Commercial solar for data centres.

Data centres carry one of the most demanding electrical loads of any building type. IT racks, cooling plant, UPS systems and back-of-house all draw power continuously, day and night, with very little daily variation. For an operator, energy is a headline operating cost and a headline sustainability question at the same time, which is why on-site generation is worth modelling carefully rather than dismissing as too small to matter.

This is commercial solar by sector over 50 kWp, designed and assured outside the MCS scheme that applies to smaller domestic systems. We build it engineer-first: an on-site survey, a half-hourly load model, and a system sized to the load you actually carry. For a data centre the demand shape is unusually simple, and that simplicity works in your favour.

The value of a commercial array sits in self-consumption. A unit of generation used on site offsets an expensive import unit, while a unit exported to the grid is paid far less. So the return tracks how well the output lines up with demand through the day rather than how many panels the roof can hold. A data centre lines up about as well as a cold store or a water treatment works: the base load is high and effectively flat across all 24 hours, so every kilowatt the array produces in daylight is consumed the moment it is generated. There is almost nothing left to export.

The honest caveat is one of scale rather than fit. Total demand at a data centre is so high that a rooftop array offsets a share of it rather than the whole. That is not a weakness in the self-consumption logic, it is the opposite: because the load dwarfs anything the roof can generate, there is no risk of overproducing or exporting cheaply, and every unit installed earns at the full value of avoided import. It also stands as a visible, measurable sustainability commitment, which tends to matter where clients and reporting frameworks ask about the energy behind the compute.

Most data centres present as a large single-occupier footprint with substantial roof area, and many carry rooftop plant and screening that has to be worked around. As an indicative orientation only, sites in this sector tend to sit in the region of 250 kWp and upward, with the firm understanding that rooftop solar offsets only a fraction of the building's IT and cooling load. Treat that band as a sense of scale rather than a quote. The real figure comes from the on-site survey and the PV*SOL model, sized to your actual roof, your shading and your half-hourly demand.

The cooling fraction moves with outside air temperature, and that works in your favour. A data centre's draw is not perfectly flat across the year. The IT load itself holds steady, but the mechanical cooling that rejects the heat off the racks rises and falls with ambient conditions. Chillers, CRAC and CRAH units and the associated pumps and fans work hardest in the warm summer afternoons, which are the same hours the array reaches its strongest output. So the part of the load that does vary climbs in step with generation, which lifts daytime self-consumption beyond what a purely flat profile would suggest. Free-cooling and economiser hours shift the balance the other way in winter, when output is low anyway, so the seasonal swing rarely costs you self-consumed generation. We confirm the actual shape from your half-hourly data and the cooling system's behaviour before sizing, rather than assume a constant draw.

The UPS and PDU layer means the data you need to model the site already exists. A data centre is metered more finely than almost any other building we work in. The supply runs through UPS systems and the power distribution units that feed each rack row, and that infrastructure is monitored continuously through the building management and DCIM systems for PUE reporting and capacity planning. The practical consequence is that the half-hourly consumption profile a PV*SOL model needs is usually already captured at a fine resolution, so the generation-against-load study can be built on real measured demand rather than on a fitted estimate, and without new sub-metering installed first. That gives an unusually defensible read on the self-consumed share before anything is specified.

A data centre never powers down, so the array is tied in around a live, resilience-critical load. There is no quiet shift, no seasonal shutdown and no maintenance window where the IT load can be dropped. The connection and commissioning are therefore planned around an estate that must stay energised throughout, with the work sequenced against your concurrent-maintainability and resilience requirements so the array is added alongside the existing UPS and standby generation rather than across them. The grid-side tie-in runs under the G99 connection process with Northern Powergrid as the network operator across our region, handled as a designed application rather than improvised on site, and the commissioning under IEC 62446-1 is staged so it never sits between the racks and their supply. A grid-tied array follows the grid and drops out in an outage, so it offsets import while standby supply stays the job of your UPS and generators, and we are explicit about that distinction from the start.

The case rests on displacing import at the full retail unit rate, with effectively nothing exported cheaply. A commercial array earns the difference between the price of a unit you would otherwise buy and the far lower price a unit fetches when it is exported. Most buildings sit somewhere between those two, exporting a real share of summer midday generation. A data centre sits at the strong end of that spectrum: because the round-the-clock load dwarfs anything a rooftop array can produce, near enough every kilowatt-hour generated displaces a bought unit at the full retail rate, and effectively none is exported cheaply. That is the strongest position a self-consumption model can take, and it is the structural reason the sector models so cleanly. The flip side, set out plainly elsewhere on this page, is one of scale: the array offsets a share of a very large bill rather than the whole of it.

Capital-allowance treatment can apply to the qualifying plant. Solar equipment installed by a business can fall within the capital-allowance regime, which changes the after-tax cost of the investment for a company that pays corporation tax. The rules, the rates and the available allowances are set by HMRC and change between fiscal events, and the position depends on your company's circumstances, so this is a matter for your accountant or tax adviser rather than something we advise on or build a figure around. We flag that it can apply so you can factor it into your own appraisal, and we make no claim about your tax position.

We model your figure from your own tariff and demand, and set the ranges out away from this page. We do not publish a payback, a return or a saving figure on a sector page, because an honest number for a data centre depends on your import tariff, your contracted capacity, the cooling profile and the roof the survey lands on, and only becomes real after the PV*SOL model is built against your half-hourly data. The indicative payback and return ranges, and how the capital case is constructed, sit in our finance material, where they can be properly substantiated rather than reduced to a headline. Any number we give you is modelled, not promised, with the basis disclosed and your own tariff and demand behind it. Where the figures do not stack up for a particular site, we tell you that straight rather than size a system that does not earn its place.