Two unrelated headlines. One uncomfortable truth about AI's future.

Two news stories landed recently. On the surface they appear unrelated. But connect them and a third story emerges — one that will reshape an entire industry.

India just recorded its highest-ever total energy consumption. And El Niño is projected to peak next year, bringing drought, heat stress and destabilised power infrastructure across large parts of the world.

The chain most people haven't connected

Data centres already consume 1–2% of global electricity, and that figure is accelerating sharply with generative AI. Cooling alone accounts for roughly 40% of a data centre's energy use — and most cooling systems are water-intensive.

El Niño doesn't just mean heat. It means scarcity — of water, of stable grids, of predictable supply chains for the energy sector. India is building out its AI ambitions at exactly the wrong climatological moment. And India is not alone.

The infrastructure powering AI may itself become the resource crisis it was meant to solve.

A natural selection event for AI architectures

What we are watching is evolutionary pressure — quiet, systemic and largely invisible until it isn't. The models that survive the next decade won't simply be the most capable. They will be the ones that operate inside tighter resource envelopes.

This isn't a prediction. It's physics meeting policy meeting climate. For leaders planning AI investments, efficiency is no longer an engineering detail — it is strategy.