Loading summary
Transcript1 lines
- [00:00]
Unnamed Host/Presenter
Mark Zuckerberg has taken two threads to make a massive announcement and that is that Meta is currently building a 5 gigawatt AI data center. Absolutely enormous. And there's some crazy news about them possibly just throwing up tents because they're trying to get this thing so built so fast. Mark Zuckerberg, who has, you know, famously been, it felt like lagging a little bit on AI, does not like that. Does not like that. People said he was, he was getting behind and is now going full tilt. I feel like we kind of have these like hot and cold cycles from Mark where he just goes really hard for like a few months, then he like chills while everyone does something. And then if anything, any crap like starts hitting the fan or Meta starts getting behind an area, he comes back full swing and goes crazy. So we're in one of those moments. He's been on an absolute tear. $100 million offers to tons of AI researchers. He's just recruited all of tons of top talent from every single, every single one of the top AI companies for his super intelligence project that he's building. And now the data center element is coming, which is going to be very interesting and allow him to compete with a lot of other big companies, including OpenAI, Google and X AI in kind of this pairing that we're seeing where these, these, you know, these software tech companies are not just building the models, but they're actually trying to build out the infrastructure as well because. Because they know it's really just a very cutthroat race. Especially when you have companies like Google that control data centers or Google Cloud and the AI models. And now you're trying to comp. And then when they're trying to compete with Google, they maybe they are so desperate they're going to go to Google for some of the compute and then all of a sudden it's like you're funding your competitors going to turn around and use that money against you. So, you know, these guys really want to build it out on their own. So we're going to be diving into everything that he's announced, where we think this is going in the future. Before I do, I wanted to mention if you want to try out all of these, all of the latest AI models, including everything from Meta, go check out AI Box AI, that's my own startup that currently is in beta. We have a playground that allows you to use the top 40 AI models all in one platform for 20 bucks a month, so you don't have to have $20 subscriptions to all of These different companies, we have tons of different open source models from Llama, including llama, including llama, 3.1405 B turbo, a bunch of other interesting models. We have Google, Deepsea, Cohere, Anthropic, OpenAI, XAI, Quen, Nvidia, a bunch of cool models and also some a bunch of image models you may not have tried and speech models and you can use them all in the same platform in the same chat. You can switch between what model you're talking to. So it's very useful. Go check it out. AI box do. Okay, let's get into the latest news from Meta and Mark Zuckerberg. So this came out on Monday, but essentially this is the newest move that he's trying to use to get ahead of OpenAI and Google. This 5 gigawatt comp of computational power essentially for his AI lab. We know that he's got a bunch of players that are helping him lead the software side. He just was able to poach the former CEO of Safe Superintelligence. That was the company that Ilya Sus, who was one of the co founders of OpenAI. He left, started his company, raised a billion dollars before a product ever existed. Everyone, it's gotten this eye watering number and he had this incredible CEO who'd come over as well, which was Daniel Gross. Now Daniel Gross's left is over with Zuckerberg. Now we also have the CEO of scale AI, Alexander Wang, who also had a bunch of data essentially to help train AI models and had OpenAI and Google as a client. And now Meta sucking up and Google said, hey, we're not going to use scale AI anymore because of kind of the way this is going down. But in any case, he's got this really kind of the dream team, this powerhouse team of people behind it. So what he needs now is the infrastructure. He's now turning all of his attention now that he's got an amazing team to building out this massive computational, all of the massive computational power that can be needed to train the actual models that they are building. So he announced the name of this, you know, this huge thing as it's called Hype Hyperion. And he said that the footprint, so the actual size of it is going to be big enough that it could cover most of Manhattan. So this is absolutely insane. Size wise. Spokesperson for Meta, Ashley Gabriel, said that it's going to be located in Louisiana, probably in Richland Parish, where Meta already announced this $10 billion data center development that they're going to be doing there. And so it seems like this is probably going to be in the same place. They're going to bring 2 gigawatts of data center capacity online by 2030. So this isn't quick, right? That's five years in the future, still a lot of, a lot of computer. So in five years they're going to have 2 gig, 2 gigawatts online, but they're going to scale it to 5 gigawatts several years later. So the plan is, you know, a lot of that is groundwork and then once that's kind of built out a few years later they'll be able to get it. So you could imagine by like 20, let's say 2033, maybe things get a little delayed, 2035, it's like 10 years in the future before they're getting this thing like really humming at full capacity. So honestly I think this is, it's kind of an interesting thing. It just shows how much work and how much time it takes to do this. It's honestly one of the reasons why a lot of people were very impressed by Xai being able to spin up a data center as fast as they did with a hundred thousand GPUs is because this stuff just takes a long time. All of the, you know, getting everything done right with the government, with, you know, getting all of the right documents and everything in place to have this thing go through it is quite a process. So Zuckerberg also said that Meta is going to bring a 1 gigawatt super cluster called Prometheus online by 2026. So this is much quicker. Next