Source:pv magazine

In Texas, a new kind of reliability standard is reshaping where data centers choose to build and how they power their operations. Under a “kill switch” law that was adopted earlier this year, utilities can forcibly disconnect large, noncritical industrial users like hyperscale data centers during grid emergencies to keep electricity flowing to the largest number of people.
“Utilities across the country are warning operators to be ready for sudden power shutoffs,” explained James Roth, the head of global policy and government affairs at Bloom Energy. He told ESS News that, for operators with contractual uptime commitments, the risk of sudden shutoffs is “simply unsustainable.”
While the Texas bill does also include a voluntary demand response program, it could still spell trouble for data center operators with 24/7 uptime guarantees or AI workloads that can’t afford a second of downtime. That’s where storage comes in. Operators begin to recognize the value of onsite systems in the face of unexpected utility power shutoffs.
The logic is simple: the more control a data center has over its power, the less exposed it is to regulatory risk and sudden disconnections. In Texas, that could look like installing batteries capable of supporting independent operations when a kill switch is triggered.
According to Roth, it’s no surprise that major cloud players like Oracle are looking to shield their operations using onsite storage and generation. What that means in practice, however, is that a new method of data center siting is emerging, where the map is drawn around storage potential rather than just transmission lines.
The general idea remains the same across state lines, but what it looks like in practice varies greatly. In California, for instance, ensuring steady access to power as quickly as possible means designing storage-backed microgrids that let a data center jump the queue and bypass long interconnection wait times.
It’s more of a structural problem, though it’s one that makes traditional grid expansion nearly impossible at the pace AI-driven workloads demand.
“Operators there are adopting islanded microgrids to bypass utility delays,” explained Roth. But, he cautioned, there’s also a political dimension.
California’s Application for Certification (AFC) process still treats onsite systems over a certain size as “thermal” plants regardless of the type of energy they use. Effectively, this locks the project into a series of slow-moving environmental reviews designed for gas infrastructure.
“Lengthy permitting constructs like California’s AFC process…must be reinterpreted,” Roth added, “to accommodate new solutions that can be deployed quickly and reliably.”
“Getting the policy framework right will determine whether the US effectively captures the economic benefits of the AI boom,” he noted.