AI's Material Footprint: Data Centers, Power, Water, and Money (Week 5)
The short version: A Rutgers lecture frames AI's data-center boom as an honest 'we don't know yet' story — but the power, water, and land bills are already due in real towns.
Our Take
Credit to Kiran Garimella (Rutgers SC&I) for a rare thing in the AI discourse: intellectual humility. Framing the buildout as bubble-and-revolution-not-opposites, and walking students through railways and dot-com fiber as precedent, is a useful corrective to both the hype merchants and the doomers. The point about chips no longer being the bottleneck — power, construction speed, and local siting are — is exactly what we track county by county, and it's good to see it land in a classroom instead of just an earnings call.
But "we don't know yet" is a fine academic posture and a lousy answer for the family whose electric bill just jumped because a colocation site or a multi-gigawatt frontier campus got fast-tracked next door. The tiers this lecture lays out — tens of megawatts up to gigawatt-scale — aren't abstractions in Loudoun County, rural Georgia, or the Arizona desert; they're substations, water permits, and tax abatements signed by people who often never asked residents first. That's the part academic frameworks tend to skip: who bears the cost while the sector figures out if this is 1990s fiber or 2000 dot-com wreckage.
If you want to see whether a lecture slide about "local siting" has a real address near you, check our facility map, and if you're ready to push back on a deal before it's finalized, our take-action page has the playbook.
This is GridWatch the USA’s original commentary. The video above is the work of gvrkirann, published on YouTube — full credit to the creator.