"Redefining the Journey: Tech-Agnostic Microgrids for Data Centres"

By Ben Pritchard, AVK CEO.

At AVK, we believe resilience comes from choice. No single technology can power tomorrow’s data alone. That’s why we design solutions across the full power chain, from turbines and engines to batteries, renewables and future fuels. Our technology agnostic approach ensures data centres can thrive in a changing energy landscape, as well as providing our clients with the bespoke designs that ensure their data centre meets regulations and requirements.

Introduction: beyond one size fits all

Data centres have rapidly become the nervous system of the digital economy. As artificial intelligence (AI), high performance computing and cloud services surge, the power profile of the modern facility is changing just as quickly. Loads are larger, power densities are higher, and AI training can drive sharp step changes and long duty cycles. At the same time, grid connections across the UK and Europe are difficult to secure on the timelines the industry needs.

In this context, the old “one size fits all” playbook (typically a single technology with a uniform design replicated from site to site) breaks down. A diesel only strategy may satisfy a narrow standby brief but struggles with sustainability and permitting. A dual fuel/gas engine only or turbine only stance can leave efficiency on the table at low load or hamper agility during AI ramp events. A ‘batteries solve everything’ mindset ignores the realities of long duration supply, fault current, and black start needs.

AVK SEG’s view is straightforward: resilience comes from choice. Technology agnosticism is not indecision; it’s a design philosophy that puts outcomes first and keeps options open – across fuels, OEMs (original equipment manufacturers), and control strategies – so each site gets the right blend for its workload, location and growth path.

Microgrids are the innovative technology that can benefit from our agnosticism. The development of a microgrid can allow data centre operators to become operational that much faster, providing the solution to today’s particular concern of lack of power availability. This combined with AVK’s technological agnosticism can allow for a bespoke data centre design that caters to the sites precise requirements. 

Why technological agnosticism matters

Agnostic design avoids the trap of choosing a single “winner” of a solution and then forcing every problem to fit it. Instead, we start with the outcomes – resilience, efficiency, sustainability and cost – and assemble the right toolset for each.

Resilience. No single technology guarantees uptime under all conditions. Blending generation and storage provides redundancy at multiple layers and avoids single points of failure.

Efficiency. Different technologies have different sweet spots. Matching assets to the load curve, especially with AI’s peaks and troughs, keeps each unit near optimal efficiency.

Sustainability. The energy transition is uneven. Some regions will prioritise hydrogen blends quickly; others will lead with biogas or aggressive renewables build out. An agnostic scheme can pivot as local fuel pathways mature.

Economics. Fuel prices, carbon costs and policy incentives move. A diversified portfolio allows operators to dispatch the lowest cost, lowest carbon mix in real time.

Load profiles are changing: AI and low load strategy

There are two practical realities that now shape microgrid design: AI workloads creating spiky demand, and the inevitability of low-load phases. 

GPU clusters can drive rapid step changes as AI training jobs start or migrate between nodes. Systems must ride through these ramps without frequency or voltage excursions. That means fast assets (engines, batteries, grid forming inverters) need to work in collaboration with slower, high efficiency plants.

Secondly, it is inevitable that there will be low load phases. Most campuses are built in phases – early halls run at partial utilisation for months, sometimes years. Turbines may operate outside their efficiency sweet spot at low load; engines can cycle inefficiently if not managed; diesel sets wet stack if idled. 

A credible low load strategy is essential: run fewer units harder, use BESS to trim peaks and fill troughs, divert surplus heat into useful cooling, and stage capacity additions to match demand. This strategy is essential to achieve the outcomes outlined above. 

The technology toolbox: Setting up for success

There are three main components when setting up power solutions for data centres: the technology, the controls systems, and the mixture of fuels and power. It is the strategic, personalised combination of these components that will provide the most optimal power solution for a data centre. The technology options are vast, and each brings their own design benefits and limitations:

Gas turbines – stable, efficient baseload

Modern aeroderivative and industrial gas turbines provide large blocks of power with strong reliability and excellent efficiency in combined cycle or cogeneration modes.

Where they shine

High output density for hyperscale sites with limited footprint.

Best in class efficiency when paired with heat recovery (steam generation or absorption chilling).

Lower local emissions per MWh than simple cycle alternatives at scale.

Fuel flexibility with pathways to hydrogen blends and, in time, higher hydrogen fractions.

Design notes

Turbines have turndown limits; heat rate worsens at low load. Pair with engines and BESS so turbines stay in an efficient band while other assets do the “agility work”.

If water is scarce or permitting is tight, favour air cooled condensers or simple cycle plus heat to chill strategies over traditional steam cycles.

