There is a moment in Jason Kelce’s new beer commercial where the former Philadelphia Eagles center stands at his toilet, fills a glass jug, walks outside, and starts singing. A crowd joins him. Everyone is holding a container of urine. The chorus goes “we want your pee, please give us your pee.”
This is an advertisement for Garage Beer, the brand Kelce co-owns, made in collaboration with Liquid Death. It is also, somehow, the clearest signal yet that opposition to AI data centers has stopped being a niche local zoning fight and turned into mainstream American consumer marketing.
The short version
- The stunt: a musical ad from Garage Beer and Liquid Death asking Americans to mail urine to AI data centers
- The product: a limited edition “Data Center Coolant Collector,” which is a resealable glass mug. It sold out.
- The disclaimer: the ad itself tells viewers “the suits want us to tell you to please don’t actually send your pee”
- The real number: data centers consumed roughly 4.5 trillion liters of water in 2025, according to a United Nations University report
- The projection: 9.3 trillion liters a year by 2030, about 2.5 trillion gallons
- The politics: 7 in 10 Americans oppose an AI data center being built in their area, per Gallup
When a beer brand decides your infrastructure is the punchline
Liquid Death has built an entire company on turning bottled water into a heavy metal joke, so a collaboration with an NFL retiree’s beer label was never going to be subtle. What makes this one notable is the target. Brands pick fights they think their customers will cheer for, and two companies just decided that dunking on AI data centers is safe territory.
“Water is the world’s most valuable resource, and data centers are a hot topic, so we got together to have some fun, while raising a little awareness,” Garage Beer co-owner Andy Sauer said of the campaign.
The marketing read is straightforward. Data center opposition is now popular enough that a mass market beverage brand can build a national campaign around it without alienating anyone it cares about. That is a meaningful shift from two years ago, when the conversation was mostly confined to county commission meetings in Virginia and Arizona.
The polling says the joke is not the fringe position
Gallup surveyed 1,000 US adults between March 2 and March 18, 2026. Seven in ten said they oppose the construction of an AI data center in their local area, and 48 percent said they strongly oppose it.
The comparison that stopped people cold: opposition to a local data center ran higher than opposition to a local nuclear power plant.
Would you oppose this being built in your area?
Source: Gallup, telephone survey of 1,000 US adults, March 2 to 18, 2026.
Gallup also asked opponents why. Resource consumption dominated the answers, and water was cited as often as electricity.
| Reason given by opponents | Share |
|---|---|
| Excessive use of resources overall | About half of opponents |
| Water use specifically | 18% |
| Energy use specifically | 18% |
| Pollution | 16% |
| Impact on local quality of life | About 1 in 5 |
| Negative economic consequences | About 1 in 5 |
What the actual water figures look like
This is where the joke stops being funny. In June 2026, the United Nations University Institute for Water, Environment and Health published a report titled “Environmental Cost of AI’s Energy Use: Carbon, Water and Land Footprints.” Its numbers are the most widely cited global estimates available right now.
Data centers consumed about 4.5 trillion liters of water in 2025. By 2030, the report projects that figure roughly doubling to 9.3 trillion liters a year.
Annual data center water consumption, in trillions of liters
Source: UNU Institute for Water, Environment and Health, June 2026. 9.3 trillion liters is roughly 2.5 trillion US gallons.
Scale is hard to feel at that magnitude, so the report offers a comparison. The projected 2030 water footprint is equivalent to the basic annual domestic water needs of all 1.3 billion people living in Sub Saharan Africa. The 2025 figure already covers the needs of more than 600 million.
Worth noting for accuracy: AI workloads are not the whole story. They currently account for roughly a fifth of data center water use, with the rest going to the ordinary cloud services that have been quietly expanding for two decades. AI is the accelerant, not the entire fire.
Why “low carbon” does not mean “low impact”
The most useful line in the UN report is a warning about how tech companies talk about sustainability. Most publish carbon numbers. Very few publish water numbers with the same granularity, and almost none publish them per facility.
From the UN report
“Low carbon is not automatically low water or low land, and evaluating sustainability through a single metric can hide trade offs and shift burdens onto places already facing water stress or land pressure.”
That is the crux of the local fight. A data center powered by a solar farm still needs cooling, and cooling still needs water, and the water has to come from somewhere specific. Facilities cannot draw from just any source either. Cooling systems generally need treated, low mineral water, which in practice often means the same municipal supply that residents drink.
There is also an indirect footprint that rarely appears in headlines. Electricity generation itself consumes water, particularly at thermal plants. A facility that reports modest on site water use can still carry a substantial water cost upstream at the power station keeping it running.
Where the water actually goes
| Draw | What it does | Usually disclosed? |
|---|---|---|
| Evaporative cooling | Water evaporates to shed heat from server halls. Cheap on power, heavy on water. | Sometimes, at company level |
| Closed loop liquid cooling | Recirculates coolant. Far less water, more electricity and higher build cost. | Rarely broken out |
| Power generation upstream | Water consumed producing the electricity the site uses. | Almost never |
| Chip manufacturing | Ultrapure water used in fabrication, well before a server is installed. | No, sits with the supplier |
The disclosure gap is the reason local opposition keeps winning arguments it technically should not win. When a company will not say how much water a specific building will draw from a specific aquifer, residents fill the blank with the worst case. That dynamic is exactly why state legislatures rather than Washington are ending up as the venue for the data center water fight, since states control the permits, the utility commissions and the water rights.
The US picture specifically
There are more than 5,000 operational data centers in the United States, with a heavy pipeline of new construction still coming. Research published earlier this year projected that US facilities alone could require as much water as a city the size of New York by the end of the decade, a framing we covered when that study first landed.
The geographic problem compounds it. Cheap land and cheap power tend to be found in the same places as thin water tables. Phoenix, west Texas and parts of Georgia have all seen large projects proposed in areas already managing drought conditions. National averages hide that entirely.
The industry’s response has ranged from serious to speculative. Closed loop and air cooled designs are being deployed on new builds. Some operators are committing to water positive targets. And a handful of companies have decided the real fix is to stop building on Earth, which is the premise behind the orbital data center pitch. That one is going to take a while.
So should you mail anything anywhere?
No. The ad says so itself, in writing, during a scene where a crowd carries jars of urine into a post office. The Coolant Collector is a glass mug, it has already sold out, and nothing about the campaign is a functioning proposal for industrial cooling.
What it is, is a temperature reading. When a beer brand and a canned water company decide that mocking AI infrastructure is the safest possible marketing bet, the industry has an audience problem that no amount of carbon reporting is going to fix. Kelce is not making an argument here. He is confirming that one has already been won in the court of public opinion.
What to watch next
- Site level water disclosure. A handful of states are weighing rules that would require per facility reporting. That single change would end most of the guesswork.
- Cooling design in new builds. Watch how quickly closed loop becomes the default rather than the premium option.
- Whether AI’s share grows past 20 percent. The UN figure assumes current workload mixes. Inference demand is not holding still.
- More brands piling on. Marketing follows sentiment. If this campaign performs, it will not be the last one.
Somewhere in Nevada there is a facility that will use more water this week than a small town, and the most effective public communication about it in 2026 turned out to be a retired offensive lineman singing about pee. Make of that what you will.
