Everyone knows now how great Texas is for renewables. Its particular combination of sun, wind, and permissive market structure has led the state to overtake even California in clean energy generation. At the same time, however, the state is nervous about data centers and their effects on the grid, even passing a law this past legislative session to more closely examine the data center industry and to establish protocols for curtailing data center operations when electricity is tight.
But what if you could do for data centers what Texas has done for renewables? That’s roughly the idea Arushi Sharma Frank, who helped bring Tesla’s energy business to Texas and is now an advisor to Nvidia-backed Emerald AI, has come up with.
On Friday, Frank filed a proposal with ERCOT, the electricity market that sits outside federal regulation and covers about 90% of the state, that would reform its rules to allow data centers to connect to the grid much faster. The rough idea is that by applying ERCOT’s existing “connect and manage” system for getting new electricity generators on the grid to new large demand sources like data centers, the data centers can get power more quickly — if they can handle not getting access to the grid sometimes.
The proposal “creates the basis of connect-and-manage of load using the existence system that ERCOT already has for generators and batteries,” Frank told me.
Her idea would reward data centers for being able to modulate how much electricity they need in the interconnection process. This could mean that data centers get credit for curtailment and for having their own generation on site. And crucially, unlike a widely-panned proposal by PJM Interconnection to essentially mandate that some customers be forced to curtail their energy use, the ability to curtail or self-power a data center would result in faster interconnection, not simply the cost of doing business.
Frank pointed to chip designer Nvidia’s recent announcement that it would back a Virginia data center using Emerald AI software to smooth out power usage, saying she wants to be “able to actually do that at scale” for “any developer in Texas.”
Getting power for data centers is one of the biggest barriers to getting them built, and so anything that can deliver faster interconnection without foisting enormous new costs on the system as a whole counts as a win-win. With this system in place, Frank told me, data centers and other large loads could “invest in firming their own power needs before major transmission upgrades get built, enabling them to voluntarily choose to be flexible participants on the grid in exchange for earlier interconnection.”
Texas has been able to deploy wind, solar, and batteries so quickly, many energy policy experts and developers say, precisely because of connect and manage, whereby new generators can get on the grid after just a local grid study, without having to examine their effect on the whole system, which most of the country’s grids require. After these system-wide studies often come expensive transmission upgrades, the costs of which are passed on to all electricity customers in the form of higher bills. This process, Duke University’s Tyler Norris has written, “can often make generators financially unviable, introduce uncertainty for project economics, and delay interconnection by years.”
That level of extensive review is partially responsible for the interconnection delays seen in the rest of the country, which can stretch to as long as several years. Projects in Texas take on average two years to complete the interconnection process, according to the trade group Advanced Energy United.
The trade-off for allowing new resources onto the grid without those upgrades means that they’re more likely to be curtailed if the amount of electricity they generate overwhelms the grid — the “manage” part of “connect and manage.” Frank made an analogy to me between a data center and an 18-wheeler, which might be allowed to start its journey sooner if it agreed in advance to get off the road in the case of heavy traffic.
Frank delivered her proposal along with support from a group of big-name and deep-pocketed stakeholders, including former Loans Program Office chief Jigar Shah, renewables developers like Cypress Creek Renewables, and a number of datacenter developers and technology providers.
In comments on the proposal, Agentic Infrastructure, which works on powering data centers, said that Frank’s plan will allow for “private capital investment to energize with dispatchable service ahead of the timeline required for expansion of firm network service,” which would ensure that “the risks of serving rapid load growth are managed privately while the economics benefits of load growth are socialized to the public rate base.”
In other words, more users of electricity would come online faster, allowing them to make payments to utilities and split up fixed costs among all customers, while the developers would take on the risk of not always being able to power their data centers.
In a best-case scenario, the proposal could be approved at an ERCOT board meeting early next year, Frank said.
Allowing flexible large loads to connect faster is “the most viable way for loads to actually invest with their complex webs of financing and technology partners in creating dispatchability,” she added.
“Everyone is talking about” how important dispatchability is, Frank told me, but “no one is doing anything about it, except for the proposal at PJM and random one-off deals that folks like Google are doing.”
“What makes ERCOT different,” Frank said, “is that it is a place that gets national attention, and it can get national attention because things generally just happen faster there.”
