Solar Power for Data Centres: What It Can Carry, What It Cannot, and the Hybrid That Works

Data centre operators on weak grids are asking a fair question. The generators burn fuel for hours a day, grid tariffs keep rising, and the sun is shining on the roof. Why not put solar on it?

The honest answer is that solar can take a real share of the energy bill on the right site, and it cannot run a data centre by itself or replace any part of the reliability design. Proposals for data centre solar power, often searched as data center solar power, tend to arrive either oversold or dismissed. Both miss the useful middle: a hybrid of solar, battery storage and generators that cuts fuel and grid spend while the UPS and generator plant keep doing the job they were designed for.

This article covers why the load is hard for solar, where solar does earn its place, what must not change, and what an owner should model before committing.

Why a Data Centre Is a Hard Load for Solar

A data centre load has three properties that work against solar.

It is dense. A server rack occupies well under a square metre of floor and commonly draws several kilowatts continuously, and high-density racks draw far more. The building exists to pack as much electrical load as possible into as little floor as possible, and the roof above it is no bigger than the floor.

It is flat. IT load barely changes between day and night, or between weekday and weekend. Solar output follows the sun.

It runs around the clock. The load runs all 8,760 hours of the year. The array produces for part of each day.

The power density arithmetic is what settles it. Under the 1,000 W/m² irradiance used for standard module ratings, a module of about 21 percent efficiency produces about 210 W per square metre, and that is its peak, reached around midday on a clear day. Walkways, spacing and rooftop plant reduce the area that can be covered. Averaged over 24 hours, with night, cloud, heat and dust included, output per square metre is a fraction of that peak. Set that against a floor where every rack position draws kilowatts all day and all night, and the comparison is not close. Even a single-storey data hall with its whole roof covered produces a small share of its own daily energy, and a multi-storey building less still.

None of this argues against solar at a data centre. It argues for being clear about which job solar is doing.

Where Solar Earns Its Place

Cutting diesel on weak-grid sites. Where the grid is unreliable and generators run for many hours a day, every kilowatt hour from solar is a kilowatt hour of diesel not bought, trucked in, stored and burned. A hybrid of solar, battery storage and generators can cut generator run hours and fuel use materially without touching the reliability design. Our guide to hybrid solar systems covers how the three sources share a load.

Offsite supply through the grid. Where regulation allows wheeling or private supply over the network, a solar plant on cheaper land elsewhere can supply energy at a scale the site's roof never could. It arrives through the grid under a contract, usually a power purchase agreement, and the data centre still needs its full on-site backup because the grid connection is how the energy reaches it. Whether this is possible at all depends on the country's rules. Our guide to solar PPAs covers the off-site structures.

Ancillary loads. Offices, lighting, security and perimeter systems, and some mechanical plant sit outside the critical power path and can be supplied from solar with simpler arrangements than the IT load needs.

Sizing follows from the role. An array meant to cut diesel on a weak-grid site should be sized against the daytime load it can actually displace, and against the minimum load the running generators need, or much of its output will be curtailed. Battery storage lets solar reach further into the evening, at the cost of the battery. Either way, the useful question is how many generator hours and litres of fuel the system removes, not how many kilowatts peak fit on the roof.

The three jobs solar can do at a data centre. The roof cannot carry a dense, flat, round-the-clock load, so the question is which job solar is doing. Source: MIMAH engineering practice.

The Reliability Architecture Stays

A data centre's power design has three layers. The UPS carries the critical load through the seconds to minutes between a grid failure and the generators taking over. The generators carry the site through long outages. Redundancy means that the failure of any single component does not drop the load: N is exactly enough capacity, N+1 adds one spare unit, and 2N provides two complete, independent systems.

The Uptime Institute's Tier Classification System describes four levels built on these ideas: Tier I is basic capacity, Tier II adds redundant capacity components, Tier III is concurrently maintainable, and Tier IV is fault tolerant. On power, Uptime is blunt. In its explanation of common Tier misconceptions it states that the only reliable source of power for a data centre is the engine-generator plant, that utility power is not even required for Tiers, and that the generator plant must be able to carry the critical load without runtime limitations. Most certified sites use the grid for economic reasons.

Read in that light, solar and grid power are both economic sources. They reduce what the site pays for energy, and neither is counted on for availability. A hybrid design that quietly shrinks the generator plant or its fuel storage because there is solar on the roof has traded reliability for fuel savings without saying so.

