Why does AI increase electricity use even as chips get efficient?

Because demand and model scale grow faster than per‑unit efficiency gains, total energy use still rises.

What can reduce AI data center power risk?

On‑site/near‑site clean generation, storage, and early interconnection planning.

AI Energy Demand: Brookfield Report Summary (2025)

Overview

Brookfield’s August 2025 report explains why AI factories—high‑density, GPU‑heavy data centers—are reshaping power markets. Even as chips get more efficient, total electricity consumption rises due to rapid scale and new AI workloads.

Key Points

Capacity growth: AI‑purpose facilities are projected to expand sharply this decade, requiring new generation and faster interconnections.
Grid bottlenecks: Interconnection queues and transmission constraints are the gating factors for new AI capacity.
On‑site & clean power: Behind‑the‑meter renewables, storage, and advanced nuclear can improve resilience and timelines.
Capital cycle: A multi‑trillion‑dollar investment wave spans compute, cooling, power, and adjacent infrastructure.
Policy & siting: Permitting reform and strategic siting are essential to meet deployment schedules.

Implications

For builders, design for power density, cooling, and modular upgrades. For policymakers, streamline permits and prioritize transmission upgrades. For investors, target grid‑constrained hubs, on‑site power, and interconnection‑advantaged sites.

References

Brookfield — Building the Backbone of AI (PDF)