AI's Carbon Footprint Becomes Harder to Ignore
A story about data centers driving big tech's carbon emissions to a third of France's total landed on the front page, and the comments weren't dismissive. One commenter added that Irish data centers now consume 23% of the country's electricity. These are no longer hypothetical future numbers, they're current operational figures.
The thread had the usual range of deflections, comparisons to automotive transport, agricultural machinery, and so on. But the underlying math is stark: AI inference and training workloads are scaling faster than renewable capacity can absorb them, and the geographic concentration in places like Ireland means local grids are getting hit hard.
This sits alongside the CoreWeave/$35B capex story. The buildout is real, the power consumption is real, and the regulatory risk in the EU in particular is real. Ireland has already started restricting new data center approvals in Dublin.
So what?
If you're building infrastructure-heavy AI products and selling into Europe, power availability and carbon reporting are becoming procurement requirements, not just PR considerations. Get ahead of this by understanding your inference provider's energy sourcing now, before a customer's procurement team asks.