This content was created by Dylan Patel. This content was uploaded by Dylan Patel. This is a YouTube video titled "Elon Musk bought an abandoned washing machine factory and made it the largest GPU cluster at the time" (created 2025-07-31). The excerpt is a short, diarized conversation between Dylan Patel and another speaker in which Dylan explains Elon Musk's unconventional conversion of an abandoned washing-machine factory in Memphis into a rapid-deployment H100 GPU data center — tapping a nearby gas main with mobile gas generators and using mobile chillers to run large-scale GPUs before grid interconnections are ready — and contrasts that speed-focused approach with Google's purpose-built data centers.
Chapters
Dylan explains Elon's 200 IQ 2024 H100 data center.
It's like what Elon is doing, right? Um, and Elon is like pretty crazy because like what he's doing is like very non-conventional. Um, so they bought a factory in Memphis, right? Tennessee. Um, which is interesting. Um, and this is like, this is like an appliance factory, right? It used to make like washing machines before it got shut down, right? Um, and it's been shut down for like a while. They bought a washing machine factory and converted it into a data center. But like the reason they bought this factory is because right along the side of it is a humongous gas line, right, which goes to a huge gas power plant. And instead of like tapping the power from the gas power plant, which they're trying to do, but the grid interconnections are like very slow, right? So you can see in the yellow that one, there's 150 megawatts that could come from the grid, but that substation, you know, the transformers and such are not built until next year. And he doesn't want to wait till next year to start training his humongous models, right? So instead he does like something incredibly different, right? He taps that main gas line and buys a bunch of mobile generators, right? Rents them, buys them, whatever. Some of these are for like disaster relief and things like that, right? But like he's just like got mobile generators plugged in to the main gas line and then converted this factory into a data center. That's crazy. And then all those trucks on the bottom, on the bottom, on the backside of the factory, right? Not the parking lot, but the other side are chillers, right? So they just cool water down. So like, it's interesting, these are all like mobile chillers as well, right? Like you see, they're all just like semi-truck, like, like there's like 60 of them, right? Versus like what Google does is like very like, I'm sorry, Google, what Google does is like very purpose-built, right? They look like they're engineered for this purpose, whereas like This is clearly not engineered for this purpose, right? But it's all about how fast I can get the model out there, right? So if I can get 100,000 GPUs working this year, then I am now competitive with, at least on a compute standpoint, with OpenAI and Google, right? And so the speed of what they're doing is interesting.
