# PUE

Token Tsunami and Power Limits Propel a Boom in Liquid‑Cooled AI Servers

Exploding AI token consumption and rising hardware costs are driving a surge in rented AI compute and accelerating adoption of liquid‑cooled, high‑density servers. Policy limits on data‑centre energy efficiency and the shift from training to widespread inference are making immersion cooling and edge deployment central to scaling AI affordably and sustainably.