focused on this. To answer your question, Matt, does this compete with the hyperscalers and the services that they provide? Potentially, but if the federal government gets involved then, it could be sort of a research test bed as well. So I don’t know, I mean, there’s so many things that are just ambiguous at this point. We’re all just sort of spit balling. That’s sort of my take on it.
Paul Smith-Goodson: Yeah, if we really understood what the project was and what’s going into it, then we could probably come up with a lot better answer too.
Matt Kimball: Yeah.
Will Townsend: But it’s fun to kick the tires on this, and speculate, and try to demonstrate how smart the three of us are.
Matt Kimball: Well, yeah, I mean, I will say this, what we said earlier. I think for folks that aren’t living in tech every day, this sounds incredible. And it is, by the way. But it sounds like this is going to change the world. I think those that are in tech have more questions than answers, and we’re all looking for those questions to be answered before weighing in. And I know, you know, Will, you mentioned, government could be dipping in. To some degree, I hope they don’t. I mean, I don’t trust the government in any way with regards to, I think they cycle innovation more than they… Maybe the breeding ground of some incredible technology.
Will Townsend: I agree. And I don’t point to be as an example of progress, because that program was fraught with lots of issues, and a lot of gaming, and that sort of thing. And I agree with you, I wouldn’t necessarily advocate that the government gets involved financially, but I mean, you never know. It’ll just be interesting to see how this all winds out over time.
Matt Kimball: Yeah, yeah. All right. So, this has been a different format because we’re kind of just hitting one topic, and we’re going to try and be a little more concise with this. We’re coming up on about 25 minutes. Closing thoughts, or predictions you might have around this in the next six months to a year? What are your thoughts? How does this evolve? And by the way, if anybody out there watching us has ever been to Abilene, trust me…
Will Townsend: Yeah, I’ve driven through Abilene, there’s not much out there. I’ll jump in, Matt. Anshel and I, we recorded, we actually live-streamed our 5G podcast, the GT on 5G, available on all podcast streaming channels. Kirtu had his plug for that. One of the topics we discussed was how 5G could play a pivotal role within Stargate. I love the name, by the way. I’m a science fiction geek, and Stargate was a great TV show. But know 5G, and I’ll just leave it, leave this as a thought for our viewers and listeners, 5G has been searching for its killer app, right? And you could argue that fixed wireless access consumer services have become that killer app, an alternative to fiber in areas that are underserved by fiber.
But what we’re beginning to see is, as AI moves hybrid, moves from large language models that are processed in the cloud to network edges where there are smaller language models and you’ve got devices and AI-enabled PCs to run these smaller language models, 5G mobility, given its latency profile, its throughput profile, and its massive device support profile, over 4G LTE, could really become part of the connectivity tissue. Because we haven’t really talked about this, but you need networking to move all this freaking data around, right? So that’s a prediction that I have, Matt, but I think there’s too much ambiguity to draw any other conclusions, at least from my perspective. But Paul, what do you think?
Paul Smith-Goodson: Well, first of all, I think it’s a good project. I mean, good for the United States, good for AI. And I think even if it turns out to be a $250 billion project, it’s still good. I think it’ll move AI a long way for us. I think they should have named it Anubis Gate instead of Stargate, written by one of my friends, Tim Powers. But we’ll go with Stargate. And I think when we get a little further along, I think they’ll flush it out and they’ll release more details, and I think we’ll be satisfied with what we see. There’s some challenges, I mean, one of the issues is, you know, there’s already a shortage of good AI people.
Will Townsend: Oh, good point.
Paul Smith-Goodson: Maybe this is a plot to use H-1B to suck off all of Chinese AI workers into the US.
Matt Kimball: A hundred thousand, a hundred thousand in five years.
Will Townsend: That would make a dent, right?
Matt Kimball: It’s a brain drain, right? Get a brain drain.
Paul Smith-Goodson: Yeah. I’m glad to see it.
Will Townsend: Matt, what are your thoughts?
Matt Kimball: So first on the name. So, you like the name. When I heard Stargate initially, I don’t know why, but my head immediately went to Heaven’s Gate, which was that cult in San Diego. I don’t know why.
Will Townsend: I’m a sci-fi nerd with all the lightsabers behind me, I went straight to the science fiction.
Matt Kimball: Oh, this is kind of science fiction, Haley’s Comet, right? Coming to get us on the comet. Anyway, here’s what I think. I think, Paul, you hit the nail on the head. Well, I think you both did. I mean, networking is obviously going to kind of come into focus, but also I think over the next six months to a year you’re going to see a lot more lot more companies involved in Stargate. This starts with three, well, three with two contributors, right? Microsoft and NVIDIA. I think this expands exponentially, not exponentially, but considerably with other contributing companies, either financially, or through technology, or some other means.
The last thing, my last prediction is, I think you’re going to see the stock of the custom silicon companies, Broadcom, Marvell, Synopsys, you’re going to see these companies, you’re going to see their stock increase considerably, because what they are building today, as this evolves over the next five years, it’s going to become finely tuned, and it’s going to require a lot of custom silicon to reach its optimal performance at scale.
Will Townsend: Matt, I would include Grok in that, if I’m pronouncing it correctly. They do the LPU, the language processing unit. They presented at that Cisco event. And I wasn’t that familiar with them, but I was quite impressed with what they’re doing. I think you’re absolutely right. And custom silicon drives a lot of this, right? Efficiency from a power perspective, a performance, latency perspective, a throughput perspective. So, yeah, I think the US wins, I mean, all around.
Matt Kimball: Yeah. All right. With that said, we hit 30 minutes on the dot, and I think we could probably go on for another few hours, but we’d just be repeating ourselves like three guys sitting around your citizens’ community drinking coffee and reminiscing about the old days. But with that said, this was a fun conversation. We need to do more of these of moving forward, find these topics and just dig deeper on them. With that said, guys, thanks for indulging and doing this, changing up the format last minute and covering this in more detail. I think it’s going to be of great use.
Again, if you are listening at home, do not take anything we said as investment advice. Yeah, we mentioned stock prices at custom Silicon companies. We do not mean you should go invest by any means whatsoever, or any other company for that matter. This is really just our positing our opinions and thoughts, and for entertainment purposes. So with that said, thank you for joining us, and I guess we’ll see you in a couple of weeks where we’ll talk about some other stuff that’s of interest.
Will Townsend: Cool.
Matt Kimball: Guys, thank you so much.
Will Townsend: Yeah. Happy Friday. Happy Friday.
