Data Center Power and Cooling Trends for the AI Era

2026-06-18 · Source: TechRepublic · Field: Technology & Digital — Cloud Computing & IT Infrastructure, Artificial Intelligence & Machine Learning · Depth: Intermediate, short

Summary

Omdia's 2026 report highlights critical trends reshaping data center power and cooling systems to support AI applications, driven by soaring power densities and gigawatt-scale expansion. Nvidia's Rubin platform exemplifies this shift, designed for 100% liquid cooling across all key components, including GPUs, CPUs, switches, and optical modules. Battery energy storage systems (BESS) are now essential for managing millisecond-level power surges from massive GPU clusters, often supplemented by supercapacitors for stability. Furthermore, high-voltage direct current (HVDC) is emerging as a necessity, with an 800V DC power backbone reducing copper costs, space, and energy losses by eliminating complex AC-DC conversion. This transition, expected to be gradual from AC to 400V and then 800V DC, will be aided by solid-state transformers (SSTs), with hyperscalers leading initial deployments and proof of concepts in 2026-27.

Key takeaway

For data center operators planning AI infrastructure, you must prioritize fundamental redesigns in power and cooling. Your existing AC systems and air cooling will likely prove insufficient for high-density GPU clusters. You should transition critical areas to 100% liquid cooling and integrate battery storage with supercapacitors to manage millisecond power surges. Begin exploring high-voltage DC architectures and solid-state transformers, as these will be crucial for future gigawatt-scale campuses. Failure to modernize risks severe performance constraints.

Key insights

AI's extreme power demands necessitate fundamental redesigns in data center cooling and power delivery.

Principles

• Advanced AI infrastructure requires 100% liquid cooling.
• GPU clusters demand large-scale on-site energy for power surges.
• HVDC architectures create high-voltage foundations for SST adoption.

In practice

• Transition high-density areas to liquid cooling.
• Integrate BESS and supercapacitors for power stability.
• Explore HVDC architectures and solid-state transformers.

Topics

• Data Center Infrastructure
• AI Power Management
• Liquid Cooling
• Battery Energy Storage
• High-Voltage DC
• Solid-State Transformers

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Architect, MLOps Engineer, IT Professional

Related on AIssential

• AI, energy and infrastructure: The future of data centre infrastructure in a high-demand world — Balancing escalating AI-driven compute demand with stringent sustainability goals requires advanced cooling, renewable energy, and intelligent optimization.
• Neutralizing the Gigascale Problem: How to Solve the Physical Power Paradox of Extreme AI Training Loads — Gigascale AI demands active, high-speed energy storage to stabilize power fluctuations, moving beyond passive backup.
• Gartner: Data Centres Face 26% Energy Spike From AI Servers — Rapid AI server growth is creating an unprecedented energy deficit, making power availability a critical constraint for data center expansion.
• Liquid-cooled AI systems expose the limits of traditional storage architecture — Hybrid air/liquid cooling for AI infrastructure is inefficient and creates thermal liabilities for storage.
• As AI data centers hit power limits, Peak XV backs Indian startup C2i to fix the bottleneck — Efficient power delivery is critical for scaling AI data centers and reducing operational costs.
• Measurement of Generative AI Workload Power Profiles for Whole-Facility Data Center Infrastructure Planning — High-resolution AI workload power profiles enable accurate whole-facility data center energy planning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by TechRepublic.