The Data Center Conversation We Should Be Having

I was dating someone not long ago who was adamantly anti-AI. They were deeply worried about the environmental impact of data centers, and I want to be clear up front: I share that concern. The environmental stakes are real, and I don’t take them lightly. But I also carry something that I think is missing from a lot of these conversations: a sense of realism.

We’re stuck in a strange place right now. AI is simultaneously the big, bad bear in the room and the promise of the future, and we can’t seem to hold both ideas at once. The question I keep coming back to is this: how do we get to a place where data centers are part of the solution?

The Media Is Selling Hype, Not Solutions

Right now, the media is sensationalist. It grabs onto anything that generates new views. And to be fair, there’s a lot of valid concern underneath the noise, data centers genuinely do strain smaller communities and large ones alike. The environmental and resource impacts are real.

But what we’re seeing in coverage is less solution-oriented and more hype-oriented. The frame is simply “data centers bad.” And that frame, while emotionally satisfying, gets us nowhere. The real conversation isn’t should we build data centers? It’s how do we build them responsibly as compute scales?

The Reality We Actually Live In

Here’s the reality we have to start from: these giant, profit-focused corporations are not going to do less. AI is now woven into international politics and military strategy. Realistically, we are not going to see a reduction in AI technology, so the honest question is what we do with that fact.

So we have a choice. We can complain about it and protest it, or we can come together and work on solutions. And I think it’s far better for the planet’s future if we collaborate rather than stay combative.

We can all acknowledge that data centers create real environmental and resource problems. But we can also acknowledge how integral the internet, AI, and connected technology have become to daily life. The data center isn’t going anywhere, so the most useful thing we can let go of is the hype.

What “Future-Viable” Actually Looks Like

The energy and cooling demands are only going to grow. As we move deeper into quantum computing, the next generation of data centers will dwarf today’s. So we need to be designing them to be both future-viable and commercially viable.

There are real ideas on the table. Microsoft spent years exploring underwater data centers through Project Natick, and the results were genuinely promising, with only six of 855 submerged servers failing compared to a much higher failure rate on land. (Microsoft has since shelved the subsea effort, but it’s folding those lessons into other sustainability work.) And the space idea is no longer science fiction: in January 2026, SpaceX filed with the FCC for approval to launch up to a million satellites as a constellation of orbital data centers, powered by constant sunlight. Elon Musk’s pitch is essentially my argument in miniature: on Earth you’re power-constrained, but in orbit the sun never sets.

Not everyone buys it. Sam Altman has called orbital data centers “ridiculous” at any meaningful scale this decade, and plenty of engineers think the timelines are wildly optimistic. But Musk isn’t alone: Google, Amazon, and a startup called Starcloud (which already put an NVIDIA GPU in orbit in late 2025) are all chasing the same idea. The problem with nearly all of these solutions is cost, and a huge driver of that cost is energy.

That’s why I find the renewed interest in small modular nuclear reactors so compelling. A growing number of tech companies are reassessing nuclear technology, and I think it’s a genuinely viable path. My belief is simple: data centers should be self-generating when it comes to power. They should contribute to their environment rather than consume from it.

AI Isn’t the Only Animal in the Room

Here’s the reframe I most want people to sit with. Data centers are absolutely a concern. But of equal concern are the massive piles of discarded fast fashion, the oil slicks on our oceans, the floating garbage patches. We don’t just have a compute consumption problem, we have a consumption problem across agriculture, vehicles, fuel, everything. Human beings are consuming on every level.

AI is not uniquely evil in a world that already consumes this way. And here’s the part worth getting excited about: artificial intelligence may be one of the most powerful tools we have to take on these species-level problems. Used well, it isn’t just something to protect the planet from; it’s something that could help us heal some of the environmental damage we’re already facing.

So What Do You Bring to the Table?

If you’re reading this with genuine environmental anxiety, if you’re not hostile to AI, just worried, I don’t want to tell you to stop worrying. I want you to worry productively.

Ask yourself this: if we can’t go backwards, how do we go forwards in a positive direction? Solving the problem was never going to mean eliminating technology, or electricity, or data centers, or the human species. Those aren’t going anywhere. We have to find a way for humanity to rise up in a different form, in unity, through collaboration.

So the real question isn’t whether data centers should exist. It’s this: what can you bring to the table that helps us solve these problems within the scope of the reality we actually live in?