The hidden challenge behind gigawatt-scale AI infrastructure

As the power demand of data centers accelerates, the challenge is more than simply supplying more power via utilities — it’s staged development of whole energy systems, while ensuring that end-user requirements for reliable supply is met at each stage of the build-out. 

When data centers become power systems

Large-scale data centers introduce engineering challenges that were once primarily associated with power system operators. 
 
Grid code compliance, fault ride-through requirements, system resilience, voltage regulation and network interoperability are becoming central design considerations for facilities that may ultimately consume as much electricity as a small country. 
 
At the same time, developers are under pressure to deliver capacity faster than ever before. The modular design approaches that supported the growth of the data center industry are increasingly being tested as projects scale beyond traditional boundaries. 
 
This raises an important question for the industry:  
 
Should future facilities continue to be designed by scaling up familiar data center models, or should they be planned more like integrated power systems from the outset? 
 
The answer is unlikely to be universal. Different projects and operating models will require different approaches. But it’s increasingly clear that decisions made early in the planning process can have significant implications for reliability, resilience, equipment selection and long-term performance. 

A system-level challenge

One of the recurring themes emerging across the energy sector is the need for more integrated planning. 

Generation, transmission, storage and demand can no longer be considered independently and connecting to the grid can no longer be treated as a "plug-and-play" exercise. Developers must navigate the complex realities of grid-code compliance, transient stability, fault levels and utility operating requirements while meeting their own expectations for rapid deployment and high availability. 

Bottom-up, modular data center designs are highly effective for scaling quickly, but they can push electrical networks toward their thermal limits while increasing fault levels. Top-down approaches better manage these system-level risks but require greater coordination and more significant upfront infrastructure investment.  

At Jacobs, we’ve seen firsthand that there is no single blueprint for connecting gigawatt-scale infrastructure. We’ve identified different approaches to integrated planning of data center energy infrastructure that successfully merge utility requirements with data center end-user requirements taking into account data centers of different scales and with and without large scale generation on the campus. This provides the energy security that end-users require, specifically during staged build-outs. 
 
The challenge is understanding which approach best aligns with the project and long-term operating strategy and expenditure appetite.  

Ultimately, the future of energy systems will be defined by how quickly and confidently they can be planned, connected and delivered.