Back to News
IndustryAI Understanding briefing

Google AI data centers partner with grids using rack‑level batteries and 800 VDC power

Google is converting backup battery assets into grid‑supportive resources, moving UPS functions to rack‑level 54 VDC batteries and planning 800 VDC distribution to boost efficiency and enable liquid‑cooling for gigawatt‑scale AI campuses.

4 min readRead the linked source
Source-provided image accompanying Google AI data centers partner with grids using rack‑level batteries and 800 VDC power
Source referenceSource recorded
Publisher
datacenters.economictimes.indiatimes.com
Source link
datacenters.economictimes.indiatimes.comhttps://datacenters.economictimes.indiatimes.com/news/ai-compute-infrastructure/google-ai-data-centers-become-grid-partners-with-advanced-power-cooling/134495767
Source type
Linked source — primary-source status has not been established.
ContextUnderstand this in 60 seconds

Start here

Key terms

Compute
The processing resources required to train and run models, often measured in FLOPS or GPU hours.
Test yourselfFuture of AI Quiz

What happened

Google announced at Data Center World Power 2026 that its AI data centers will act as active grid partners, repurposing battery energy storage systems for peak‑shaving, shifting UPS functions to rack‑level 54 VDC batteries, and targeting 800 VDC power distribution within racks. The company also detailed a move toward advanced liquid‑cooling and workload‑shifting across its global sites to manage power demand during heat‑wave events.

At the Data Center World Power 2026 conference, Tom Garvens, Google’s VP of advanced technology innovation, described a full‑stack redesign of AI data‑center power and cooling. Backup assets such as battery energy storage systems (BESS) will be repurposed to provide grid‑supportive services, smoothing peak demand and reducing power fluctuations.

Google is replacing traditional centralized UPS units with rack‑level battery backups operating at 54 VDC. This eliminates an AC‑DC‑AC conversion step, delivering a 1.5 % efficiency gain. The company plans a gradual migration to 800 VDC distribution inside racks, supporting future rack densities near 1 MW and reducing current requirements.

To manage the higher heat output from denser racks, Google is deploying advanced liquid‑cooling solutions and closed‑loop cooling loops that minimize water use. Designs will be adapted to local conditions such as altitude, humidity, and drought risk.

Workload shifting across Google’s global data‑center portfolio will be used to balance power demand during regional peak events, like heat waves, further reducing strain on local grids.

Source details: datacenters.economictimes.indiatimes.com ↗

Why it matters

The shift reduces conversion losses, improves overall data‑center efficiency by about 1.5 %, and enables higher rack densities approaching 1 MW, which is critical for the massive needs of modern AI models. By integrating battery storage directly with the grid, Google can help stabilize local power networks while lowering its own energy costs and carbon footprint. The approach also sets new standards for the industry through Open Compute Project specifications, potentially influencing other hyperscale operators.

The efficiency gains from eliminating AC‑DC‑AC conversion and moving to higher voltage distribution directly lower operational costs and carbon emissions, addressing sustainability concerns for AI that is otherwise energy‑intensive.

By acting as a grid partner, Google can provide ancillary services such as frequency regulation and peak shaving, which can improve grid reliability and enable greater integration of intermittent renewable sources.

Standardising these designs through the Open Project creates a reusable blueprint for the industry, potentially accelerating the deployment of more energy‑efficient AI data centers worldwide.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Model Parameter Size:8B Parameters
VRAM Required5.5 GBGPU memory footprint
Target HardwareMacBook / Single GPUDeployment tier
Privacy100% Air-GappedLocal device capability
Core takeaway: Small, quantized models (3B–8B) now run directly inside smartphones and laptops with complete data privacy, while mammoth 400B+ models remain the domain of datacenter clusters.
Interactive Concept Check+10 Points
Future of AI Quiz

What did neural scaling law research (e.g. Kaplan et al., 2020) observe?

What to watch next

Future monitoring should focus on Google’s rollout of 800 VDC distribution, adoption rates of rack‑level battery backup, and the performance of closed‑loop liquid‑cooling systems in varied climates. Regulators and utilities will watch how Google’s grid‑interaction model affects regional grid stability and renewable integration. Competitors may adopt similar designs, and OCP’s evolving specifications will indicate broader industry uptake.

Implementation timelines for 800 VDC power distribution and the scale‑up of rack‑level battery systems across Google’s AI campuses.

Performance data from the new liquid‑cooling loops, especially in regions with water scarcity, to assess real‑world water‑use reductions.

Regulatory responses and utility partnership agreements that may arise as Google’s grid‑interactive model expands.

Adoption of Google’s OCP specifications by other hyperscale providers, indicating broader industry impact.

Related guides & quizzes

Future of AIAI TrainingAI EthicsTest what you know — try a free AI quizLook up an AI term in our glossaryFollow the AI funding tracker
Found this useful?