Oct 28, 2026 · 1h 9m
AI data centers outrun oversight and community consent
The Truth About Data Centers w/ Paris Marx
The infrastructure powering generative AI is expanding faster than public scrutiny, reliable data, and regulation of its immediate harms.
- 1AI infrastructure has shifted from conventional facilities to enormous GPU-driven clusters with uncertain definitions and rapidly escalating resource demands.
- 2Communities increasingly resist data-center expansion over water use, noise, pollution, tax incentives, and opaque corporate promises.
- 3Public debate fixates on hypothetical superintelligence while current environmental, economic, labor, and social harms remain weakly regulated.
Don't miss
Paris Marx and Ed Zitron contrast bipartisan local resistance to data centers with national political alliances focused on speculative AI dangers.
The brief
Ed Zitron and Paris Marx, author of Hyperscale, trace data centers from Google’s early facilities to the massive GPU clusters built for generative AI.
The Dalles, Oregon, becomes an early case study in water use and scrutiny, while newer AI facilities push power demands toward 100-megawatt and gigawatt-scale clusters.
Across the United States, France, Ireland, and elsewhere, residents challenge facilities over noise, ugliness, pollution, tax incentives, and the loss of local control.
The industry’s secrecy makes those disputes harder to resolve: NDAs, undisclosed GPU inventories, uncertain power consumption, and opaque financing limit independent verification.
The episode’s central warning is that speculative superintelligence distracts from harms already arriving—and from the political alliances and media hype helping expansion continue.
What was said on this episode
29 statements · 4 positive · 20 negative · 1 mixed · 4 neutral
Data-center operators are switching from evaporative to closed-loop cooling
“they're switching to what's called closed-loop cooling”
Listen at 8:13
Closed-loop cooling reduces water use but increases electricity consumption.
“you're using more electricity because you need to cool the water down”
Listen at 8:35
PFAS and coolant pollution will receive more data-center attention within 6–12 months
“we're going to see become part of the data center conversation a lot more in kind of the coming six to 12 months is my thinking”
Listen at 9:25
Data centers consume more than 20% of Ireland’s electricity
“more than 20% of the electricity in Ireland goes to data centers”
Listen at 11:51
Ireland’s data-center electricity demand has reduced grid reliability and raised consumer prices.
“that has impacted grid reliability. It has pushed up electricity prices for people”
Listen at 11:55
Cloud computing enabled generative AI to emerge in its current form.
“without the beginning of the cloud, I would argue that you wouldn't really see generative AI in the form that we have it today”
Listen at 23:33
The industry commonly uses 50 megawatts as a hyperscale threshold.
“the 50 megawatts is one that has been gravitated around”
Listen at 24:29
Fifty megawatts is commonly used as the hyperscale threshold
“50 megawatts is one that has been gravitated around”
Listen at 24:29
AI GPUs require more electricity and cooling than traditional data-center hardware
“The GPUs do run harder. They require more electricity. They throw off more heat, which means they need more cooling as well.”
Listen at 25:57
AI-focused data centers draw more electricity and water than traditional facilities
“the AI-focused data center is going to have a bigger draw on electricity, on the water system”
Listen at 26:14
Tech companies are building substantial fossil-fuel capacity for data centers
“the level of fossil fuels that they are building to power their data centers is just staggering”
Listen at 27:00
The tech industry has aggressively expanded its data-center ambitions
“We have really seen the tech industry push hard with its ambitions the past few years.”
Listen at 27:45
Hundreds of U.S. local governments have enacted data-center construction moratoriums.
“We have seen across the United States hundreds of local moratoriums being passed on data center construction.”
Listen at 28:07
Hundreds of U.S. localities have passed data-center construction moratoriums
“hundreds of local moratoriums being passed on data center construction”
Listen at 28:10
Oracle and CoreWeave’s future viability is increasingly questioned
“whether companies like Oracle or CoreWeave are actually going to be viable into the future”
Listen at 32:19
AI and data-center companies actively maintain secrecy around their infrastructure.
“they work very hard to maintain that secrecy”
Listen at 36:02
Data-center companies work hard to conceal projects and operations
“they work really hard to try to keep these things secret”
Listen at 36:22
Many Republican supporters oppose local data-center projects.
“so many Republicans, are also opposed to these data center projects”
Listen at 37:33
Many Republican supporters oppose the party leadership’s AI and data-center agenda
“so many of its supporters are not buying into that project”
Listen at 37:58
AI regulation is being directed toward hypothetical rather than current harms
“the regulation is based on fantasy and not on reality and kind of, you know, actual harms that people are feeling right now”
Listen at 45:44
Bernie Sanders advocated a nationwide data-center construction moratorium
“I was completely behind the effort to to bring forward a nationwide moratorium on data centers that he was advocating end of last year”
Listen at 47:27
Alexandria Ocasio-Cortez has addressed concrete AI harms better than Bernie Sanders
“Alexandria Ocasio-Cortez has been better on this”
Listen at 47:50
Alexandria Ocasio-Cortez focuses AI policy more on deepfakes and mental-health harms.
“Alexandria Ocasio-Cortez has been better on this and has been more focused on deepfakes and mental health”
Listen at 47:50
Chatbots are not thinking or developing independent consciousness
“The public is led to believe that these chatbots are actually thinking and developing independent consciousness and all this kind of stuff when we know very tangibly that that is not occurring.”
Listen at 54:58
Anthropic does not have proven five-gigawatt capacity
“Anthropic has five gigawatts of capacity. Even though they don't, there's no proof of that.”
Listen at 57:01
Generative AI has not delivered many promised social benefits
“We've not seen many, if any, of these positive benefits be realized.”
Listen at 58:55
Generative AI technologies have produced widespread social and environmental harms
“The amount of social, community, environmental harms that have proliferated because of these technologies.”
Listen at 59:01
AI and data-center expansion may eventually undergo a large-scale collapse
“Hopefully because of the scale of what we're seeing. what we see with AI and data centers and the scale of what the collapse will ultimately have to be when that comes.”
Listen at 1:04:24
The AI and data-center boom may produce a major reckoning after collapse.
“Hopefully that means that it is different this time and that there will have to be a proper reckoning.”
Listen at 1:04:38
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Books & mentions
Some links are affiliate links — PodLume may earn a commission if you buy.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.