Data Centres, Utility Power and the Path to Decarbonisation

Author: Ian Bitterlin, Portman Partners Associate

In this fourth and final article in the series, we turn our attention to the relationship between data centres and utility power — not only the impact that national power grids have on data centres, but increasingly the impact that rapidly expanding data centre estates have on national grids and their decarbonisation ambitions.

Data centres are purchasing significant quantities of renewable energy and, in some cases, investing directly in new generation capacity. The question, however, is whether these initiatives always deliver the environmental benefits they appear to promise — or whether some risk becoming another form of greenwashing.

At the time of writing, a growing number of locations have either introduced restrictions on new data centre developments or are openly debating their future expansion. These include Amsterdam, Frankfurt, Dublin, Sydney and Singapore.

Ireland and Singapore: Record-Breaking Data Centre Nations?

Two particularly interesting examples are Ireland and Singapore, where national power grids are facing immediate challenges associated with continued data centre growth.

Ireland has rapidly expanded to become one of Europe’s largest data centre hubs. Operators have increasingly invested in Irish wind generation in support of their renewable energy ambitions.

Singapore faces a very different challenge. Its limited land availability restricts both solar and wind generation opportunities, while the country’s electricity system remains heavily dependent on natural gas.

Both nations also have relatively limited interconnection capacity through which electricity can be imported or exported.

Their neighbouring countries face their own pressure to expand renewable capacity for electric transport, domestic heating and grid decarbonisation. As a result, exporting low-carbon electricity may become less attractive as governments prioritise their own national climate objectives.

Ireland

Ireland — and Dublin in particular — has emerged alongside London, Frankfurt, Amsterdam and Paris as one of Europe’s most important data centre markets.

Relative to its population and national grid capacity, Ireland has arguably become one of the most concentrated data centre markets in EMEA.

There are several reasons for Dublin’s success.

The city offers an educated, English-speaking workforce, a strong technology ecosystem and historically attractive conditions for international technology companies. The country’s business tax environment has also encouraged international investment, while planning frameworks and access to power have supported data centre construction.

Direct employment within data centres may be relatively modest compared with the capital invested, but a substantial network of construction, engineering, maintenance and technology businesses has developed around the sector.

Dublin also benefits from excellent fibre connectivity between the United States, Ireland and continental Europe.

It is not necessarily a low-cost location when either capital expenditure or operating expenditure is considered. For hyperscale operators, however, low cost is only one part of a much broader location strategy.

Ireland’s Electricity Challenge

Ireland’s electricity system has historically depended heavily on fossil fuels, although onshore wind capacity has expanded significantly.

Plans have also progressed for the Celtic Interconnector between Ireland and France, designed to strengthen electricity links between the two countries.

Such an interconnector could allow Ireland to export surplus wind energy during periods of high generation while importing lower-carbon French electricity when renewable output falls.

The challenge is that continued data centre growth places additional pressure on an electricity system that must already expand and decarbonise to accommodate transport electrification, heating and wider economic growth.

During research for a Channel 4 Dispatches documentary examining video streaming and the impact of data centres on the Irish grid, I analysed Ireland’s ten-minute grid statistics.

At 15:10 on 5 November 2020, Ireland was consuming approximately 3.42GW of electricity. Wind generation was extremely low, resulting in a grid carbon intensity of approximately 512g CO₂/kWh.

The generation mix at that moment was approximately:

• Coal: 18% — 620MW from 1.2GW of capacity
• Wind: 1.2% — 38MW from 4.2GW of capacity
• Hydro: 5% — 173MW from 216MW of capacity
• Gas: 49% — 1.67GW from 4.27GW of capacity
• Oil: 12% — 394MW from 1.27GW of capacity
• Other: 2.6%
• Imports from the UK: approximately 15% or 530MW
• Exports to Northern Ireland: approximately 1.8%

The carbon-intensity figures used for the individual generation technologies were based on internationally recognised IPCC figures available at the time.

One striking observation is that Ireland already had approximately 4.2GW of installed wind capacity, greater than the entire national grid demand at that particular moment.

That raises an important question: if substantially more wind capacity is installed, what happens to the surplus electricity during periods of strong wind and low demand?

One potential answer is large-scale energy storage or hydrogen production.

No Electricity Source Is Truly “Zero Carbon”

It is also important to distinguish between low-carbon and genuinely zero-carbon electricity.

Every generation technology has an associated carbon footprint.

Even nuclear, wind and solar require raw materials, construction, manufacturing, transmission infrastructure, maintenance and eventual decommissioning.

