A new era of load growth: How data centers are becoming part of our communities and energy system

The demand is here. How will you serve it?

What was once thought to be stagnating to decreasing load growth has begun to shift. Onshoring of manufacturing capacity, building and vehicle electrification, and the expansion of computational needs are creating a new era of load growth. Data centers may encompass future US electricity demand as high as 12% by 2028¹.

Society continues to demand more cloud computing power, be it smoother video streaming, wider-reaching cloud enterprise services, and forward-looking applications such as autonomous vehicles. These trends were already driving data center growth before the AI boom of recent years. Power is an essential, if not critical path, to this growth.

Not all grid participants will be impacted the same. Connection requests of 15 GW in AEP Ohio by 2030² (>40% increase from peak 2023 load) and 80 GW in Oncor (59 GW for data centers³, >90% increase from peak 2023 load), indicate duplicative requests in the queue. Knowing whether “real” annual demand growth is 4% or 15%⁴ is a grid planning team’s constant struggle. Looking across known builds highlights the primary markets where data center demand is growing most as well as robust activity in secondary markets.

This trend of data centers being closer to local customer demand will only increase as cloud computing and AI inference expand due to real-time response needs. Conversely, AI training has greater optionality for site location. While much attention remains focused on energy for AI training needs, at PwC we have seen increased energy needs for AI model inferencing, like many corporate customers that rely on pretrained models⁵.

As power providers seek to serve this load, the ability to supply reliable, cost effective, and preferably sustainable energy continues to be differentiators to attract data center demand.

The rules of the game work but aren’t enough for near-term growth

Centralized, integrated planning over a set timeframe has been successful in managing infrastructure risks and customer impacts with utility regulators – but this model is being challenged with more dynamic market needs. The PJM capacity auction closing at a total cost increase to consumers from $2.2 billion to $14.7 billion⁶ (nearly a seven-fold increase) this summer was just one example of what the future may hold. Utilities and system operators are now quickly updating demand forecasts to account for the increase in resource requirements to service new loads.

When a single data center can use as much power as 80,000 houses⁷, utilities may have to find a way to accelerate their development timelines, which are traditionally measured in years, not months.

Companies are not waiting on utilities to solve the supply-side challenges and are now finding their own demand-side solutions, including developing on-site generation or other grid solutions, or relocating to markets with power availability. Solutions leveraging gas capacity, battery storage, geothermal generation, and fuel cells are creating a bridge to long-term electricity grid solutions. This may be the start of a trend where modular demand needs are matched to distributed generation.

Benefits beyond power alone

A PwC study found that data centers provide a 6x multiplier of indirect or induced jobs, across the US, for every direct job they provide. Coupling data centers with a larger industrial base could help address siting needs and enable robust economic hubs.

In addition to contributing taxes and jobs, data centers can also become part of the grid by helping alleviate congestion and, over time, becoming a load-balancing asset. While there may be pushback from DC customers reticent to flex IT capacity, the ask may be achievable. For example, data center load flexibility would be most needed for two to three hours during high peak season to trim demand as solar ramps up and down⁸.

To achieve this objective, data centers will require a mix of engineering, operational, and commercial solutions. Engineering solutions might include data center virtualization, shifting AI training loads or leveraging multiple data centers, ASIC chip efficiency gains, and better cooling technologies. Operational advances could involve leveraging back-up generators or on-site battery storage. On the commercial side, tariffs or market participation contracts that create mutually incentivized results to benefit the grid and local communities may become more common.

As utilities grapple with managing current demand from data center customers and/or attracting additional data center demand, they should seek to leverage the scale and capabilities of the technology sector and large C&I customers and consider evaluating these five capabilities:

Thank you to Tara Woram and Arsen Akopian for their contributions to this article.