Onsite Energy for Data Centers: Finding Solutions for Future Power

Data centers are looking beyond the traditional power grid to generate the electricity needed to support advanced artificial intelligence. Despite the introduction of new renewable energy generation, the current grid infrastructure makes it difficult to keep up with sharply rising demand across multiple sectors. Data centers, in particular, account for a large increase in the amount of electricity needed.

The current grid simply cannot support the scale and pace of that expansion. It can take up to a decade just to permit and integrate new large-scale transmission projects into the current grid. Instead of waiting for that to happen, technology companies are turning to solutions that develop power generation sources and hyperscale data centers side by side.

This trend will likely wind up being more than just a stop-gap solution. It will also present opportunities for innovation and expansion of technologies that could support a more flexible grid in the future. But realizing those possibilities will require close collaboration among engineers developing technology across multiple domains, from electrical supply to data architecture, power monitoring and cybersecurity.

The sheer demand for new power means co-location projects are unlikely to favor any one source of electricity. As long as it’s possible to build a generation or storage facility on site, any economical source of power will do. Solar and battery storage are likely to see strong near-term adoption, as will turbines that run on a range of fuel sources, including natural gas and other traditional fossil fuels. Generally, the more easily available a given fuel source is, the greater the potential for economic efficiencies from an on-site energy source. Government policies such as the U.S. Inflation Reduction Act, European Green Deal and China’s AI energy efficiency mandates, will also play a large role in shaping how and where adjacent projects will be built.

This trend exists in part because most of the challenges involved in delivering on-site power have reasonable solutions available. The main difference between co-location projects for an AI-focused data center and other industrial users tends to be an increased emphasis on redundancies in the power supply. That desired level of redundancy requires substantially more switchgear compared to similar projects — typically triple the amount.

Data center operators don’t necessarily need to restrict themselves to stock parts, however. In some cases, these new entrants into the energy generation business are developing their own specifications could produce better results. For example, reducing the footprint of a piece of switchgear can make it possible to fit more of them in a single cabinet. When you triple the amount of gear going into a facility, those space savings add up.

Engineers at TE Connectivity developed a compact elbow that allows exactly this type of efficiency. Even though it may seem like a small change, it took multiple years to develop the product and required expert manufacturing engineering and new molds to produce it.

Given the rate of change in the industry, developing innovations to accelerate data center co-location projects will require very close collaboration. For TE, that means introducing a more consultative approach to the traditional vendor model. Creating more engineer-to-engineer touchpoints allows teams to stay ahead of trends and innovate in advance of changing specifications.

As on-site generation facilities develop, the way data centers hook up to their power supplies could change further. A move to higher-voltage electricity supply could produce efficiencies since data centers could potentially transmit the power they generate with fewer losses, fewer parts, less current and less heat generation. Onshore wind and solar generation plants have already begun to shift toward increased use of higher-voltage connections.

However, the more power you pass through a part, the more challenging it is to keep the system within required safety specifications. Solving those increased technical challenges typically leads to more expensive parts. Data center operators will need assistance from trusted engineers to understand what options are available to them and to help weigh the potential cost of higher voltage parts across a system against the longer-term potential efficiencies. That collaboration will ultimately help data center operators make informed decisions that meet the unique needs of their project’s generation source and location.

Connecting engineers across different industrial domains can also help accelerate development in a rapidly changing space. Because TE has a wealth of experience with the systems used throughout data centers, engineers have opportunities to look at ways to make those systems interact more efficiently. Advanced monitoring systems, in particular, could bring together information from the local smart grid and the voltage and current at the end of each row in the data center to improve reliability across the site.

Co-located sites are not happening in a vacuum. The ongoing need to improve grid transmission to the point where it can support much larger electrical demand is itself likely to be a decades-long project. The exact shape of the future grid remains unclear, but power generation from the widest possible variety of sources will be part of the solution. To the extent that adjacent projects spur continued advancements in developing areas such as battery storage or nuclear power supplied by small modular reactors, they could play a role in advancing overall supply.

On-site solutions also will help buy time for additional innovations needed to solve larger transmission challenges. Even after grid operators develop the ability to keep up with surging electricity demand, co-located facilities are likely to remain a valuable layer of much-needed redundant supply on a greatly expanded grid.