With the news that Seattle was the proposed location for five large (“hyperscale”) data centers, we were presented with an urgent question: Under what conditions, if any, would this be an acceptable proposition?
In the face of powerful opposition (including over 80,000 letters sent to the City Council and the mayor), two companies have dropped out — but the question is the same.
The argument that data centers are economic development drivers has been disproved: A study that compared job creation claims of developers with data gathered from centers indicates that $33 million of investment is necessary to create a single job. And that doesn’t take into account the fact that artificial intelligence companies are telling us AI will cause massive job loss in every industry.
If we treat these like value-neutral business widgets, what’s our baseline for ensuring that this makes sense for Seattle?
The proposed data centers are very, very large: their power use would be many times the size of the largest Seattle energy user — the data center in the Westin Building downtown. Data centers can also use as much as 5 million gallons of water per day, they require a lot of land, and produce noise pollution. Power, water, land, ambient sound: those are our most basic resources.
Seattle City Light cannot provide clean power for these while serving existing residential and commercial customers well; it’s already under strain from drought. Puget Sound Energy might be able to do so, but that would involve major new pollution (perhaps a new gas “peaker plant” in King County and a substation bringing the power to the city) when we need to reduce fossil fuel use.
Other options frequently mentioned for data centers, like fuel cells, small modular reactors and geothermal, are similarly dirty, dangerous and/or impractical for an urban location. Better utility-scale options, like solar or wind, are also somewhat impractical. So could we even find the necessary power?
Maybe yes. One option is to require these hyperscale center builders to share the load, making use of centers’ need for power capacity, not necessarily generation. Instead of buying turbines or paying for new utility infrastructure, hyperscalers would pay for things like heat pumps and weatherization for homes, and for rooftop solar and batteries for homes and government buildings, and maybe even to place them in large parking lots, as France requires; this has the added benefit of reducing heat islands.
Because all of this greatly reduces the existing City Light load while increasing distributed generation (which also adds resilience), City Light would have enough capacity left over for data centers. In other words, your new heat pump and solar panel means you’re using (and paying) less, so someone else can use more.
Best of all, because the hyperscalers would be handing the city a large sum of money to implement this — they don’t want to manage your heat pump installation — we’d see enormous economies of scale, and we could ensure that all of this work was unionized.
If feasible, it will address pollution, affordability and grid stability concerns and greatly increase work for the trades. We would still need to assess solutions for protecting our water supply and ensuring that centers’ wastewater is properly treated. We’d also need to protect neighboring communities from noise impacts, and decide if we can afford to lose land to this purpose. Finally, we’d need to decide if we want to participate in building this technology at all.
I was an early adopter of everything internet, and I know machine learning has great promise. Nevertheless, I hate data centers because they represent everything that’s wrong with modern tech: consolidation, haste and profit at any cost. They are not an inevitable future — the future is likely to be distributed, resilient, and sustainable — and we’ll be sorry if we let tech companies roll over us on their current juggernaut.
So why even consider these data centers? Because one or two of them will almost certainly be built, and we don’t want a pristine Seattle in a sea of fossil-fueled, noisy, paved-over rural communities: that’s pure dystopia. As a place with a people-friendly government and a proud tradition of renewable energy, we’re well-placed to show that it can be done differently.
First, we have to figure out if and precisely how it can be done differently. If it can, we make AI developers do it. If it can’t, well — as communities across the country have proved, the option of a simple “no” is still very much on the table.
