Alright DERTF crew — next up in our DERVOS 2026 panel announcements: data center flexibility.
For years, the working assumption has been that a data center is the ultimate firm load. It runs 24/7, it never blinks, and it needs the grid to be there for every last megawatt, every hour of the year. So utilities planned for it that way, and data centers waited in the interconnection queue until the wires and generation got built.
That assumption is obviously starting to crack. Utilities can’t build fast enough, and the developers who connect first are the ones who can tell the utility: you don’t have to serve all of me, all the time. Flexibility has transformed from a nice-to-have sustainability story into a speed-to-power strategy.
That flex can come from a few places. The compute itself can bend, shifting workloads across time and across a fleet of sites. On-site generation or storage can carry the load when the grid is stressed, so the servers never notice. And new interconnection approaches and standards are emerging that let utilities and grid operators actually count on that flexibility, with faster energy-only and other non-firm service as the reward.
That raises some questions. How much can compute really flex once uptime commitments and customer contracts are on the line? Who pays for on-site assets, and who carries the risk when the grid calls on them? Is non-firm service a permanent model, or just a bridge until firm capacity arrives? Is ERCOT’s or SPP's approach a template for the rest of the country? Is the data center becoming the biggest DER on the grid? And zooming way out: for a century we’ve built the grid to meet peak demand, always. If the largest new loads can flex, is data center growth the tip of the iceberg for a more dynamic, efficient way of building infrastructure across the board?
We have an incredible lineup joining us to unpack this. Varun Mehra (Technical Lead, Google) leads data center flexibility work spanning compute, on-site assets, demand response, VPPs and flexible interconnection. Victoria Fleming (CBO, Scale) develops the data center deals where on-site energy provides the resource adequacy utilities need to offer energy-only interconnection, after a decade at NextEra working with data center customers. And Arushi Sharma Frank (Founder and Principal of Luminary Strategies, advisor to Emerald AI and Base Power) has pioneered flexible interconnection standards work, including the PCLR standard in ERCOT, after leading energy market policy at Tesla Energy.
Bring your hardest questions for the Q&A next week.





On whether ERCOT is a template, the clock flexibility has to beat is long: DC Hub models Dallas time-to-power at about 38 months, and Austin sits in the CAUTION band on DCPI at 42.8. The resource mix behind non-firm service also differs by grid, with SPP at about 25% wind and 40% gas this morning against ERCOT's 48% gas. Curious whether the panel sees non-firm interconnection shortening that clock or mostly reordering the queue. DC Hub (dchub.cloud), as of Oct 30, 2026: https://dchub.cloud/connect
Thanks for framing this so plainly. Flexibility only counts once the grid operator can count on it, so the interconnection standards may end up mattering as much as the hardware. From what we can tell, large loads can be listed well before the rules for curtailing them exist, which makes the ERCOT versus PJM contrast worth watching. Glad to be reading along.