Data Centers: Can't Live With Em, Can't Live Without Em
Episode
48 min
Read time
2 min
Topics
Productivity, Investing, Fundraising & VC
AI-Generated Summary
Key Takeaways
- ✓AI Infrastructure Scale: ChatGPT trained on 20,000 Nvidia GPUs; modern AI data centers string together 100,000-200,000 GPUs to form supercomputers. Single facilities consume electricity equivalent to towns of 50,000 people, with Meta's Hyperion using five gigawatts at peak.
- ✓Investment Bubble Risk: Tech companies pledged $3 trillion for data centers between 2025-2030, yet only 5% of business AI pilots secure returns on investment. Private credit financing remains largely unregulated, creating potential economic instability similar to 2008.
- ✓Local Resource Depletion: Data centers in Northern Virginia increased electricity prices 267% since 2020. Meta and Microsoft consume 7 million gallons daily in Phoenix. Newton County Georgia facilities use 10% of local water supply, with some communities experiencing actual water shortages.
- ✓Employment Misconception: A £10 billion Northumberland data center creates only 400 full-time jobs. Hyperscale facilities require minimal staffing due to automation efficiency, providing limited economic benefit to host communities while consuming disproportionate resources and raising utility costs for residents.
What It Covers
Data centers evolved from mainframes to hyperscale facilities consuming massive electricity and water resources. AI computing drives exponential growth requiring hundreds of thousands of GPUs, creating environmental concerns and infrastructure strain worldwide.
Key Questions Answered
- •AI Infrastructure Scale: ChatGPT trained on 20,000 Nvidia GPUs; modern AI data centers string together 100,000-200,000 GPUs to form supercomputers. Single facilities consume electricity equivalent to towns of 50,000 people, with Meta's Hyperion using five gigawatts at peak.
- •Investment Bubble Risk: Tech companies pledged $3 trillion for data centers between 2025-2030, yet only 5% of business AI pilots secure returns on investment. Private credit financing remains largely unregulated, creating potential economic instability similar to 2008.
- •Local Resource Depletion: Data centers in Northern Virginia increased electricity prices 267% since 2020. Meta and Microsoft consume 7 million gallons daily in Phoenix. Newton County Georgia facilities use 10% of local water supply, with some communities experiencing actual water shortages.
- •Employment Misconception: A £10 billion Northumberland data center creates only 400 full-time jobs. Hyperscale facilities require minimal staffing due to automation efficiency, providing limited economic benefit to host communities while consuming disproportionate resources and raising utility costs for residents.
Notable Moment
The hosts reveal they have never personally used AI chatbots or large language models despite the technology touching every aspect of modern life, choosing to continue working methods that function well without contributing to resource consumption.
Episode Transcript
This is an iHeart podcast. Guaranteed human. Refresh your bathroom with big savings at Lowe's. Reimagine your bathroom with up to 40% off select faucets during our bath savings event. Need it today? Order by 2PM for same day delivery by 8PM. Shop for your bathroom refresh at Lowe's. We help. You save. While supplies last, selection varies by location. Same day delivery on eligible in stock items subject to availability. Hey, and welcome to the podcast. I'm Josh, and there's Chuck, and Jerry's here too. And we've got our pocket protectors and tape on the bridge of our glasses. Hey. And this is stuff you should know. Nice work. Did you just hear something? No. Oh, weird. I heard you say nice work. Yeah. But then I stop abruptly. Well, I stopped abruptly because I thought I heard a little digital glitch. Oh, no. I didn't hear anything. I might be losing my mind then. I I you know, I'm curious whether we end up editing this out or not. Any other podcast on the planet would edit that out without even thinking about it, but there's, like, a 50% chance it'll stay in with us. I mean, this is why we didn't get that Golden Globe nomination. That's right. This kind of classic stuff you should know unprofessionality. That's exactly right. That and the Italian accent, the mispronunciations, there's a whole laundry list. Yeah. That's okay, Chuck. I think we're we're golden regardless. Agreed. So, we're talking about data centers, which I had a very, very rough idea about, but actually, no. I I knew that they existed essentially and that they were Yeah. Becoming a problem with, the rise of AI. Yes. But that was about it. How about you? Are you are you date a data centerphiliac? No. You know me. I'm not I'm not super technology minded, so I don't know a lot about this stuff. Mhmm. I I remember walking by our server room back in the day when we were at Ponce City Market and seeing, our colleague Izzy in there, hard at work. Yeah. And when that door was unlocked and open, hearing the whir Mhmm. Of the of the servers and the cooling machines and, you know, that that's a on a smaller scale, that's a a data center. Absolutely. A 100% that's a data center. It was also a great place to curl up and take a nap in the middle of the day. Like, the the warmth of the server. Yeah. And and the WER puts you right into sleep. Yeah. So, yeah. That definitely counts as a data server. If you have, like, one of those, little home networking setups in, like, a closet in your house, data center. Sure. Technically, the PC is a data center. Anywhere you can store and access data, that's technically a data center. Yeah. And you're like, well, that's stupid. Why did you even say that, Josh? It's quibbling. That's quotidian. Shut …
Get the full transcript (8,635 words) + summary by email — free
One-time email with the complete transcript and AI summary of this episode. No account needed.
One email, no spam. We’ll also show you what SignalCast does.
You just read a 3-minute summary of a 45-minute episode.
Get Stuff You Should Know summarized like this every Monday — plus up to 2 more podcasts, free.
Pick Your Podcasts — FreeKeep Reading
Books, tools, and gear mentioned in this episode
SignalCast may earn commission on purchases via these links. As an Amazon Associate, SignalCast earns from qualifying purchases.
Gear
by NVIDIA
“ChatGPT trained on 20,000 Nvidia GPUs; modern AI data centers string together 100,000-200,000 GPUs to form supercomputers.”
More from Stuff You Should Know
We summarize every new episode. Want them in your inbox?
Similar Episodes
Related episodes from other podcasts
The Vergecast
Sep 2
What’s happening with US data centers?
Odd Lots
Aug 21
Jasmine Sun on What the AI Industry Got Wrong About the Public Backlash
Eye on AI
Aug 10
In 5 Years, 90% of What You Use AI For Will Run on Your Smartphone | Paolo Ardoino, Tether
Odd Lots
Jul 15
NY Governor Kathy Hochul on Her One Year Data Center Moratorium
NVIDIA AI Podcast
Mar 4
Powering the AI Inference Wave with EPRI's Ben Sooter - Ep. 292
Explore Related Topics
This podcast is featured in Best Science Podcasts (2026) — ranked and reviewed with AI summaries.
Read this week's Investing & Markets Podcast Insights — cross-podcast analysis updated weekly.
You're clearly into Stuff You Should Know.
Every Monday, we deliver AI summaries of the latest episodes from Stuff You Should Know and 192+ other podcasts. Free for one show.
Start My Monday DigestNo credit card · Unsubscribe anytime