Google AI buildout pushes electricity use up 37% in 2025

Google’s electricity consumption jumped 37 percent in 2025, its largest year‑on‑year increase, as AI workloads drove a surge in data‑center power use【2】. The rise threatens Google’s climate goals, forcing the company to accelerate clean‑energy procurement while operational emissions fell 2 percent year over year.

AI demand and supply‑chain impact

Google attributes the electricity surge primarily to AI‑driven compute growth, noting that total electricity consumption is now more than 250 % of its 2019 level【2】. While operational emissions dropped 2 % thanks to efficiency gains, Scope 3 emissions—dominated by data‑center construction and semiconductor procurement—rose 25 % year over year, adding roughly 2.3 million t CO₂e in 2025【2】. The company points to suppliers in Taiwan, Japan, Vietnam and India, where power grids remain fossil‑fuel heavy, as a key factor in the emissions increase【2】.

Clean‑energy response and efficiency gains

To offset the higher demand, Google signed power‑supply agreements for more than 12 GW of net‑new clean energy—the largest annual total in its history—covering projects such as a 600 MW nuclear restart in Iowa, a 3 GW hydropower framework across PJM and MISO, and a fusion offtake with Commonwealth Fusion Systems【2】. The firm also integrated 1 GW of demand‑response capacity into long‑term utility contracts, allowing it to shift machine‑learning workloads during grid stress periods【2】. Data‑center efficiency improved, with an average power usage effectiveness (PUE) of 1.09, and the energy footprint of a median Gemini text prompt fell 33‑fold over the previous year【2】.

Comparative context with peers

Google’s supply‑chain emissions growth mirrors a broader industry trend; Amazon reported a 16 % rise in total carbon footprint for 2025, driven largely by similar AI‑related data‑center expansion and a 34 % jump in electricity‑linked emissions【1】. Both firms matched 100 % of their operational electricity with renewable purchases, but the rapid AI buildout outpaces grid decarbonisation, highlighting a sector‑wide challenge of reconciling compute growth with climate commitments【2】.

What to watch

• Clean‑energy procurement milestones – Google’s next annual clean‑energy signing target and the commissioning dates of the Iowa nuclear plant and the 3 GW hydropower framework.
• Supply‑chain emissions trends – Quarterly updates on Scope 3 emissions, especially from semiconductor suppliers in high‑carbon regions.
• AI workload scheduling – Deployment of demand‑response contracts and any reported curtailment of machine‑learning jobs during peak grid stress periods.

The 37 % electricity surge underscores the tension between AI‑driven compute expansion and corporate climate pledges, raising questions about whether clean‑energy sourcing and supply‑chain reforms can keep pace with the accelerating demand for AI infrastructure.