Concrete, steel, copper, carbon fibre and other materials all carry embodied emissions.

Therefore, describing an electricity source as “zero carbon” can be misleading.

Very low carbon is a valuable objective and one towards which electricity systems should continue moving. But technically, as low carbon as practicable is often a more accurate description.

Managing Intermittent Renewable Power

In theory, if wind output were sufficiently strong and stable while electricity demand was low, Ireland could potentially meet a substantial proportion of its demand through wind generation.

The challenge is grid stability.

Electricity systems must respond instantly to changes in demand and generation. Large proportions of intermittent generation increase the need for balancing resources, storage, interconnection or dependable low-carbon generation.

Experience in European electricity markets has demonstrated that high renewable penetration requires careful grid management.

Wind generation alone cannot guarantee supply when the wind stops blowing.

How Large Is Ireland’s Data Centre Load?

Ireland’s population is approximately five million.

At the 3.42GW grid load referenced earlier, national electricity consumption was equivalent to roughly 0.68kW per person.

If approximately 500MW is attributed to data centres, the remaining demand falls closer to 0.58kW per person.

For comparison, at approximately the same time the UK was consuming around 38.3GW, equivalent to approximately 0.58kW per person across a population of roughly 66.6 million.

After estimating the UK’s data centre demand, the non-data-centre figure becomes approximately 0.55kW per person.

These broad comparisons suggest that Ireland’s electricity system supports a significantly larger proportion of data centre demand than the UK’s.

Dublin provides another useful comparison.

With approximately 1.2 million residents, its theoretical share of national demand would have been around 800MW. If approximately 500MW of Irish data centre demand was concentrated in and around Dublin, the remaining metropolitan load would be approximately 300MW.

Bristol in the UK, with a similar urban profile and around 90% of Dublin’s population, traditionally consumes in the region of 230MW on average and approximately 300MW during winter peaks.

These comparisons suggest that an estimated 500MW of actual Irish data centre load was within a plausible range and may at the time have represented approximately 15% of national electricity demand.

Further planned development could increase that proportion considerably.

The Renewable Energy Paradox

This creates an important policy question.

If data centre operators finance significant amounts of wind generation but also consume all or most of the additional renewable output, has the wider national electricity system genuinely decarbonised?

When wind generation is low, the data centres still require power.

That electricity must then come from gas, imports, storage, another low-carbon baseload source or emergency generation.

At the same time, Ireland must expand its electricity system to accommodate electric vehicles, electrified heating and broader economic growth.

Data centre investment may therefore help accelerate renewable infrastructure, but it can simultaneously increase the overall scale of the decarbonisation challenge.

Ultimately, some dependable low-carbon source must support the grid when intermittent renewable generation is unavailable.

Singapore

Singapore presents a different but equally important case.

The island nation has a population of approximately 5.7 million and has become one of Southeast Asia’s most important digital infrastructure hubs.

At 10:45 on 4 March 2021, Singapore’s electricity demand was approximately 6.9GW, with a carbon intensity of around 492g CO₂/kWh.

That equates to approximately 1.2kW per person.

Direct comparison with Ireland is difficult because Singapore’s hot, humid equatorial climate creates substantial year-round air-conditioning demand.

Based on regional market estimates available at the time, Singapore’s data centre estate was approximately 600MW, equivalent to around 9% of grid demand.

Continued expansion could push that figure significantly higher.

Singapore possesses many of the characteristics that have made Dublin successful: strong connectivity, an educated workforce, widespread English usage and an attractive environment for international technology companies.

Its main challenge is energy.

Singapore’s Energy Constraints

Singapore has historically generated more than 97% of its electricity from natural gas.

Land is scarce and expensive, which constrains large-scale solar projects.

Onshore wind opportunities are limited, while offshore wind development is complicated by some of the busiest shipping routes in the region.

Imports can supplement domestic generation but do not automatically solve the carbon problem if imported electricity itself comes from carbon-intensive sources.

Singapore therefore faces a difficult challenge: maintaining the energy security required for a major global digital hub while simultaneously reducing the carbon intensity of its electricity system.

For data centres, however, the issue is arguably less about whether Singapore is an appropriate location and more about how the national electricity system itself evolves.

The country’s connectivity, business environment and regional importance continue to make it highly attractive.

As a result, operators increasingly rely on mechanisms such as Power Purchase Agreements and Virtual Power Purchase Agreements to pursue renewable energy targets.

