Data centres have always had onsite generation: standby diesel sets sized to carry the load while the grid is down. What is changing is the role of that generation. Increasingly it is the primary supply, with the grid — if there is one — as the backup.
Behind the meter, in plain terms
"The meter" is the boundary between the utility network and the customer's installation. Power generated behind the meter is produced on the customer's side of that boundary and consumed on site. It does not pass through the transmission or distribution network, so it does not need network capacity to reach the load.
That is why behind-the-meter supply is the natural answer to grid constraints: it avoids the part of the system that is congested.
From standby generation to a microgrid
A microgrid is a local energy system that can operate connected to the wider grid or independently of it ("islanded"). For a data centre, the building blocks are:
- Generation. Continuous-duty power blocks sized for the load plus redundancy.
- Storage. Batteries and UPS systems that hold the load during transfers and smooth fast changes in demand.
- Switchgear and protection. Medium-voltage switchboards that connect sources to the load and isolate faults.
- Controls. A microgrid controller that dispatches generation, manages storage and handles the transition between grid-connected and islanded operation.
- Grid interface (optional). A connection for backup supply or export.
Three design choices that matter
1. Redundancy level
The target availability drives how much spare capacity is built in. N+1 means one more generating unit than the load needs; 2N means a complete duplicate supply path. The right answer depends on the facility's own tier design and customer contracts. See What 99.999% availability really means.
2. The role of the grid
- No grid. The microgrid is the only supply. Redundancy has to be complete onsite.
- Grid as backup. Onsite generation is primary; the grid is an extra path during maintenance or faults.
- Grid as export. Surplus capacity — often available during early phases before the load has grown — can be sold.
3. How it grows
A data centre built in phases needs a power system built in phases. Modular generation blocks can be added as halls come online, with the switchgear and controls designed from the start for the final capacity.
Continuous-duty generation, not standby sets
Standby generators are designed to run for a few hundred hours a year. A primary supply runs around the clock. The equipment, fuel logistics, emissions controls and maintenance regime are all different. When evaluating a microgrid, check that the generation is rated for continuous duty and that the operator has a maintenance model for 8,000+ running hours a year.
A typical single-line arrangement
In simple terms, a data centre microgrid built around onsite generation looks like this:
- Generation blocks feed a medium-voltage switchboard. Each block can be isolated independently.
- The switchboard feeds the facility through redundant transformers and feeders — typically A and B paths, so each rack can be supplied from two sides.
- UPS and batteries sit between the supply and the IT load, holding it through any transfer.
- A microgrid controller dispatches generation to match the load, keeps spare capacity online and manages storage.
- An optional grid connection ties into the switchboard for backup or export.
The details — voltage levels, number of blocks, N+1 or 2N — follow from the facility's tier design and load.
Phasing: power the first hall first
Most data centres are built in phases. A sensible power plan mirrors that:
- Day one: enough generation for the first hall plus redundancy, with the switchboard, foundations and cable routes sized for the final capacity.
- Each subsequent phase: add generation blocks and commissioning, without a shutdown of the running halls.
- Final build-out: full capacity, with redundancy maintained throughout.
This avoids paying for — or waiting on — the full capacity before the first customer is live.
Questions for your design team
- What availability are we committing to customers, and at what point is it measured?
- What is the day-one load and the expected ramp?
- Do we need the grid at all, and if so, in what role?
- How much space is available for generation, and where can it sit relative to the halls?
- Who will operate and maintain the power system — and do we want that responsibility?
How CNP delivers it
CNP develops, finances, builds and operates modular onsite generation of 1–20 MW per site for data centres. Plants are co-located with the facility, designed with full backup and redundancy, and can run fully off-grid. CNP offers contractual availability of up to 99.999% at the point of supply, under an Energy-as-a-Service agreement or with the plant owned by the customer.
Next step
Assess your site — tell us your load, phasing and grid position, and we will outline a microgrid configuration.
