The Cloud Has a Water Bill: What Data Centers Mean for Washington

We call it "the cloud," which makes it sound weightless. Like your photos and emails and AI chats float somewhere in the sky.

They don't. They live in buildings. Big ones. Full of servers that run hot, all day, every day, and need enormous amounts of electricity and, often, water to stay cool.

Washington has a lot of those buildings. And the AI boom is bringing more.

How big is this in Washington?

According to comments filed with the state's Utilities and Transportation Commission by Climate Solutions, Washington is home to 127 data centers estimated to use roughly 1,500 megawatts of electricity, about six percent of statewide demand (UTC docket filing).

And the growth ahead is the real story. A state Data Center Workgroup found that data centers are the largest source of expected electricity load growth in the Pacific Northwest. The Northwest Power and Conservation Council has projected that data centers and chip fabrication could add 2,200 to 4,800 average megawatts of load by 2030, as summarized by the Municipal Research and Services Center (MRSC). Climate Solutions describes that as enough new electricity to power two to four Seattle-sized cities (Climate Solutions).

Why AI makes this bigger

Here's the technical part, simplified.

Training a large AI model means running huge numbers of calculations across thousands of specialized chips for weeks or months. Then, every time someone uses the model, that's "inference," more computation. Those chips draw a lot of power and pack it densely, which creates a lot of heat in a small space.

Heat is the enemy of servers. So data centers cool them, often with:

• Air cooling with big fans and chillers, which uses more electricity.
• Evaporative cooling, which uses water. In evaporative systems, water is lost to the air rather than returned to the local watershed (Climate Solutions).
• Liquid cooling that brings coolant right to the chips, increasingly common for AI hardware.

The trade-off is real: designs that save water can use more power, and vice versa. That's why transparency on both matters.

For a sense of scale, the International Energy Agency estimates that a 100-megawatt data center may consume about the same amount of water as 2,600 households, according to Sen. Dick Durbin's office. Many proposed data centers are much larger than that.

The benefits are real too

I'm a data engineer. I use the cloud every day. My work depends on it. So let me be fair.

Data centers bring construction jobs and significant tax revenue, especially to rural communities, which Washington's own legislation acknowledges (SB 6171 text). And consolidating computing into large, modern facilities has improved energy efficiency compared to every company running its own server closet (MRSC).

The question isn't whether we have data centers. It's who pays for the infrastructure they need, and whether communities get to see the numbers.

What Washington tried in 2026

Lawmakers took a real swing this year. House Bill 2515 would have required utilities to create tariffs so large energy users pay the full cost of serving them, and required data centers to report their energy and water use. Its sponsor, Rep. Beth Doglio, said the policies "seek to protect ratepayers by ensuring new data centers are picking up the whole tab for new growth" (OPB).

The bill was weakened along the way. A proposed per-kilowatt-hour fee projected to raise over $30 million a year was cut, and clean-energy deadlines were pushed back (OPB). The Seattle Times reported that tech industry lobbyists successfully weakened many parts of it (Seattle Times).

In the end, HB 2515 did not pass. MRSC expects the debate to return in the 2027 session. One related change did happen: under SB 6231, a sales tax exemption no longer applies to replacement and refurbished data center equipment (MRSC).

Why this is a community issue

When a utility has to build new substations and transmission lines, someone pays. If the rules don't assign those costs to the customers who caused them, they can end up spread across everyone's bills. The state's workgroup flagged that large new loads "may present risks for other utility customers in the form of higher rates" (MRSC).

That's a digital equity issue, even if it doesn't look like one. The family deciding between the internet bill and the light bill shouldn't be subsidizing the infrastructure behind someone else's AI product.

Questions every community should ask

If a data center is proposed near you, or your utility is planning for one, here's what I'd ask:

1. Water: How much water will it use at peak, from what source, and what cooling technology? Is any water returned to the watershed?
2. Power: What's the expected load in megawatts? Who pays for new substations and transmission?
3. Rates: Is there a separate tariff for large loads, so costs aren't shifted to households?
4. Grid stress: During a heat wave or cold snap, will the facility reduce load, so hospitals and homes aren't put at risk?
5. Transparency: Will the operator publish annual energy, water and refrigerant reports?
6. Jobs: How many permanent jobs, not just construction jobs, and will local residents get trained for them?

The bigger point

Every AI prompt, every streamed video, every "save to cloud" runs somewhere physical, on real power, sometimes with real water. That doesn't mean we stop using technology. It means we stop pretending it's free.

Pour me another: Would you want to know how much water and power your AI tools use? Tell me through the contact form.

Sip slow, stay curious.D