Gas engines – agility, modularity and part load efficiency

Medium speed reciprocating engines offer rapid start, strong part load performance and modular scalability.

Where they shine

Fast starts and high ramp rates for AI step loads and spinning reserve.

High efficiency across a wide load range—ideal for early phases and variable demand.

Modularity allows “right sizing” the running set to today’s load.

Design notes

Engines provide valuable fault current and inertia in islanded operation, simplifying protection.

Plan acoustic treatment early; engine farms in urban locations face tight noise limits.

Fuel transitions (biogas, synthetic methane, hydrogen blends) are a realistic near term decarbonisation path.

Battery energy storage systems (BESS) – instant stability and black start

Utility scale batteries bring capabilities no rotating machine can match.

Where they shine

Millisecond response to arrest frequency excursions during load steps or generator trips.

Grid forming operation to provide virtual inertia, voltage control and fast fault ride through.

Black start of auxiliaries and first up generation after a complete outage.

Load smoothing to keep turbines and engines in stable operating regions.

Design notes

Size BESS for power (MW) to handle ramp rates, and for energy (MWh) to ride through short disturbances or support renewable smoothing.

Integrate with the UPS strategy; avoid duplicated storage by coordinating DC side and AC side reserves where appropriate.

Include lifecycle planning for augmentation, replacement and recycling.

Renewables – lower carbon and operating cost

On site solar PV and, where feasible, wind reduce net emissions and hedge against market volatility.

Where they shine

Near zero marginal cost once installed.

Direct emissions reduction without reliance on certificates alone.

Demand charge management in markets that penalise peaks.

Design notes

Renewables need firming. Pair with BESS and engines/turbines for smooth integration.

Consider behind the meter PPAs with adjacent renewable assets when on site space is limited.

Hydrogen, biogas and synthetic fuels – credible transition pathways

Clean molecules are a practical route to lower carbon without rebuilding the plant.

Where they shine

Hydrogen blends in turbines and engines as supply improves.

Biogas or synthetic methane using existing gas infrastructure for a drop in reduction in lifecycle emissions.

Design notes

Design for fuel flexibility from day one: materials compatibility, metering, safety zoning and ventilation for hydrogen.

Secure long term offtake arrangements for biogenic fuels; availability is regional.

Diesel/HVO standby – niche but still useful

While not a prime power solution for most new campuses, modern diesel or HVO sets remain relevant as last line standby or for specific geographies with limited gas infrastructure.

Design notes

If used, plan for emissions control (SCR/oxidation catalysts) and load banking or BESS assisted loading to prevent wet stacking in low load tests.

HVO improves local air quality metrics and lifecycle carbon versus fossil diesel.

Heat recovery, absorption chilling and district energy – turning waste into value

Microgrids unlock thermal synergies that conventional standby cannot. 

Where they shine

Absorption chillers convert waste heat into chilled water, reducing electrical cooling load—valuable for GPU dense halls.

District heating/industrial heat exports surplus energy to neighbours, creating community benefit and improving utilisation.

Thermal storage (chilled water/ice) time shifts cooling, flattening electrical peaks.

Carbon capture – deep decarbonisation option

Where CO₂ transport and storage infrastructure exists or is planned, post combustion capture on turbines or engines can materially reduce stack emissions.

Design notes

Allow space and ductwork “stubs” for future capture units; progress policy and offtake discussions early.

Control and protection: the brain of the microgrid

In combination with the technology, controls must be considered during the design stage to ensure optimal performance. Whatever the mix, the architecture of the controls system makes or breaks the scheme.

Microgrid controller (MGC). Coordinates dispatch, islanding, synchronisation, power quality and black start sequences. Implements droop control, set point optimisation and contingency logic.

Grid forming inverters. Enable stable island operation with virtual inertia, fault ride through and fast voltage control.

Protection and selectivity. Islanded systems need intentional fault current sources and carefully graded protection; engines and synchronous condensers can help.

Harmonics and power factor. High density IT loads are largely power electronic. Specify harmonic limits (e.g. THDv/THDi) and reactive support via inverters or STATCOM like functions.

Cybersecurity. Treat the MGC and SCADA as critical infrastructure: network segmentation, secure protocols, role based access, monitored remote connections.

Testing. Prove black start, islanding, re sync and AI load step response during IST—not just nameplate capacity.

How to combine technologies: proven patterns

It is unwise to rely on one form of technology when designing a power solution. The combination of fuel and power types allow for back-up generation, providing reliable power to make sure downtime is non-existent. The proven, successful patterns in technology combinations are as follows:

1) Turbine + engines + BESS (hyperscale baseload, AI peaks).

