Santa Clara Implements Flexible Power Program for Data Centers Amid Grid Strain
Santa Clara has introduced a Flexible Load Link Program (FLIP) allowing its municipal utility, Silicon Valley Power (SVP), to reduce or cut off electricity to data centers during periods of grid insufficiency. This program allows large power users like data centers to connect to the grid faster, addressing delays experienced by developers such as Stack Infrastructure. A pilot program with Nvidia and Emerald AI is also underway to dynamically adjust AI workloads based on grid conditions.
Santa Clara, California, is implementing an innovative solution to address the growing power demands of artificial intelligence (AI) data centers, which are outstripping the expansion rate of the local power grid. The city's municipal power company, Silicon Valley Power (SVP), is advancing a Flexible Load Link Program (FLIP) that permits large-scale power users, including data centers, to connect to the grid more swiftly. The condition for this accelerated connection is an agreement by these users to reduce their power consumption when the grid is under strain, with SVP reserving the right to directly limit power supply if reductions are insufficient.
This initiative was prompted by issues faced by data center developer Stack Infrastructure. Its special purpose corporation, 1220 Santa Clara Propco, constructed a 48MW data center in Santa Clara, but delays in power supply from SVP hindered its operation. Stack filed a damages claim, citing uncertainty over power supply as a detriment to its ability to lease the facility. In response, the Santa Clara City Council approved a contract that connects Stack's data center to SVP's Remote Surveillance Control (SCADA) system, enabling SVP to adjust or limit the data center's power consumption as needed.
The FLIP program is being extended to approximately five other major power customers, moving away from the traditional model of requiring extensive grid expansion before data center connection. The goal is to maximize the flexible use of the existing power infrastructure. Furthermore, SVP, in collaboration with Emerald AI and Nvidia, is piloting a technology to dynamically adjust AI operational loads within commercial data centers based on real-time grid signals. This allows GPUs to operate at full capacity when power is abundant and to scale down or postpone non-urgent AI tasks during peak demand to alleviate grid pressure.