How Do We Measure the Size of a Data Centre Market?

Industry reports frequently publish highly specific estimates of data centre capacity in megawatts.

These figures should be interpreted carefully.

For example, a report might state that a particular city contains 400MW or 600MW of data centre capacity.

But what exactly does the figure represent?

It could refer to:

• utility connection capacity;
• maximum facility demand;
• installed transformer capacity;
• design IT load;
• operational IT load;
• carrier-neutral colocation capacity;
• hyperscale capacity; or
• a combination of different measurements.

A facility described as “50MW” therefore does not necessarily have 50MW of active IT equipment.

What Does “50MW” Actually Mean?

Several interpretations are possible.

A planning application may reference utility capacity rather than net IT load.

Power companies may specify capacity in MVA rather than MW, meaning that assumptions regarding power factor become relevant.

The grid connection must also support the entire facility — including cooling, pumps, lighting, losses and supporting infrastructure — not simply the IT load.

Peak facility efficiency is therefore important when translating utility capacity into actual computing capacity.

Finally, facilities rarely operate continuously at their theoretical design maximum.

Hyperscale facilities may eventually achieve high utilisation because of the economics associated with large-scale infrastructure investment.

Retail colocation facilities, by contrast, may begin with a relatively small anchor tenant and gradually increase utilisation over several years.

A theoretically large facility may therefore carry a significantly smaller actual IT load.

This is why published market-capacity statistics should always be treated with caution unless their underlying methodology is clearly explained.

Hardware Efficiency Also Changes the Picture

Data centre electricity demand does not necessarily increase in direct proportion to digital activity.

Modern computing hardware can perform substantially more work for each unit of electricity consumed than previous generations.

A facility may therefore accommodate growing digital workloads while its electrical demand remains stable — or even declines — as hardware is replaced.

This creates further uncertainty around market estimates based only on historic facility ratings.

Hyperscale operations are generally easier to predict because utilisation tends to be higher and infrastructure is designed around large, rapidly deployed workloads.

Renewable Power

Renewable electricity is an important part of achieving genuine data centre sustainability — but it should come after efficiency and optimisation.

Available renewable technologies include:

• Wind
• Solar photovoltaic
• Solar thermal
• Sustainable biomass
• Geothermal
• Hydroelectric power
• Tidal power
• Wave power

Renewable electricity can also be used to create secondary fuels such as green hydrogen.

Other electricity sources, including nuclear and some forms of waste-to-energy, may not technically be renewable but can nevertheless provide significantly lower carbon intensity than fossil-fuel generation.

Low Carbon Does Not Mean Zero Carbon

Every electricity source carries embodied carbon.

Generation facilities must be manufactured and constructed.

Transmission infrastructure must be installed and maintained.

Equipment must eventually be replaced, recycled or decommissioned.

Solar panels, wind turbines, nuclear facilities and hydroelectric schemes all require materials and industrial processes.

The appropriate objective is therefore not simply a marketing claim of “zero carbon”, but the progressive reduction of lifecycle emissions across the electricity system.

How Electricity Actually Moves Through the Grid

One of the most important concepts in discussions about renewable electricity procurement is that power from a national grid does not travel directly from one named generator to one named consumer.

A London data centre may purchase enough Scottish hydroelectric generation to match its annual electricity usage.

That does not mean the electrons entering the facility physically came from a specific Scottish hydroelectric plant.

The data centre consumes electricity from the overall grid mix at that particular moment.

Renewable electricity procurement is therefore fundamentally an accounting and commercial mechanism.

A simple way to visualise a national grid is as a large shared system into which hundreds of generating sources feed electricity while thousands or millions of consumers simultaneously withdraw it.

Generation and consumption must remain balanced continuously.

Voltage and frequency must also remain within extremely tight operating tolerances.

When demand rises, generation must respond.

When demand falls, generation must reduce or the excess electricity must be stored, curtailed or exported.

The electricity entering an individual data centre therefore represents the instantaneous mixture of generation operating across the connected grid.

There are no separate “green electrons.”

This distinction becomes particularly important when a company claims to be powered by 100% renewable electricity based on annual certificates while continuing to consume grid electricity during periods when renewable generation is unavailable.

Transparency around these claims is essential.

On-Site Renewable Generation

For most large data centres, generating all required renewable electricity on-site is impractical.

Large-scale wind turbines require considerable land.

Rooftop solar installations may contribute only a small percentage of the facility’s overall electricity demand.