year at some point we're going to get, he's going to get one gigawatt online and so he's going to be one of the first tech companies to have an AI data center that's this big. So Prometheus is going to be located in New Albany, Ohio. Meta's AI data center build out is probably going to make them very competitive with OpenAI, Google, Anthropic and kind of what they're actually able to train and do. You know, you kind of see one of two strategies where these AI companies are building out like, like Meta or Google the, you know, the data centers or you see the strategy of something like Anthropic where they partnering with aws who already has a lot of the infrastructure in place to use their infrastructure and go and do it themselves. So overall you have a couple different options. Meta obviously is not going to rely on any other tech company as they know that they're quite literally competing Directly with, with many of them. So between the two of them, what a lot of people are talking about, I don't know, it's, it's not really my favorite talking point on all of this. I, I tend to think we're gonna, we're gonna solve a lot of the energy problems. Like we know we have to figure them out but, but there is the energy issue, right? So we're building all these data centers. They're gonna use a lot of energy. We, it's, it's again another interesting thing that we found with xi. When they spun up their data center very quickly, it's like, where did they get the energy? They just imported tons of diesel generators, fired them all up and are running, you know, I think in Louisiana somewhere. They're near Memphis, Tennessee actually. This huge data center just powered off of diesel generators. So like there is a way to do this quite quickly. I don't think that's going to be met as strategy. So they're going to have to figure that out. People also have kind of accused them of like, like, oh, it's going to be bad for electricity or it's going to be bad for water because the data centers use a lot of water. My one point I'll make on water and I think energy is a legitimate concern. We got to figure out where we're, what we're doing and by what we're doing we've got to figure out how to build more energy compute in America. That's probably nuclear. That's also though, something that takes a long time to build up. China's building up all the energy capacity that they need. They're going to completely smoke us in AI if we don't build it up. So I think the energy problem is not one to take lightly. It's very serious and you know, we got to figure out how to get more energy creation inside of America. The water one concerns me a lot less. They talk about, I don't know. Apparently in the New York Times they said that in Newton County, Georgia, Meta's data center project they have over there has made it so that the taps have run dry for some people's homes. I don't know what exactly the issue is there. I don't imagine this is a long term thing where people's taps in their house are just going to run out of water because the data center soaked up so much water. You'd imagine you probably want to build these things in places where you have easy access to a lot of water. So in any case, I'M not sure what the problem is there. Also, the other thing that I would love to bring up is just the fact that when, when we talk about data centers using water, it's not like they, they take in like infinite water and then just like use it and it disappears. What, what the water is used for is cooling the machines down so it's actually circulating in a reservoir and recycling. So yes, like maybe when Meta's building this, you know, data center in Georgia, filling up, like maybe filling up their reservoirs take a lot, takes a lot of time, but like it's not. The water all turns into steam and disappears. Like they keep it inside of a system, they're using it for cooling, they're recycling the water, they're reusing the water. It's not like the water disappears. So all I'm saying is the water concerns me a lot less because you can keep it inside of the system, the electricity, obviously you use it and burn it and it's gone. We got to figure out how to make more electrical compute. So that's me trying to be as, I don't know, middle of the road on that, is on that as possible. In any case, we know that there's a ton of competition. So whether Meta does this or not, people are going to criticize them for those types of issues. Whether they do it or not, other players are going to do it and so they're obviously going to compet. We have the AI Hyperscaler core weave is planning a data center expansion in Dallas as well. We have, you know, the huge Stargate project that was announced by OpenAI and Oracle. We have the massive data center by Xai, the Colossus supercomputer. So all of this is going down right now. It's going to be interesting. There was a column recently featured in the Economist on Monday and the US Secretary of Energy, which is Chris Wright, called for the U.S. to quote, lead the next major energy intensive frontier, artificial intelligence. He said that AI transforms electricity into quote, most valuable output imaginable intelligence. And he also is, you know, essentially talking about the fact that the federal government is going to accelerate the production of high energy derived from coal, nuclear, geothermal and natural gas. So evidently people in the, in the government are focus. They can see what is happening. The writings on the wall, everyone's scaling and scrambling as fast as they can to build these huge data centers. We need energy to power it. And so I think we're, I think we're going to get there. But, but China is also doing this to quite a high degree. So we have, we have competition and it's going to be interesting to see what happens with all of that. So in any case, I will keep you up to date as new companies announce these data centers as they get closer to completion, as the we're grappling with the energy side of it, the crisis that may arise there, all of that. Thank you so much for tuning into the podcast. And if you want to try all of the latest AI models, make sure to go check out AI box AI. Thanks so much for tuning in. Leave a review if you enjoyed the episode, and as always, I hope you have a fantastic rest of your day.