The demand for energy and water at data centers

How can we sustainably power data centers?

The first challenge for data centers involves energy consumption. Data centers require a significant amount of power to operate, and that power must be reliable as these facilities are often mission critical and must operate 24 hours a day, seven days a week. We can reliably produce the energy required for data centers from traditional fossil fuels like oil, coal, or natural gas. But those sources lead to a large amount of greenhouse gas (GHG) emissions being released into our atmosphere. As we are in the midst of the climate crisis, the entire world is working to reduce our carbon footprint. That must include the data storage industry, which could be the largest consumer of energy by 2030. So, how can we power these facilities more sustainably?

Oftentimes, the first thought that comes to mind is replacing conventional energy with renewable energy sources. But the non-negotiable need for reliability—typically, 99.999% or less than five minutes of downtime per year—conflicts with the intermittent nature of renewables like solar and wind. We want to use those when and where we can, but what happens when the wind doesn’t blow, or the sun doesn’t shine? Some companies are building solar facilities to support their data centers. But that needs to be combined with a lot of battery storage for reliability. Really, we need to have backup power in those situations. That’s why microgrids are a great fit here.

Microgrids pull together distributed energy resources (DERs)—such as renewables, battery storage, and generators—and they can act alone when generation conditions are good. When they’re not, the facility can draw power from the traditional electrical grid. It’s more emissions intensive, but it’s also a consistent supply of power. Modern microgrid control systems can be programmed to monitor system conditions and optimize the use of renewable power sources. By using microgrids, data centers can operate more sustainably when possible, while having access to a reliable energy supply when it’s not.

Blending in hydrogen as a fuel source for combustion turbines (CTs) could also be an option. Currently, most major producers of CTs offer flexible fuel systems that can burn a blend of hydrogen and conventional fuels (some up to 50% hydrogen), allowing owners to burn conventional fuels now while being prepared for hydrogen blends in the future. While you can fuel a CT today with natural gas, the hope is that the blending of hydrogen into natural gas or other hydrocarbon fuels will eventually get to the point where more than 90% of the fuel is hydrogen, which would result in cleaner emissions.

As of late, we are seeing another interesting development in the data center industry: The use of nuclear power. Nuclear power has seen a recent resurgence as we work towards our net zero goals. After all, it can produce copious amounts of energy without the GHG emissions. Nuclear power can also be ramped up and down quickly to support varying load demands. That’s why it’s being considered for data centers, either alone—as traditional nuclear or small modular reactors (SMRs)—or as part of a microgrid arrangement to offset the downtime of renewables. In September, Microsoft announced it was purchasing the Three-Mile-Island nuclear plant to power its data centers. And in October, Amazon declared it was investing more than $500 million to develop SMRs. Clearly, the chips are falling (so to speak). And it’s exciting to see how the industry is adjusting.

We know the demand for data centers will only continue to rise. And worldwide energy consumption will rise with it. By implementing microgrids, DERs, hydrogen, and potentially nuclear power and SMRs, we can meet energy demands while reducing emissions. 

How can we responsibly source the water for data centers?

The other key challenge that data centers face revolves around water usage. Data centers house an incredible amount of equipment, including servers, routers, storage units, security systems, and more. This equipment generates heat and requires lots of cooling to remain operational.

Whether by circulating air or liquid coolant (as is becoming more common with increasing load densities due to AI, machine learning, and high performance computing) through them, a lot of supporting mechanical systems are required to reject that heat. Data centers utilize different types of mechanical systems for this, and some of the commonly seen types can consume as much as 100,000 gallons of water per day during peak conditions.

The most water-hungry types of systems utilize evaporation. Direct evaporative systems introduce water into the airstream where it evaporates, cooling an airstream that is circulated through equipment to pick up heat. Indirect evaporative systems operate similarly, except the evaporatively cooled airstream is separated from the airstream that directly touches equipment via a heat exchanger.

Evaporative cooling is also used in chilled water systems to reject heat from chillers—large pieces of equipment that generate cold water, which can be circulated to directly or indirectly cool equipment via air or fluid—if cooling towers are utilized.

As noted previously, data centers that use evaporative cooling or evaporative heat rejection systems can have large water requirements. Various factors affect water consumption. They include site weather conditions, site water quality, the data center’s allowable thermal operating envelope, and the exact type of cooling/heat rejection system used.

There are many questions a data storage facility must ask itself. Where do we source that much water? Does the geographic location play a role? Once we source the water, how do we treat it? How do we perpetually cool it, and how do we recycle and return this water to the environment? These are all questions that must be answered if we hope to operate these data centers in a responsible manner.

Water availability is a big concern, and the need to conserve water at data centers is becoming even more critical. Ideal locations for data centers that use evaporative systems are often water scarce. Many municipalities either don’t have the utility water infrastructure to support data centers or have constraints that can affect the feasibility of data centers.

As we move forward with the development of data centers, it’s important to keep water top of mind. This is especially true as we are already facing water shortages in some regions within the US and around the world. Water is becoming a scarce resource and water resources are being continuously strained, especially due to climate change.

With great power demands, comes greater environmental responsibility

Technological innovations like digital communications, online shopping and banking, and artificial intelligence can and do bring several benefits to our communities. But as we noted earlier, these kinds of disruptions also bring challenges. In this case, the challenge is storing and protecting an endless amount of data. Data centers are enabling this, but these facilities must be able to operate reliably and sustainably.

The data center industry is aware of its challenges regarding water availability. Some providers have transitioned to cooling systems in new data centers that don’t need water. This presents an additional challenge, as systems that don’t use water often compensate by using more energy to reject heat. But if electricity can be sourced from renewable sources, the ecological and carbon footprint of the data center can still be reduced.

When it comes to power, we must develop ways of weaning data centers off fossil fuels and towards cleaner forms of energy—without sacrificing reliability. We also need to find the most efficient ways of sourcing, cooling, reusing, and treating the water required. As engineers, we believe it is possible. And we’re already making strides towards sustainable solutions. But it will take robust planning, creative ideas, and significant investment to keep up with energy demands while reducing GHG emissions at data storage facilities.