A data centre requiring tens or hundreds of megawatts cannot generally power itself entirely from rooftop generation.

Wind also introduces further complications.

To operate a data centre entirely on wind generation, substantial excess generation capacity, grid support, energy storage and emergency backup would be required because wind output varies continuously.

As a result, utility-scale renewable generation connected through the wider grid remains the more realistic option for most large facilities.

Wind and Solar Generation

Solar generation has the advantage of being comparatively predictable.

However, it requires substantial physical space and only generates electricity during daylight hours.

Wind is often more attractive for very large electricity consumers because utility-scale projects can produce substantially greater output.

Nevertheless, wind generation remains variable.

A turbine’s rated output represents its maximum operating capacity, not its average annual output.

Onshore wind farms may typically achieve annual capacity factors of approximately 35–40%, depending on location.

Offshore wind can achieve higher capacity factors, potentially around 45–50%, although construction, connection and maintenance costs are generally greater.

This difference between rated capacity and annual output is crucial when assessing renewable energy claims.

A 4MW wind turbine does not produce 4MW continuously.

It may generate nothing during periods of low wind and its full rated output only under favourable conditions.

As wind becomes a larger proportion of electricity generation, grid balancing becomes increasingly important.

Purchasing Renewable Energy

For data centres, the final step towards reducing operational carbon is procuring electricity from lower-carbon or renewable sources.

As the renewable proportion of national electricity systems increases, all consumers connected to those grids gradually benefit from lower carbon intensity.

The ISO/IEC 30134 series includes measures intended to assess the resource effectiveness of data centres.

Carbon Usage Effectiveness, or CUE, considers the carbon associated with the electricity used relative to the facility’s IT energy consumption.

A data centre connected to a national grid without any special electricity agreement can legitimately claim the renewable proportion already contained within that grid’s generation mix.

However, renewable electricity certificates and procurement agreements allow individual organisations to go further.

Large technology companies increasingly highlight these arrangements within sustainability reports and corporate communications.

Some hyperscale operators have also achieved substantial efficiency improvements through high infrastructure utilisation and low PUE.

However, improving efficiency is fundamentally different from reducing total digital demand.

The commercial objective of these companies remains the sale and expansion of digital services.

Who Should Receive the Renewable Electricity?

This raises a broader ethical question.

Should data centres receive a disproportionately large share of limited renewable electricity?

Or should scarce low-carbon electricity instead be prioritised across society — including hospitals, schools, care facilities, public infrastructure, transport and domestic heating?

There is no simple answer.

Large data centre operators can make renewable energy projects financially viable by committing to substantial long-term purchases.

This can accelerate the construction of new generation.

But if newly built renewable projects simply meet additional data centre demand, the rest of the grid may remain dependent on its previous generation mix.

In that scenario, total renewable capacity has increased but national decarbonisation may have progressed less than headline figures suggest.

The Challenge of Large Data Centre Clusters

When data centres represent only a small percentage of overall grid demand, the system can generally accommodate their growth alongside changing renewable generation.

As their share rises towards 20%, 30% or more of local or national demand, the relationship becomes more complicated.

Several challenges emerge.

First, data centres require electricity around the clock regardless of renewable generation conditions.

Second, electricity systems must accommodate periods when renewable output exceeds immediate demand.

Third, sufficient dependable generation must remain available for periods when renewable output falls.

Fourth, substantial investment may be required in transmission infrastructure, interconnectors, energy storage and backup generation.

Finally, policymakers must determine who ultimately pays for these investments.

Power Purchase Agreements

A Power Purchase Agreement (PPA) allows an electricity consumer to contract directly or indirectly for electricity from a particular generation source over a defined period.

As national grids continue to decarbonise, the need for organisations to purchase separate renewable electricity contracts may eventually diminish.

In a genuinely low-carbon grid, all connected consumers would automatically benefit.

Achieving that outcome, however, will require enormous investment.

Electricity systems must not only replace existing fossil-fuel generation but also expand to support electric vehicles, electrified heating, industrial decarbonisation and growing digital infrastructure.

Until then, PPAs can provide a mechanism through which businesses place commercial pressure on electricity markets to develop additional low-carbon capacity.

Virtual Power Purchase Agreements

A more complex mechanism is the Virtual Power Purchase Agreement, or VPPA.

A VPPA allows an organisation consuming electricity in one grid to financially support renewable generation located in another electricity market.

The organisation obtains the associated renewable-energy certificates and may use them to match some or all of its electricity consumption.

