Jason Kelce’s Pee-Mail Stunt Puts AI Data Center Water Use in the Spotlight

Former NFL player Jason Kelce is the face of a new marketing campaign, launched Tuesday, that asks people to bottle their urine and mail it to AI data centers. The stunt is silly by design, but the underlying issues are not. AI data centers consume up to 5 million gallons of water per day according to a 2025 report, and a Gallup poll now shows 71% of Americans oppose building new ones in their communities. The campaign has turned a genuine public anger into a viral moment.

What happened

Data point	Detail
Campaign launch	Tuesday (recent)
Max daily water use (large data center)	5 million gallons
Equivalent population water use	10,000 to 50,000 people
Projected power, Ports-Pike Campus (OH)	~8 gigawatts
Meta Hyperion campus size	4 million square feet, Louisiana
Americans opposing new data centers locally	71% (Gallup)
Total US AI data centers	~4,000

Jason Kelce, co-owner of Garage Beer and co-host of the New Heights podcast, stars in a now-viral ad that opens with him in the bathroom. He then holds up a jar of yellow liquid and invites the public to send urine to “the AI data center of your choice,” framed as a way to help cool the servers. The ad was made in partnership with Liquid Death, the beverage brand known for its irreverent marketing. The collectible mug for bottling your contribution is part of the campaign merchandise.

The joke has a factual hook. According to the Cleveland Clinic, urine is 91% to 96% water, and water is exactly what data centers burn through at scale.

Why the water and power numbers actually matter

AI data centers house the high-performance graphics processing units (GPUs) that run and train AI models. Those chips generate enormous heat. Evaporative cooling towers, the most common solution, absorb that heat using fresh water, and large facilities can consume up to 5 million gallons a day, according to a 2025 report from the Environmental and Energy Study Institute. That is comparable to the daily water consumption of a town with 10,000 to 50,000 residents.

The electricity picture is just as stark. Two planned projects illustrate the scale:

• The Ports-Pike Technology Campus in Pike County, Ohio (an OpenAI and Nvidia project) is projected to consume roughly 8 gigawatts, equal to the annual energy use of 6 million average US households.
• Meta’s Hyperion campus in Louisiana, a 4-million-square-foot facility built on former soybean and river cane land, is projected to use three times the power New Orleans consumes on a typical day, according to the Institute on Taxation and Economic Policy.

Environmental activist Erin Brockovich has built a public database tracking data center locations. Most are going up in rural areas. In a blog post she described a pattern of backdoor deals and limited community disclosure, with residents noticing rising utility bills and unexplained sick animals near facilities.

Is AI data center opposition really a mainstream issue?

Yes, and it crosses party lines. A recent Gallup poll put opposition to new local data center construction at 71% of Americans. With US midterm elections approaching, that number has become politically significant. The National Republican Senatorial Committee circulated a memo, reported by Axios, warning that Ohio incumbent Senator Jon Husted could lose unless Republicans improve their messaging on data centers. The NRSC stated: “If voters’ perceptions of data centers are not fixed quickly, the campaign against them will expand far beyond Ohio.”

Protests have already taken place in cities including Imperial, California. The issue connects resource anxiety, utility cost fears, and distrust of opaque corporate deals into a single flashpoint. Our earlier coverage on the public trust crisis around AI shows this sentiment has been building for a while.

Our take

Kelce’s ad is a marketing stunt. Nobody is actually cooling a GPU rack with a mug of urine. But the campaign works because it attaches absurdist humor to a grievance that is both real and broadly felt. The water and power figures cited are sourced from credible research, not activist exaggeration, and a 71% opposition rate in a Gallup poll is not a fringe position.

For businesses building AI-dependent products or advising clients on AI integration, this is worth watching for two reasons. First, growing regulatory pressure on data center siting and resource use could affect where and how AI infrastructure gets built, and by extension, availability and pricing of cloud AI services. Second, the public mood around AI is souring in ways that go beyond abstract ethics. Opposition is now local, practical, and tied to utility bills and water supplies.

That is a harder narrative to counter than concerns about robots taking jobs. Any business that publicly leans into AI should be thinking about how it communicates the tradeoffs, not just the benefits.

What to do about it

1. Track proposed data center projects near your operations or clients, since local utility rate changes may follow.
2. Review AI vendor sustainability disclosures. Major providers like Google, Microsoft, and Amazon publish water and energy usage data; use it when evaluating platforms.
3. If your business communicates publicly about AI use, get ahead of the resource question with honest, specific answers rather than waiting for it to become a customer relations problem.
4. Watch midterm election outcomes in Ohio and other data-center-heavy states for early signals of regulatory shifts that could affect cloud infrastructure costs.

The joke about mailing pee will fade. The pressure on AI infrastructure resource use will not.

Source: Bing News · OpenAI