Google Saint-Ghislain: Grid-Interactive Battery Backup at Hyperscale
How Google proved that a data center battery doesn't have to sit idle waiting for disaster — and what that means for every AI infrastructure project that follows.
LegacyGrid AI Infrastructure Advisory
GOOGLE SAINT-GHISLAIN — SAINT-GHISLAIN / MONS, BELGIUM — FLUENCE GRIDSTACK BESS
LG-CS-001
Google Saint-Ghislain BESS
Saint-Ghislain / Mons, Belgium
Fluence (Gridstack)
Centrica (FlexPond)
Conceptual reconstruction — not Google engineering drawings
Background: Google's First European Data Center
Google began investing in Saint-Ghislain in 2007, making it the company's first data center outside the United States. The Belgian facility has since grown into one of Google's major European hyperscale campuses, with Google reporting more than €5 billion invested in Belgium in data centers and associated infrastructure as of 2025. In October 2025, Google announced another €5 billion investment for 2026–2027, including expansion of Saint-Ghislain to support increasing Google Cloud and AI demand and approximately 300 additional full-time jobs.
The campus is not simply a large building full of servers. It has become a collection of integrated efficiency technologies — each addressing a different dimension of the data center energy problem. The BESS installation is one layer of that system, not the whole story.
The System: What Google Actually Built
| COMPONENT | VERIFIED SPECIFICATION |
|---|---|
| Owner / Operator | |
| Location | Saint-Ghislain, Belgium |
| Facility Type | Hyperscale data center — Google's first outside the U.S. |
| BESS Supplier | Fluence |
| BESS Product | Fluence Gridstack |
| Battery Power | 2.75 MW |
| Stored Energy | 5.5 MWh (approx. 2-hour duration at rated output) |
| Grid / Optimization Partner | Centrica Business Solutions |
| Control Platform | Centrica FlexPond |
| Primary Purpose | Zero-emission backup power during grid outages |
| Secondary Purpose | Belgian grid frequency regulation and ancillary services |
| Diesel Strategy | Replace / reduce a portion of generator capacity (proof-of-concept) |
| Battery Chemistry | Lithium-ion |
| On-Site Solar | 2.8 MW — 10,665 ground-mounted panels — ~2.9 GWh annual generation |
| Cooling Architecture | Evaporative cooling — no conventional mechanical chillers |
| Water Source | Grey water from nearby industrial canal — treated on site |
| PUE (2017) | 1.09 — EU Code of Conduct Award winner 2018 |
| BESS Announcement | December 2020 (Google) / April 21, 2022 (Fluence + Centrica) |
Why Google Built It: The Stranded Asset Problem
Hyperscale data centers need backup electricity capable of coming online extremely quickly when utility power disappears. Historically, the industry solved that problem with diesel generators. But those generators spend almost their entire lives doing nothing — idling, waiting for a grid failure that may never come during their operational life. They are expensive stranded assets that also happen to produce emissions when they do run.
Google's question at Saint-Ghislain was straightforward: what if the backup-power asset could perform useful work when there is no outage? Google wrote when announcing the project that batteries could both protect the data center during an outage and strengthen the broader electrical grid when emergency backup was not needed. That creates two distinct operating roles from a single piece of infrastructure.
"Backup generation does not have to be a stranded asset. A battery can serve resilience, grid services, and emissions reduction from the same infrastructure." — LegacyGrid AI analysis of the Saint-Ghislain proof-of-concept
Two Operating Modes: Resilience and Grid Services
The Saint-Ghislain BESS operates in two distinct modes depending on grid conditions.
Mode 1 — Data Center Resilience: When the Belgian grid fails, the 2.75 MW / 5.5 MWh battery provides zero-emission electricity to critical data center loads. At rated output, the system provides approximately two hours of backup capacity — meaningfully longer than a UPS bridge designed only to cover the seconds until a generator starts. This is the primary design intent.
Mode 2 — Grid Asset: When there is no emergency, Centrica's FlexPond platform connects the batteries to the Belgian electricity market. The system can charge and discharge in response to grid frequency regulation requirements and ancillary service market signals. Reporting on the project states that up to half of the 5.5 MWh battery capacity could be made available to the grid as needed. Fluence now cites Saint-Ghislain as an example of storage integrated into a virtual power plant and participating in ancillary-service markets.
GOOGLE SAINT-GHISLAIN — AERIAL VIEW — BESS INSTALLATION AND SOLAR FARM (CONCEPTUAL RECONSTRUCTION BASED ON PUBLICLY DISCLOSED SYSTEM ARCHITECTURE — NOT GOOGLE ENGINEERING DRAWINGS)
The Full Integrated Energy System
The BESS is one layer of a broader integrated energy system at Saint-Ghislain. Understanding the full architecture is important because it shows that no single technology solved the efficiency and resilience problem — the combination of technologies is what makes the campus exceptional.
Project Chronology
FLUENCE GRIDSTACK — THE BESS PRODUCT DEPLOYED AT GOOGLE SAINT-GHISLAIN (IMAGE: FLUENCE ENERGY)
What This Case Study Proves — and What It Does Not
- Backup generation does not have to be a stranded asset
- A single BESS can serve resilience, grid services, and emissions reduction simultaneously
- Data centers can be active participants in the energy system, not passive consumers
- Grid + generation + BESS + controls can function as an integrated system
- On-site solar co-located with a data center is technically feasible at hyperscale
- Chillerless data center operation is achievable with advanced cooling design
- Non-potable water sources can reduce data center freshwater consumption
- Google itself called Saint-Ghislain "an alternative backup power model for all data centres"
- That a 2.75 MW battery can power a hyperscale data center indefinitely
- That BESS eliminates the need for all backup generation at every facility
- That the exact 2.75 MW / 5.5 MWh design should be copied into another project
- That BESS sizing, duration, and grid-service strategy are universal — they must be engineered for each site and utility environment
- That the Belgian grid-services model translates directly to ERCOT or other U.S. markets without modification
Primary Sources and Evidence Base
LegacyGrid AI's analysis of this case study is based exclusively on publicly available primary sources. The following sources were used and are recommended for independent verification:
- Google — Saint-Ghislain official facility page: Google identifies Saint-Ghislain as its first battery-based hyperscale generator-replacement system and describes the BESS as "an alternative backup power model for all data centres."
- Google — Original battery announcement (December 2020): "Cleaner data centers, batteries included" — establishes Google's original intent and dual-mode operating concept.
- Fluence + Centrica deployment announcement (April 21, 2022): The primary technical source confirming the 2.75 MW Gridstack installation and FlexPond grid integration.
- Data Center Dynamics: The clearest published source for the 5.5 MWh figure and the portion of capacity potentially available for grid support.
- Google Belgium solar and efficiency case study: Documents the 2.8 MW solar installation, 10,665 panels, 2.9 GWh annual generation estimate, PUE, industrial-water cooling, and chillerless design.
Note: LegacyGrid AI has not obtained or reviewed Google engineering drawings, one-line electrical diagrams, BESS site plans, battery cell chemistry specifications, inverter specifications, transformer ratings, protection schemes, fire-suppression drawings, or interconnection agreements. Diagrams in this case study are conceptual reconstructions based on publicly disclosed system architecture and are not Google engineering documents.
Mapping to LegacyGrid AI Methodology
The Saint-Ghislain case study maps directly to LegacyGrid AI's five-layer AI Infrastructure Systems Engineering methodology. It is the most complete publicly documented example of what responsible AI infrastructure integration looks like at hyperscale.