Sustainability in Engineering: Transforming Data Centres into Eco-Friendly Hubs

A Green Data Centre

If Michelangelo could come back, and be tasked with creating another masterpiece to represent our age, his best bet would be painting a Data Centre (DC). Thousands of tonnes of material, hundreds, if not thousands of kilometers of pipework, ducts and cables, racks after racks of computational machinery. And what the aggregated machine does is even more remarkable; holding up the fabrics of 21^st century life; science, commerce, entertainment, governance and much more.

Digitalisation growth statistics are astonishing. Statista estimates Internet of Things (IoT) active devices connected to the internet in 2023 to be 15.15 bn, and predicts a growth to 29.42bn in 2030[1]. These devices, in return, generate data exchange and storage requirements that hit 120 zettabytes in 2023 and are heading for 180+ Zettabytes in 2025[2].  Nearly all of this data is stored, and most importantly processed for new insights in DCs. DCs in turn require colossal amounts of materials, logistics and energy at their inception (embodied Carbon) and during their operation (in the form of power, cooling and refreshed equipment). This means global DC energy consumption now exceeds national energy consumption of most countries[3]. How could these giants of 21^st century be made to have low or zero environmental impact? The following is only a starter menu for solving one of the most challenging assignments an engineer could be tasked with.

Figure 1 Historical and predicted total global data generated in a year (zettabyte: 10²¹ bytes)

1.     A Sustainable & Low Impact Design

• Local renewables at scale

Having all its operational energy sourced from renewables is the high water mark, but conceals the fact that geographically dispersed renewables will involve substantial transmission losses, even if we assumed the electrons they mobilise will end up performing work in the target DC. Renewable energy certificates (RECs) or  energy attribution certificates (EACs) can be sold to a DC that is connected to a different electricity grid. DCs have vast surface areas that can Nearby land may allow additional PV deployment with potentially a private wire that eliminates adding power flow to already congested national grids. Wind and hydro where and if feasible, and finally topping up the operational energy shortfall with PPA[5] is the gold standard. Hourly or half-hourly matching PPAs ensures that DCs are not importing high CO₂ grid electricity at peak times and buying the equivalent green power produced at times of low grid CO₂.

• Water saving design measures

DC water usage is mostly dictated by local climate. Where available, sea and river water cooling as well as heat export eliminates the need to use water-intensive evaporative cooling. Rain, grey or surface water use, when used in combination with recycling[6] is another feasible measure in areas with high rainfall. However for water-stressed regions and providing that low Carbon energy is abundantly available locally (i.e. extensive renewables or SMRs [7]), the guaranteed clean electrical output can allow chillers to be deployed to enable water-free heat rejection at much higher temperatures. This of course comes with higher cooling loads and embodied Carbon of additional infrastructure, so a delicate balance exists between additional water versus electricity use and added infrastructure. Water preservation, recycling and replenishment remains an area of great focus with many large corporates committed to  water replenishment schemes.

• Environmental pollutants

From electrical switchgear to cooling refrigerants and modern two phase immersion cooling systems, a whole host of substances can be deployed in DCs that inevitably have adverse environmental impacts. Some are being phased out (i.e. SF6 gas in switchgear, high GWP refrigerants or PFA Substances), but some mediums still remain that if leaked into the atmosphere, have adverse environmental impacts (i.e. Hydrocarbon-based oils or Fluorocarbon based refrigerants). It is possible to minimise or replace most of these working or isolating fluids with natural substances (i.e. ammonia or propane as refrigerants or, natural ester dielectric oils) to ensure the operational phase is as environmentally benign as possible.

• Carbon assessment

Up to 80% of a product’s environmental impact are determined at the design phase. Whether the capabilities exists to perform full WLCA or assessments or not, WLCA will have to become a KPI that generates insights, guides the procurement and technology selection and ultimately shapes DC design. In its best form, it should enable the design team to replace Carbon-intensive elements of design with more environmentally friendly solutions. RED Engineering Design has its own in house tool (REDuce-to-Zero) where in combination with other platforms we conduct WLCA for fabric and equipment of DCs with the intention of making clear recommendations to our clients.

• Zero waste to landfill policies

From construction to operational phase, and from initial deployment of equipment to refurbishment and server replacements, DCs must adopt a no waste policy. High levels of replicability and standardisation can be achieved in DCs. Modularisation of data halls is standard practice now and not only the fabric, but also HVAC components can be modularised, manufactured and factory tested and plugged in on site to guarantee performance and eradicate site waste. Equally all computational machinery (some of which contain multiple rare-earth materials) in DCs have a high replacement cycle, and should have a 100% end of life recyclability planned into their original design and deployment. This will somewhat chip away at humankind’s colossal e-waste (estimated by WHO [8] to be 53.6m tonnes in 2019).

• NZC readiness

It is clear to many DC designers that climatic and local constraints may leave a 100% Carbon free solution beyond the reach for existing applications. However it is still possible to aim for a highly configurable design which from a power and cooling demand as well as surplus heat management has the potential to transition and accept appropriate NZC solutions as they come to market. This informed our extensive ‘program of research in which we mapped how highly integrated decentralized microgrids with reliable power and thermal outputs can enable future DCs to remove stranded back-up assets, assist constraint local grids and in time adopt to the most promising future fuels and technologies (i.e. fuel cells). Some of our energy system level considerations are outlined next.