Integration is where the engineering effort goes. On a grid failure, the transfer sequence of UPS, generator start and synchronisation must behave exactly as it did before solar was added, and the solar inverters must follow the generators rather than disturb them. That sequence should be tested under real failure scenarios before handover, not assumed from the drawings.

The reliability architecture stays when solar is added. Solar and grid reduce what the site pays for energy; neither is counted on for availability. Source: MIMAH engineering practice, after the Uptime Institute Tier Classification System.

PUE and Why Cooling Matters in Hot Climates

Power usage effectiveness, as defined by The Green Grid, is the total energy used by a data centre divided by the energy used by its IT equipment. The ideal value is 1. Everything above it is overhead: cooling, power conversion losses, lighting and the rest of the building. The metric has since been standardised internationally as ISO/IEC 30134-2.

In hot climates the cooling plant works harder. Condensers reject heat into hotter air, there are fewer hours when outside air can do the cooling for free, and dust fouls coils and filters faster. The overhead is correspondingly larger. PUE is only meaningful when measured at the same points across a full year; a reading taken on a cool night, or on the hottest afternoon, says little about the annual energy bill.

That has two consequences for a solar decision. Cooling is the one part of a data centre's load that is shaped a little like the sun, rising with afternoon heat, so solar output and cooling demand line up better than solar and IT load do. And every kilowatt hour of overhead removed through better containment, cooling design and maintenance is energy that never has to be generated, stored or backed up. Efficiency belongs first in the model, before sizing any array.

Batteries Do Two Different Jobs

The word battery covers two different duties in a data centre, and confusing them causes bad designs.

UPS batteries bridge the gap between a power failure and the generators taking the load. They discharge rarely, at high power, for minutes, and they have to be full at every moment because nobody knows when the grid will fail.

Battery energy storage for energy shifting cycles every day, delivering hours of energy at moderate power. It stores midday solar for the evening, lets generators run at efficient load or switch off, and shaves grid demand peaks. Its value comes from being used.

These duties conflict. A battery cycled daily for energy shifting cannot also be guaranteed full for the UPS moment unless a UPS reserve is ring-fenced in its controls and sized on top of the shifting capacity. The cleaner route is usually to keep them separate. They also differ in cell selection, protection and fire design, which local codes will govern. Our guide to commercial battery storage covers the energy shifting side in detail.

Generators Still Need Testing and Fuel

A generator that runs less because solar is carrying the daytime load does not need less maintenance. It needs to be proven more deliberately, because it is no longer being exercised by daily running. That means regular testing under real load, with load bank testing where the site cannot safely offer enough load, and hybrid controls that keep running generators above the minimum load their manufacturer recommends, since diesel engines run lightly for long periods suffer from wet stacking. Our guide to generator load bank testing sets out how that is done.

Fuel needs the same discipline. Storage should stay sized for the outage the site is designed to survive, not the shorter one solar makes more likely. Stored diesel degrades and needs testing and polishing, and on sites supplied over long or disrupted roads the delivery chain is part of the reliability design. The principle is the one we apply to black-start sets at power plants, such as the 1,500 kVA diesel generator sets MIMAH has supplied in Nigeria: the set that rarely runs is the set that most needs testing.

What an Owner Should Model Before Committing

Interval load data, with IT and facility loads separated. The IT load sets the floor and the facility load sets the shape. Without both, any solar proposal is sized against an assumption.

Generator run hours, fuel use and delivered fuel cost. On weak-grid sites this is where the savings come from, so it needs real records, not estimates.

Grid availability pattern and tariff. When the grid fails, for how long, and what grid energy costs at different times of day.

Available space and realistic yield. Roof, car park canopies and adjacent land, with a generation estimate from site irradiance data and honest allowances for heat and dust.

The regulatory route for offsite supply. Whether wheeling or private supply is permitted, and on what terms.

The battery duty. Energy shifting, UPS, or both with a ring-fenced reserve, written down before anyone sizes a battery.

Control and integration. How the solar and storage system interacts with the UPS and generator sequence, and how that will be tested before handover.

Monitoring. A platform that reports fuel displaced, generator loading and battery state alongside solar output. Our guide to solar monitoring systems covers what to specify.

What an owner should model before committing to solar at a data centre. Efficiency comes first, before any array is sized. Source: MIMAH engineering practice.

MIMAH's renewable energy team designs and builds hybrid solar, storage and generator systems for critical facilities, and our maintenance team looks after the generators they depend on. If you are weighing solar for a data centre and want the numbers modelled honestly before you commit, get in touch.