Google & Voltus Expand to 1 GW: Can "Bring Your Own Capacity" Solve the Data Center Grid Crunch?
Data centers will pay almost anything to connect to the grid and access capacity. So why couldn’t they buy that needed capacity from other customers, by getting them to change their behavior during system peaks, or by adding batteries on site?
The AI industry needs the grid, while the grid needs help to modernize and become more efficient, while keeping costs in check. Meanwhile, customers have existing capacity that could in theory be purchased at some price, as a perfect arbitrage opportunity.
Today, customers don’t free up much capacity during peaks because they don’t get paid enough. But they might, if the incentive is right.
As utility CEO John Rowe once famously said in the 1990s of utilities and efficiency incentive programs, “the rat must smell the cheese.” Pay a rational economic actor enough and they will do nearly anything.
We may be beginning to see the first signs of a shift in that direction, as data centers increasingly focus on capacity at the grid edge. Distributed energy resource companies such as Sunrun and Tesla recently announced a Google-funded program to enlist batteries in California and test the concept.
In June DER company Voltus and Google announced a collaboration to develop a 100 MW virtual power plant capacity, with a
“bring your own capacity” approach. With this, data centers finance distributed capacity, which is registered with the ISO and then delivered to the load-serving entity that provides electricity to the data center.
Last month, the two parties announced a significantly expanded approach, with a gigawatt-scale (Voltus CEO Guernsey put this number in an interview at 1,070 MW) VPP across the PJM grid. This would aggregate flexible capacity from technologies such as smart thermostats, batteries, and industrial assets. Participating customers would receive economic benefits including bill savings and cash payments.
Voltus later indicated that the program will ramp up to the full 1,070 over three or four years and that it will then manage the capacity through 2040 under a contract with Google.
CEO Dana Guernsey commented the deal is designed to help catalyze the market and create a durable commercial structure that could be emulated by others. She also commented that much of this capacity would come from commercial-scale on-site batteries. Other end-uses are more difficult to harness, but it’s disappointing that they are not really trying to address them.
The announcements were short on detail, offering little of value to utilities, grid planners, or regulators. It would have been helpful to see potential milestones, timelines for delivery of capacity, and anticipated technologies by capacity, and perhaps a commitment to report on progress at various milestones.
With something of this magnitude and potential importance, more detail would let us know if it’s worth emulating at scale. We don’t need all of the competitive financial details. But it sure would be nice to get a general sense as to the size of the cheese needed, how many rats are eventually attracted to it, and what this might mean for our energy future.