2.     Integration into Local Energy Ecosystems

• Add value to local community

Most large scale planning applicants are required to state how their proposed development enhances the neighborhood and benefits its residents. A community scale integrated approach can enable a DC to offer great enhancement to its residents, from heating local swimming pools and homes, to partnering with other businesses to create highly integrated industrial clusters with strong and long term local employment opportunities, DCs can become not only an energy prosumer, but also the anchor around which a community grows and prospers.

• Consolidate e- and bio-fuel supply chains

DC owners and operators have vast purchasing power, and are collectively so significant that they can become a market signal for phasing out fossil fuels. The EU is working hard to support and consolidate the supply chain of biofuels. DCs can use a blend of biofuels to power their backup and potentially continuous power plants, with multiple examples now of HVO[9] and biodiesels deployed in DCs. A greater adoption of green fuels can consolidate and encourage the supply chain for these fuels as well as suitable power plants. Even more ambitious is emerging interests in a partnership between DCs and SMRs[10] which can see DCs becoming the anchor around which an economically vibrant industrial cluster of power and heat intensive stakeholders form. This can potentially include a Hydrogen fuel supply chain as a byproduct, with Fuel Cell incorporation to enhance the economic attractiveness for initial investors (link to SMR white paper).

• Elevate cooling temperatures for heat export

Nearly all of the electricity that DCs use to perform computational tasks is manifested as heat. Increased IT densities and adoption of direct liquid cooling driven by AI is now enabling this heat to be captured at higher temperatures with two major benefits, firstly waterless free-cooling in most climates and secondly improved techno-economic viability of heat export. Typical liquid cooling return temperatures are 50°C (potentially even higher). At these temperatures coupling the data centre heat rejection to external energy systems such as district heating, agricultural, process and factory energy inputs can be a viable option. The added benefits are improved fuel utilisation factors, enhanced environmental KPIs, and freeing up valuable space which otherwise is taken up by heat rejection equipment.

• Support for utilities

The speed of transition to a NZC status by worldwide utilities is limited by the speed by which renewables can be incorporated into the energy systems. Their intermittency and lack of inertia brings power quality challenges to the grid operators. Our own parent company (Engie), one of the biggest RES[11] operators globally, estimates that when RES penetration reaches 30% of overall mix, for each new GW of additional RES about 0.15GW of storage is required to perform balancing duties. DCs are able to be active participants in all the balancing services that modern grids with high RES input require, especially so in grid-constrained regions. Participation in flexibility markets (DSR/FFR), load balancing and utility grid reinforcement not only offers the ability to monetise some of the flexibility assets DCs own, but also present DCs as an enabling and integral part of future energy systems.

3.     Set an example

• Record, report and reduce

RED maintains that environmental reporting allows the DCs to stay ahead of the curve. While a lot of it has so far been voluntary, new regulatory forces such as EU EED[12] and CSRD[13] do and will require reporting KPIs such as Carbon, energy and water use, power and waste heat utilisation, renewable share and temperature set points. More stringent sustainability requirements will make it harder for those lagging behind to attract tenancy and investment, and as such Uptime Institute predicts a challenging period for the sector in the run-up to 2030 when publicly stated NZC goals are due. Ultimately though the regulatory aim is to trigger a program of deep retrofit, so it pays to remain at the forefront.

• Turn visibility into advocation

Milestones that large consortiums set themselves – such as Climate Neutral DC Pact – generate corporate momentum. It is a vital part of demonstrating the ability of business to voluntarily enshrine stewardship and environmental advocacy as an integral corporate mission. When Coca Cola became the first to launch its water mandate program in 2007, it only took the next giant (PepsiCo) 2 years to follow. Now over 40 fortune 500 including all GAFAM members have a water replenishment scheme. Similarly most DC operators are aiming for climate neutrality by 2030. That momentum and advocacy overcomes system inertia and directs investment. And beyond 2030 the challenge doesn’t stop, as understanding and quantifying scope 3[14] bring in a lot of down and upstream emissions that will test the resolve of the industry to find solutions for its wholistic environmental impact.

Our mission at RED is to turn data centres into the engines of change that liberate 21^st century economic activity from fossil fuels. Our starting point is finding the data centre’s place in the wider energy ecosystem, and planning it so that it is NZC-ready, can perform its computational duties optimally and is of the greatest value to its neighborhood.

[1] Statista IoT forecast (Source)

[2] Bernard Marr & Co (Source)

[3] Global DC demand reported by IEA to be 240-340 TWh in 2022 (Source).

[4] Building Integrated PVs.

[5] Power Purchase Agreement.

[6] Up to 75% of cooling tower bleed water can be recycled.

[7] Small Modular Reactors

[8] World Health Organization (source)

[9] Hydrogenated Vegetable Oil

[10] Small Modular Reactors

[11] Renewable Energy Source

[12] Energy Efficiency Directive

[13] Corporate Sustainability Reporting Directive

[14] The third emission category as defined by the GHG Protocol