Physically, however, its data centre continues consuming electricity from its local grid.

This distinction is critical.

The renewable electricity supported through the VPPA may be generated hundreds or thousands of kilometres away and may never enter the electricity system supplying the data centre itself.

VPPAs are legal and widely used commercial instruments.

They can also help finance new renewable developments.

However, sustainability claims based on them should clearly explain the difference between physically consuming renewable electricity and financially matching electricity consumption with renewable generation elsewhere.

Hyperscale Operators and Global Renewable Procurement

Some of the world’s largest digital infrastructure companies aggregate their global electricity demand and enter into PPAs and VPPAs across numerous regions.

The attraction is clear.

A business can locate its data centres where connectivity, tax policy, land, customers and commercial conditions are most favourable while purchasing renewable certificates from locations where wind or solar generation is more abundant and economical.

The local electricity provider continues supplying the physical electricity.

The renewable project provides the financial and environmental attributes used to offset or match the consumption.

This can produce credible investment in renewable infrastructure, but the distinction between contractual renewable energy and the physical carbon intensity of the local grid should remain transparent.

Large-Scale PPAs and Electricity Markets

The growth of intermittent renewable generation also creates challenges for electricity pricing.

Wind and solar have no conventional fuel cost once installed.

During periods of strong generation and low demand, wholesale electricity prices can fall dramatically — sometimes even becoming negative.

This creates difficult questions regarding how generation capacity, reliability and grid stability should be valued.

Traditional electricity markets were designed largely around generators whose operating costs included fuel.

As renewable generation becomes dominant, electricity-market structures will need to evolve.

The Netherlands: Another Example

The Netherlands has faced similar debates around the growth of very large data centre developments.

Major technology facilities can bring substantial investment and employment, but they can also consume enormous amounts of electricity and land.

This creates public debate when renewable generation such as wind farms appears to be developed primarily to support large technology campuses.

Grid operators must also plan major network upgrades to accommodate such facilities.

The issue is therefore not simply whether a data centre uses renewable electricity.

It is whether the entire electricity system can expand sustainably while continuing to decarbonise for every consumer.

Re-Using Waste Heat

One potentially valuable contribution data centres can make is the reuse of waste heat.

Servers generate large quantities of low-grade heat that must normally be removed through cooling systems.

In suitable locations, this heat can potentially be transferred into district-heating networks and used to warm homes, commercial buildings or other facilities.

Projects in parts of Europe are actively exploring this opportunity.

However, successful implementation requires appropriate infrastructure, nearby heat demand and cooperation between data centre operators, utilities and local authorities.

It is therefore promising, but not automatically practical for every facility.

Conclusion

Data centres are becoming critical infrastructure for the modern digital economy.

At the same time, their electricity consumption is becoming increasingly important to national energy policy.

Renewable procurement, PPAs and VPPAs can play an important role in financing the transition towards lower-carbon electricity systems.

But the industry must remain transparent about what these mechanisms actually achieve.

A data centre purchasing enough renewable certificates to match its annual consumption is not necessarily consuming renewable electricity every hour of every day.

Likewise, building additional renewable generation specifically to satisfy new data centre demand does not automatically mean that the wider national electricity system has decarbonised.

The challenge ahead is therefore much broader than simply making individual data centres appear green.

It is about creating electricity systems capable of supporting digital growth, transport electrification, heating and wider economic activity while simultaneously delivering dependable, affordable and genuinely lower-carbon energy.

That will require collaboration between data centre operators, utilities, governments, regulators, engineers and energy investors.

This concludes the series, although it certainly does not conclude the discussion.

The data centre industry faces a period in which questions of energy, infrastructure, sustainability and ethics will become increasingly interconnected.

About the Author

Ian Bitterlin is a Portman Partners Associate and Chartered Engineer with extensive experience in data centre power and cooling.

His career includes more than three decades working within the data centre industry, following earlier experience in rotating electrical machines and power systems.

Ian has previously served as CTO for Emerson Network Power Systems in EMEA and as a Visiting Professor within the University of Leeds School of Mechanical Engineering.

He is currently Principal Consultant at Critical Facilities Consulting.

Throughout his career, Ian has held senior positions with organisations including Anton Piller, Liebert, Emerson Network Power, Active Power and Chloride, contributing to the development of important technologies and products within the critical infrastructure and data centre sectors.

Ian has twice received recognition for his Outstanding Contribution to the Data Centre Industry — from Data Centres Europe in 2009 and DataCenterDynamics EMEA in 2015.