Maybe We Should Stop Building Data Centers For 5 Minutes
[full disclosure, I'm a tree hugger]
Every time somebody raises concerns about data centers, the conversation immediately jumps to AI.
Will AI take our jobs?
Will AI become sentient?
Will AI write poetry?
Will AI cure cancer?
Maybe.
But I think we’re asking the wrong question.
The question I’m interested in is much simpler:
How much farmland, water, energy, & public infrastructure should private companies be allowed to consume chasing growth projections that may or may not pan out?
Because here’s the thing nobody seems eager to discuss.
Technology gets better.
Every cycle.
Computers get smaller.
Storage gets denser.
Processing gets faster.
Models get more efficient.
The same thing that required an entire room full of computers decades ago now fits in your pocket.
Yet we’re being told we need more & more land for bigger & bigger data centers.
Maybe that’s true.
Or maybe we’re watching another industry convince itself that exponential growth will continue forever.
We’ve seen this movie before.
Railroads.
Telecom.
The dot-com boom.
Housing.
Office parks.
Shopping malls.
Every generation has a sector that insists it must build at breakneck speed or risk catastrophe.
Then reality shows up.
Demand slows.
Technology changes.
The market consolidates.
And suddenly there are empty buildings nobody quite knows what to do with.
The assumption behind the current AI buildout is that demand will continue growing faster than efficiency improves.
Maybe it will.
Maybe it won’t.
What I know for certain is that once farmland is converted, paved over, connected to utility corridors, & integrated into industrial infrastructure, putting it back isn’t nearly as easy as the developers make it sound.
Land isn’t a spreadsheet.
You don’t just hit Undo.
Which brings me to a question that should be asked every single time a new data center is proposed:
What happens if you’re wrong?
Not wrong about AI.
Wrong about demand.
Wrong about growth.
Wrong about how much infrastructure you’ll actually need ten years from now.
Because communities are constantly being asked to absorb the risk while private investors collect the reward.
The company gets tax incentives.
The company gets infrastructure upgrades.
The company gets the upside.
But if the projections don’t materialize?
The community gets the abandoned facility.
The community gets the altered landscape.
The community gets the long-term consequences.
Seems like a bad deal.
Personally, I don’t understand why data centers shouldn’t face some of the same requirements imposed on other industries that have significant environmental impacts.
Want to build?
Fine.
Post a rehabilitation bond.
Set aside restoration funds.
Demonstrate long-term need.
Show your work.
If you’re confident your project makes sense, none of that should be a problem.
And if those requirements suddenly make the project financially unviable, maybe the project wasn’t as inevitable as we were told. [that last bit I said in my loud voice in my head]
What frustrates me most is the assumption that growth itself is always good.
That if a company wants to expand, our only responsibility is figuring out how quickly we can accommodate it.
Why?
Why is “grow as fast as possible” treated like a law of nature?
Communities have priorities too.
Food production.
Water security.
Energy reliability.
Open space.
Long-term sustainability.
A company has a fiduciary duty to investors.
A community has a duty to the people who live there.
Those aren’t the same thing.
And while we’re on the subject of AI, let’s talk about jobs for a second.
Everyone is busy arguing over whether AI will replace workers.
I’m more interested in what happens if it does.
Because if automation allows companies to produce more value with fewer workers, eventually we have to confront an uncomfortable question:
Who gets the benefit?
Shouldn’t that mean the prices all drop?
The economy still needs customers.
People still need purchasing power.
Companies still need someone to buy their products.
If technology allows a handful of corporations to automate enormous portions of economic activity, then maybe the conversation shouldn’t be “How do we stop AI?”
Maybe the conversation should be “How do we ensure the benefits of automation are broadly shared?”
Instead, we’re acting like the only acceptable answer is endless expansion.
More servers.
More power consumption.
More land conversion.
More water use.
More growth.
Faster.
Always faster.
Maybe that’s wisdom.
Or maybe it’s another bubble convincing itself it’s different this time.
I don’t know whether AI demand will justify every data center currently being proposed.
No one does, not the executives, investors & analysts, they’re making educated guesses.
They’re making educated guesses.
What I do know is that when the forecast is uncertain, caution isn’t anti-technology.
It’s just common sense.
Maybe before we pave over another chunk of farmland based on projections stretching decades into the future, we should slow down long enough to ask a basic question:
What if we’re building too much?
And if the answer is “we don’t know,” then perhaps patience is a better strategy than treating every growth forecast like prophecy.
And if it involves cutting down a single Redwood tree, I am 100% against it.
