Three emerging trends in data center sustainability

By Cummins Inc., Global Power Technology Leader

Three emerging trends in data center sustainability

Data centers are expected to undergo tremendous growth. As the demand for artificial intelligence (AI) expands, so does the demand for AI’s computing power. Not only do AI-specialized servers have more computing capabilities, but they are also more potent than conventional servers. In 2023, total data center energy usage was estimated at 4.4% (176 TWh) of US consumption. This is projected to rise as high as 12% (400 TWh) of the nation’s energy usage by 2028, depending on AI adoption, infrastructure efficiency, and sustainability initiatives.

Seeing this projection, data center investors and operators recognize the effort needed to maximize the sustainability of data centers and limit their carbon emissions. Therefore, data center sustainability is a significant talking point, and its global attention is expected to grow. 

Let’s look at three emerging trends in data center sustainability and how Cummins works in this space.

1. AI energy and cooling optimization

The rapidly growing usage of AI is driving up the energy usage of data centers. As AI capabilities continue to improve, it is increasingly becoming an embedded tool in business and consumer processes. Practically, an AI chatbot can usually provide a more accessible and useful answer than a traditional search engine (i.e. Google) does, with an almost instant response. However, the key challenge is that this query via AI uses roughly ten times the energy compared to a similar query via a search engine. Not only does this increase the server power usage, but it also increases the heat load to be rejected by the cooling system, increasing the energy needed for cooling. 

Therefore, this growing usage of AI places more focus on improving data center energy efficiency and lowering their specific energy consumption per operation. Data centers are tasked with finding ways to sustainably and efficiently deliver more computing power in the same package. There are several approaches to tackle this challenge.

A key focus in reducing the energy consumption of data centers is to improve power usage effectiveness (PUE). A lower PUE indicates that more of the site's total energy goes directly to powering servers rather than being lost to cooling and other overhead services. A focus on lowering PUE relentlessly drives the optimization of cooling efficiency. For instance, implementing hot/cold aisle containment and having closer control of ambient temperature to avoid excessive cooling.

There is a shift to direct liquid cooling, which has a higher cooling efficiency than standard air cooling installations. However, these are expensive systems, and AI GPUs' increased heat rejection requirements primarily drive their adoption.

A significant opportunity for energy efficiency in data centers is heat reuse, or using the waste heat from server cooling to provide energy for other valuable operations. These synergistic relationships can exist between data centers, district heating systems, agricultural greenhouses, or industrial processes.

However, a more fundamental approach to reducing energy consumption is to improve the chip and computing efficiency of the servers. This allows more computing capability for the same power draw and thus improves the data center's efficiency. Modern chips can lower their consumption when not operating at full capacity, and energy efficiency is a focus area for chip development.

It is even possible to apply AI to the operations and management of a data center, where AI may optimize its operations to maximize energy efficiency in ways that humans can’t. These AI-supported systems could identify stranded servers, take them offline, or potentially find software improvements to reduce load. AI will likely be applied to other data center support services, such as optimizing predictive maintenance—scheduling maintenance only when needed, with improved prediction accuracy. AI could even play a part in future data center design and layout improvements.

2. Renewable and low carbon energy adoption

The reliability of the energy supply is a key concern for every data center. Several pressures exist on energy sources available to a typical data center via local utilities. The utility’s infrastructure is aging and becoming less reliable, and it is struggling to keep up with the expanding demand in areas where data centers cluster. A data center operator also has neither control over the energy mix of the local utility nor the carbon footprint of that energy. 

One of the trends in sustainability for data centers is to source low-carbon energy that can be dedicated to the data center. Such generation plants are constructed by an Independent Power Producer (IPP), which sells power to the data center under a Power Purchase Agreement (PPA). Typically, these would be ‘behind-the-meter’ (BTM) power plants that sit on the customer’s side of the utility meter and generate renewable energy from sources like solar, wind, and gas turbines. Newer technologies, such as small-scale nuclear or hydrogen fuel cells, could also be considered by IPPs. The focus here for the data center is two-fold: to have autonomy over the energy generation and confidence that the energy is from a low-carbon source.

A related trend in IPPs and utilities is using on-site energy storage—essentially large batteries called BESS (Battery Energy Storage Systems). BESS provides a fast demand response, which is valuable for smoothing out highly variable load profiles and softening the peak demand on constrained utility grids. BESS also provides interim power while bringing gas generators online if the data center needs to switch to standby power. 

3. Circular economies

The growing expectation in the data center industry is that operators set sustainability targets for which to strive. All large-scale data center operators now have a sustainability goal, typically targeting 2030. Progress toward achieving these goals requires measuring impacts such as carbon emissions and, in particular, life cycle carbon analysis of their entire operations. 

A data center (and its power supply) is a complicated entity whose impact on the environment is much more than just its power draw at the operating state. To optimize for sustainability, it is essential to understand a data center’s impact across its entire life cycle.

Carbon emissions are calculated in three categories. Scope 1 emissions are those that arise directly from the core operation of the facility. Scope 2 emissions are those created in the provision of purchased power and, therefore, are heavily determined by the generation technology of the IPP or utility. Scope 3 emissions do not come directly from the data center's day-to-day operation but from the production and transport of purchased goods, such as electronic equipment. For a data center, Scope 3 emissions include activities such as construction, waste, recycling, and shipping goods.  

These emissions are comprehensively and systematically considered using the approach of a Life Cycle Assessment (LCA). LCA takes the “cradle to grave” approach, considering the impact of the data center at every stage of its operation. This begins even at the extraction of the raw materials used in its construction, and all activities throughout its operation are systematically considered until its decommissioning and demolition. Over this life cycle, the bulk of emissions from a data center are classified as Scope 3.

To reduce Scope 3 emissions, data centers are increasingly looking to source equipment from suppliers focused on measuring and lowering their products' carbon footprint. This requires focusing on efficient production, reducing waste, lowering water usage, and championing reuse and recycling. 

Cummins’ sustainability road map: Planet 2050, Destination Zero, and Green Innovations

Cummins is focused on improving the sustainability of data centers and is actively partnering with customers and suppliers to make a difference in our world. Under its Planet 2050 and Destination Zero, Cummins has set its own internal sustainability targets. 

Practically, Cummins is making meaningful progress by reducing energy consumption, waste, and water usage, while also being a true partner in the communities in which we operate. In line with our Destination Zero strategy, Cummins is developing innovative low and no-emission technologies for use in various applications, including data center demand response programs. 

Learn more about how Cummins is sustainably powering the Data Centers industry: https://www.cummins.com/generators/data-centers/sustainability
 

Author Profiles

Cummins Office Building

Cummins Inc., Global Power Technology Leader

Cummins Inc., a global power solutions leader, comprises five business segments – Components, Engine, Distribution, Power Systems, and Accelera by Cummins – supported by its global manufacturing and extensive service and support network, skilled workforce and vast technological expertise. Cummins is committed to its Destination Zero strategy – the company’s commitment to sustainability and helping its customers successfully navigate the energy transition with its broad portfolio of products. Cummins has approximately 69,900 employees and earned $3.9 billion on sales of $34.1 billion in 2024. See how Cummins is leading the world toward a future of smarter, cleaner power at www.cummins.com.

Want to find out more?

Receive the latest from Cummins in technologies, products, news, and insights.

Redirecting to
cummins.com

The information you are looking for is on
cummins.com

We are launching that site for you now.

Thank you.