We Didn't Have an Answer. That's a Problem.

At our recent all-company AI session, one of our dev team asked a question that stopped me in my tracks.

Dave, now 16 years at Vibe (congratulations, and genuinely, thank you), raised his hand and asked: what's our position on the environmental impact of AI?

It blindsided me. I didn't have a good answer. That's not a feeling I'm used to.

The gap we missed

We've done rigorous legislative validation of AI across the business. We've thought carefully about model selection: using lighter models where the use-case allows, restricting AI in production to where it genuinely earns its place, not burning inference on tasks that traditional automation handles better.

But we hadn't assessed our AI suppliers on their environmental footprint. That's an oversight. And it's out of character.

We're not starting from zero

Our datacentre power is green. Our office energy suppliers are green. We're conscious about recycling, about choosing public transport for meetings. The culture is there.

But AI puts new pressure on the question.

The problem is upstream

The rare earth metals mined for IT hardware. The vast new datacentres being built at scale. The power stations required to run them, not all of which are renewable. The water. The heat. The supply chains.

Optimising our own inference is necessary. It's not sufficient.

So Dave, here's my honest answer

I don't have a complete solution yet. But your question has stayed with me, and that's exactly the point. These are the questions we all need to be pushing, internally and externally, on our suppliers, on our choices, on the industry.

There's a more optimistic frame too. AI is already accelerating investment in fusion energy. Clean, limitless power. It may yet solve the very problem it's creating.

But that's a long game. In the short term, we act on what we can control, and we keep asking the uncomfortable questions.

Thanks for asking yours.