Bohemian Swarm AI Podcast - Episode 2

Data Centers and Water: Who Holds the Risk?

The following is an AI-generated summary of the interview transcript.

We got together to talk about one of AI's most contested topics right now: data centers and water usage. The panel: Jess McDonald (AI Advisor, background in data analytics and impact/causality assessment), her father Brion McDonald (commercial general contractor who has built data centers and clean rooms), Steve Idowu (business and technology consultant who has deployed 25+ AI solutions), and Peaches the Cockatiel. Thomas Ott, AI Advisor, is today's moderator.

Notes from the Bohemian AI Swarm podcast — not affiliated with WWT, just coworkers talking shop.

Here's what came out of it.

The most misunderstood fact

• Steve: The fight isn't really about gallons — it's about who holds the risk. Tech companies treat water like a line item on a budget. Communities treat it like a birthright. Those two views are on a collision course.
• Jess: At the national level, data center water use is arguably overplayed — semiconductors and agriculture consume far more. But these decisions aren't made nationally. They're made locally, often for profit, and that's where real communities are feeling real impact.
• Brion: Water is just the medium for heat transfer — it doesn't have to be used and lost. New e-coolants (Shell has one in development for circuits, batteries, and radiator systems) point toward a future where the real question is conservation and reuse, not consumption.

How cooling has changed

• Old-generation CPUs ran hot but were largely air-cooled. The new breed of GPUs is pushing the industry hard toward liquid cooling.
• Brion described a prototype server rack — roughly 4 feet square, 6-7 feet tall — that draws a full megawatt, with hollow conductors actually plumbing fluid directly into the power and server lines.
• Two parallel paths forward: better liquid cooling integration, and semiconductor designs that are simply less energy-intensive to begin with.

Should water use be reported like emissions?

• The panel's consensus: yes, and it's already starting to happen. Up to 10 states — including Illinois, California, and South Carolina — have bills moving through their legislatures, and a federal bill (reportedly introduced by an Illinois senator) would require data centers to report water and energy use, with some states pushing for public dashboards.
• Steve's comparison: emissions reporting didn't stop climate change, but it ended companies' ability to say "we had no idea." Water reporting could do the same — turn accountability into something measurable, and something good actors can be proud of.
• Brion's point on messaging: a lot of community fear could be defused simply by explaining where the water comes from. His town is building a system that recycles water from sewage/wastewater treatment for drinking and landscaping — if a data center's water draw starts as recycled water, that changes the entire community conversation.

Are we moving too fast?

• Lowered federal thresholds have made it easier to fast-track projects, sometimes skipping the deeper environmental impact review that used to be standard — including in ecologically sensitive areas like deserts.
• CEOs are incentivized to move fast because they "live and die by the quarter."
• The group's frustration: we already know where the best sites are — near hydro power, for instance — but haven't built the incentives to steer development there rather than wherever's fastest to permit.

What does a "good deal" look like? (Steve's framework)

• Commit to recycled or non-potable water before construction starts, not after community backlash forces a promise. (Steve cited Google's Ohio facility, which reportedly pledged 100% recycled water only after pushback.)
• Real, enforceable limits on water withdrawal — not vague marketing language.
• Clarity up front on accountability: who's responsible when something goes wrong.
• Real investment in local infrastructure as a good-faith down payment — Steve referenced a Kansas City project where a company reportedly invested $500M+ in local water recycling infrastructure to get approved.
• Proactive, honest community engagement — Meta's approach in one community (town halls, direct mail, explaining the benefits) was described as effectively running "a political campaign."
• A cautionary counterexample: AWS reportedly withdrew as anchor tenant on a Tucson, AZ project after public pushback over Colorado River water rights — a sign that public response can and does change outcomes.

Should water-stressed regions be off-limits entirely?

• Steve's take (he called it controversial): don't ban the location, ban the method. If a company can prove it isn't competing with local water resources, location shouldn't automatically be the fight.
• Jess pushed for case-by-case, data-driven review rather than blanket rules, with multi-disciplinary teams — data scientists, environmental engineers, and contractors — at the table together.

Geothermal as a possible path forward

• Brion is currently working on a geo-exchange building where the ground stays around 66-67°F year-round, letting the system heat and cool by overcoming only a few degrees of difference rather than fighting outside air temperature.
• Applied to a data center, a closed-loop geothermal system could theoretically take water use from something like a million gallons a year down to a couple hundred (illustrative numbers, not measured figures).
• The honest caveat: unlike a building that both heats and cools (balancing the ground temperature over a year), a data center's cooling-only, one-directional heat load could have a bigger localized thermal impact on the ground beneath it — an open question the panel flagged rather than resolved.
• Other approaches mentioned in passing: underwater data centers (already happening in China) and early exploration of space-based data centers.

Ten years from now — reckless or well-managed?

• Jess: Optimistic. She sees signs that tech leadership is waking up to the policy gap and believes the next year will be a forcing function — but warns that if the current pace continues unchecked, billions will get wasted and public trust will erode further.
• Steve: "Right now it looks reckless." He expects history to judge the first five years of AI-scale buildout as reckless, but believes companies that correct course over the next two years will be the ones that survive — and that conversations like this one genuinely move the needle.
• Brion: Called it "helter skelter" in the near term, with success depending on treating this as an iterative process — identify mistakes, correct, repeat.
• Thomas (me): I'll be honest — I think we're being reckless right now. My background is in civil engineering and environmental impact statements, and I lived in New Mexico for six years, which taught me how fragile water resources really are. The same week we recorded this, Angel Fire, NM declared a water emergency — and someone on Reddit sarcastically joked about building a data center there. That's the gap we're operating in. I'd like to see environmental engineers, civil engineers, and contractors sitting alongside data scientists on every one of these projects, not brought in after the fact.

Where we landed

• Jess: It's not too late. Get involved, be vocal, be authentic — we've solved harder problems than this before.
• Brion: Call your representatives. Push for data centers that represent community interests as well as business interests, and lean on universities to help drive real solutions.
• Steve: Don't watch the sustainability pledges — watch the deal terms. Every hyperscaler has a nice "water positive by 2030" promise. What actually matters is whether they agreed to real, enforceable withdrawal limits before being asked to. That information is public — go look it up.
• Thomas: Grassroots, local involvement is still the fastest way to see results, and it bubbles up. This isn't a simple problem, but it's a solvable one if everyone brings their actual expertise to the table.