Turbines carry the steady core load efficiently. Engines handle ramps and provide spinning reserve. BESS manages instant transients and provides black start. Add heat recovery to drive absorption chilling for high density halls.

2) Engines + BESS + renewables (phased growth, grid constrained sites).

Engines scale in modular blocks as the campus grows. BESS smooths and provides power quality. PV/wind offsets daytime demand and lowers OPEX. Ideal where early phases run at lower utilisation.

3) Engine only with BESS (edge or water constrained).

Where water for steam cycles is limited or planning favours smaller footprints, an engine centric scheme offers agility, strong part load efficiency and simpler permitting.

4) Turbines + carbon capture + thermal integration (industrial clusters).

In regions with CO₂ infrastructure and heat users, combined cycle with capture plus district energy delivers very low operational carbon and high overall efficiency.

5) Hybrid with HVO standby (strict urban air quality limits).

Gas engines or turbines for prime power, BESS for quality and resilience, and HVO fuelled standby narrowed to rare events—balancing reliability with local emissions constraints.

Designing for geography: legislation and local reality

Agnosticism really earns its keep when policy and permitting differ by country and city. We have diligently expanded our reach throughout Europe, and in doing so it has been important to factor in these differences in legistlation to ensure beneficial outcomes for our clients. 

United Kingdom. Compliance with G99 (grid code) and the Medium Combustion Plant Directive/Industrial Emissions Directive (as implemented in UK law) drives emissions and monitoring obligations for prime plant. Air quality (AQMA) status will influence abatement, stack height and operating hours. Planning conditions often prioritise noise, visual impact and traffic during construction.

European Union. The Industrial Emissions Directive and national transpositions govern thresholds, BAT conclusions and permitting cycles. Fit for 55 measures and ETS costs push design towards higher efficiency and cleaner fuels.

Local incentives and markets. Capacity markets, flexibility services, guarantees of origin and green hydrogen pilots vary widely. An agnostic scheme can monetise these where present without depending on them where they are not.

Infrastructure realities. Gas availability, CO₂ pipeline proximity, water constraints, and solar resource all shape the optimal mix.

The practical takeaway: define a policy aware reference design for each region, then tailor per site. The reference ensures repeatability and speed; the tailoring preserves performance and permitting success.

A credible low load playbook

The early phases are often where inflexible designs stumble. A robust low load strategy is important, minimising energy consumption and operational costs while enhancing system reliability and sustainability. A low-load strategy typically includes:

1. Staging capacity. Commission fewer prime units initially; use modular engines to “right size” running plant and add turbines later if baseload stabilises.

2. Run fewer, run harder. Avoid idling multiple machines at low efficiency. Keep a small number of units near their sweet spot; let BESS handle short spikes.

3. Thermal utilisation. Use absorption chilling or thermal storage so recovered heat remains productive even at modest electrical load.

4. Smart testing. Use BESS assisted testing to avoid wet stacking on diesel standby and to minimise waste fuel burn.

5. Controls tuning. Retune droop, ramp limits and set points as each phase adds capacity; today’s parameters will not be right tomorrow.

Procurement without lock in

Our agnosticism in technology also extends to our supply chain relationships and contracts. This agnostic approach is seen by prioritising multi-OEM optionality, open interfaces, performance-based outcomes, and lifecycle clarity. This allows us to de-risk manufacturing slots, integrate suppliers seamlessly, tailor contracts for the best outcomes and plan for the future. Above all else, our agnosticism is one of the reasons for our high rate of client satisfaction.

 

Conclusion: flexibility is the advantage

The next decade will reward data centre operators who treat power not as a constraint but as a strategic capability. There is no single technology that can deliver absolute resilience, leading efficiency, and credible decarbonisation across every geography and every workload. But the right combination can. That combination will differ by site, by phase, and over time.

Technology agnosticism is how you keep that combination optimal. It’s how you accommodate AI’s volatile demand without over building, how you navigate permitting from Dublin to Düsseldorf, and how you migrate to cleaner fuels without stranding assets. It is, in short, how you build microgrids that are as dynamic as the digital world they power.

The data centre market is booming thanks to AI. According to Goldman Sachs, the explosion in...
By Caff Allen, Global Head of Learning and Development, Black & White Engineering.
By Conrad Purcell, partner, and Kayley Rousell, associate, in the Energy, Power and Natural...
By Anna Pavlova, Senior Vice President for Strategy, Market Development, and Sustainability,...
By Juan Colina, EMEA Data Centre & IT Segment Leader at Eaton.
By Matt Powers, vice president of technology & support services, Wesco.