Episode Summary
Executive Summary: Michael Thomas of CleanView explains how AI-driven data center demand is reshaping U.S. power development, pushing companies toward off-grid gas, rapid-build “tent” facilities, and other extreme measures to secure capacity. He details how his data-plus-investigative workflow surfaces these projects, why some bypass the grid while others still benefit from it, and how the boom raises complex questions about prices, water, noise, emissions, and public trust.
Main Topics: CleanView’s evolution into data center tracking (Priority: 5/5): Thomas describes how policy uncertainty around renewable energy forced CleanView to broaden beyond clean energy project tracking into a second product focused on data centers and grid demand. Investigative workflow and public-record sleuthing (Priority: 5/5): The discussion explains how CleanView combines permits, SEC filings, satellite imagery, and construction tracking to uncover what companies are actually building, often before public disclosure. Off-grid gas and the rise of “Mad Max” power development (Priority: 5/5): They examine how AI companies—especially xAI and Meta—are using mobile turbines, behind-the-meter generation, and tent-based deployments to shortcut long grid interconnect timelines. Grid vs. off-grid economics and project decision-making (Priority: 4/5): Thomas compares cases where off-grid generation sped up deployment with cases where grid-connected sites or pre-permitted locations were faster and more viable, showing the tradeoffs are highly site-specific. Public backlash: water, bills, noise, and trust (Priority: 4/5): The conversation covers why communities oppose data centers, including concerns over water scarcity, electricity affordability, local pollution, and a broader distrust of big tech and AI. Emissions, fossil-fuel lock-in, and policy limits (Priority: 5/5): Thomas argues the current data center boom is driving more gas buildout, delaying coal retirements, and raising near-term emissions, while federal policy to constrain impacts appears politically unlikely. Long-term outlook: clean tech opportunity or fossil resurgence (Priority: 4/5): The episode closes with uncertainty over whether AI demand will accelerate clean firm technologies like geothermal and nuclear or instead entrench a more fossil-heavy power system.
Key Arguments: CleanView survived a period of renewable-sector freeze by expanding into data center tracking, because AI infrastructure became a major new driver of grid demand. Public records and satellite imagery can reveal major infrastructure projects before companies acknowledge them, making data analysis a powerful form of investigative reporting. Off-grid gas data centers exist because grid interconnections can take years; companies are paying a premium for speed to power, not necessarily for a permanent operating model. xAI’s Colossus showed that mobile turbines can bring a gigawatt-scale facility online far faster than waiting for utility power, inspiring similar strategies across the sector. Not all off-grid strategies win: Amazon’s New Carlisle site shows that a well-sited, pre-permitted grid-connected project can still be faster than building standalone power. The data center boom is intensifying local opposition because communities experience direct impacts—noise, water use, pollution—while distrusting company promises. Electricity bills are nuanced: some regions may see price pressure from data centers, but in other cases new load can lower costs for everyone else if infrastructure is already in place. Water use is a local-siting issue more than a total-volume issue; in drought-prone regions, even relatively small increases can trigger major concern. The sector’s emissions impact is likely large and growing, especially as companies rely on gas turbines and other fossil generation to meet AI demand quickly. Federal legislation to protect ratepayers or cap emissions from data centers is politically unlikely right now, so the debate will remain local and uneven.
Data Points: CleanView tracked clean energy projects: 10,000+ - Size of the company’s clean energy project database CleanView tracked developers: 500+ - Number of clean energy developers in the database Data center projects tracked nationally: 3,500+ - Scale of CleanView’s data center tracking Operating data center projects: 1,300 - Tracked data centers already in operation Planned data center projects: 2,000+ - Tracked data centers in development Off-grid / behind-the-meter data centers tracked: 100 gigawatts+ - Universe of off-grid data center announcements being tracked Operating off-grid capacity: a couple of gigawatts - Rough current operating behind-the-meter data center capacity Off-grid capacity under construction: 5–10 gigawatts - Estimated pipeline of off-grid projects actively being built Amazon Gigawatt Ranch natural gas plant: 7.65 gigawatts - Size of the proposed off-grid natural gas power plant tied to Amazon’s project New Carlisle Amazon data center timeline: about 2 years - Grid-connected gigawatt-scale data center built unusually fast in Indiana Dominion Virginia interconnect timeline: about 7 years - Utility estimate for connecting large new loads in data center-heavy Virginia XAI Colossus timeline: within a few months - Mobile turbines enabled rapid startup of a very large data center Meta New Albany gas buildout: 200 megawatts initial + another 200 megawatts planned - Off-grid gas powering Meta’s Prometheus cluster in Ohio Meta New Albany ultimate scale: more than 1 gigawatt - Projected final size of the off-grid gas-powered campus Anthropic annualized revenue growth: $10 billion to more than $65 billion - Used as an example of AI compute demand growth in the discussion Revenue per gigawatt economics: $10 billion to $50 billion per gigawatt - Estimated economics for data center / compute capacity in current AI market PJM battery buildout: less than 1 gigawatt - Used to contrast underbuilt storage in the largest U.S. grid region Texas battery capacity: more than 10x PJM - Illustrates Texas’s much larger storage buildout relative to PJM Amazon project emissions claim: potentially largest single source of pollution in America - If run at industry-average utilization, the site could out-emit many coal plants XAI expansion plan: 10 to 20 gigawatts by end of 2027 - Musk’s stated goal for off-grid data center capacity, with timeline uncertainty Vantage VA2 turbines: 7 gas turbines - Illustrative example of noise and neighborhood proximity concerns Water concern ranking: #1 public concern - Polling cited by Thomas on why people oppose data centers
Pivotal Quotes: "There are a lot of these like nameplate gigawatt scale projects that get announced." — Michael Thomas: On the gap between announced AI infrastructure and what is actually under construction or operating "We’re now tracking more than 100 gigawatts of these behind-the-meter data centers." — Michael Thomas: On the rapid explosion of off-grid / off-electric-grid AI power development "I think the trajectory right now is we’re going to see a lot more gas electricity generation and a much slower phase out of coal." — Michael Thomas: On the climate and power-system implications of the AI data center boom
Implications: Listeners should expect AI infrastructure to keep stressing grids, accelerating gas buildout, and intensifying local fights over water, noise, and bills. The near-term winners will be companies that can secure power fastest; the big unanswered question is whether policy or clean firm tech can redirect that demand.