Flexible Data Centers: A Faster, More Affordable Path to Power

Can site-level flexibility really allow us to connect huge amounts of new data center demand to the grid? 

That question has been at the heart of one of the biggest issues in energy, as we grapple with the challenge of adapting our existing energy infrastructure to serve the needs of the massive amount of new data centers set to come on line in the next few years. 

Like many in the industry, we at Camus paid close attention to the Duke report which suggested that we could fit up to 100GW of new load into the existing US grid if it was possible to occasionally curtail loads. Our team has been delivering flexible interconnection solutions based on our grid orchestration platform for the last few years, and we’ve seen firsthand that a little curtailment can massively increase utilization of the grid’s resources. 

This is also reflected in experiences with “connect and manage” models in markets like Australia and Texas. In Texas, providing a “fast path” for new generation which requires curtailment in exchange for an expedited interconnection has enabled ERCOT to add more than 25GW of new generation in the last few years - almost double that of any other US market.

That kind of flexible “fast path” option is not available to large loads today, and up until now, most datacenter power solutions are “either/or”. Either they’re connecting to the grid which will serve 100% of their demand through the existing energy system (but waiting several years for new capacity and upgrades), or they are designing sites to be entirely self-sufficient by colocating with generation or building new onsite generation (which is cost intensive and also takes several years to build). As timelines increase for new grid connections and available capacity (and turbines!) are spoken for, the existing options aren’t going to get data centers connected fast enough.

As Camus started applying our flexible interconnection work to data centers, we wanted to answer the question: what is the practical, real-world blueprint for connecting data centers with site-level flexibility? Whether it’s reducing load, adding onsite generation, or switching to batteries, there are a range of options which could enable data centers to reduce their usage of the grid during key times - depending on how often, how long, and how much curtailment is required. 

Given the very real concerns about data centers driving up system costs for everyone else, we also wanted to know - can adding site flexibility help manage the costs of serving these new loads?

With the support of our partners encoord (Carlo Brancucci and team), the Princeton University ZERO Lab (Jesse Jenkins and team), and financial backing from Google, we were able to dig into this question. We used real transmission system data to model six sites in PJM, to first understand whether site-level flexibility could speed up their path to power, and then to understand what kinds of flexibility options would best meet their power needs. 

The answer to the question: can flexibility allows data centers to connect faster? A resounding “yes.” Building on existing “connect and manage” models, our results showed that a small amount of local flexibility could enable loads to connect 3-5 years sooner. Further, bringing a set of onsite power solutions (including load flexibility, onsite battery, and gas generation), when paired with procured capacity from renewable, battery, or VPP resources, can enable sites to entirely offset the additional system costs of adding new loads. The result is a better grid - one which can support a lot more load, in which energy and capacity prices both remain stable, and in which there are new sources of system flexibility which can increase the overall resilience of the system. 

Download our white paper here.

If you’re interested in applying this work to your data centers, or to sites in your queue, Camus has a flexible interconnection solution for utilities and data centers (named FlexConnect) that can help you put this work into practice.