When Texas Governor Greg Abbott signed Senate Bill 6 into law last summer, rewriting the rules for how large loads connect to and operate on the grid, it made ERCOT the country’s most visible laboratory for managing data center load growth.
The core philosophy of SB6 is simple, and has echoed as far as major tech companies and utilities, as well as the White House: Data centers should pay their own way for grid upgrades, and stand down during emergencies so that ratepayers don’t end up footing the bill.
But as the rulemaking process for SB6 has unfolded, it’s becoming clear just how far apart regulators and the AI industry are when it comes to the price of grid admission. Other markets around the country are closely watching the back-and-forth between Texas and developers over the size of financial commitments, and whether to grandfather in existing projects, explained Anne Liu, the research lead for ERCOT at Aurora Energy Research.
“I think this is going to significantly impact how other markets are going to [manage large loads],” she said. While ERCOT isn’t directly regulated by the Federal Energy Regulatory Commission, federal policies often reflect the trends that start in Texas’ islanded grid, she added.
ERCOT’s approach to connecting large loads — which mimics the process for generator interconnection — is simple and familiar enough that other markets are likely to mirror it.
Paying your own way is expensive
In Texas, the work to implement SB6 is playing out on five different topic tracks at the public utility commission: interconnection standards, net-metering arrangements for co-location, large load forecasting, demand reduction, and transmission cost allocation. Each is operating on a slightly different timeline, but all are expected to land on final rules by the end of 2026.
Furthest along is the foundational rulemaking on interconnection standards. The PUC published proposed rules in mid-March that would set steep financial security and project disclosure requirements for loads over 75 megawatts — including those projects that were already in the queue before the passage of SB6. The proposal is almost certain to change; comments from developers continue to pour in ahead of April deadlines.
Before a project can begin the official interconnection study, a developer must disclose extensive information about their site, such as property interests like deeds or leases, load ramp schedules, and what type of backup generation they intend to use. They must then pay a study fee of between $100,000 and $300,000, depending on the project size, and post a financial deposit of $50,000 per MW. Notably, that amount is down from the $100,000 per MW security deposit initially proposed, but the scale remains daunting for many developers.
As Monarch Energy, a green hydrogen-turned-data center developer, pointed out in the docket, a 500 MW project hoping to connect in ERCOT would have to post $25 million just to find out if the grid can support it.
Once the study is completed, developers have 30 days to decide whether to move forward. Doing so requires executing a final interconnection agreement, paying a non-refundable interconnection fee of $50,000 per MW (this can come from the deposit) as well as 100% of the direct construction costs for their specific hookup, and posting additional financial security for any required system upgrades or long-lead-time equipment.
If a project withdraws, it loses 80% of its remaining security balance, which the transmission service provider would then keep as an offset to its rate base. If a review concludes that the grid can’t support a particular project, the study fee is non-refundable. If a project does come online and use the electricity it planned for, the remaining security balance is refunded over time with scheduled load milestones.
In early responses, developers are pushing back against nearly every line item. They argue the 80% forfeiture rule, for example, is a “penalty” that isn’t related to actual costs a project puts on the system. Instead, they propose that 20% of the security deposit be nonrefundable in the case of withdrawal.
Google and data center developer Lancium (whose projects include the Stargate campus in Abilene), submitted a response earlier in March suggesting capping the non-refundable portion of the security deposit at $6.5 million, which only becomes non-refundable after the initial study shows a project to be viable. As proposed, the rules put too much financial risk on customers too early in the process, they explained, adding that developers want to see the technical results of an interconnection study before writing massive, non-refundable checks mandated in the current draft rules.
The filings also reflect a concern that Liu, at Aurora, said comes up frequently in conversations: retroactivity. ERCOT is in the middle of implementing a transition to its new batch review process for large loads, which it is calling “Batch Zero”. Monarch, Google, and others are urging the PUC against applying the new interconnection standards to this initial batch of projects, arguing that doing so would be unfair to those that have already spent years and millions of dollars on development.
Beyond ERCOT
Of course, Texas is far from the only place writing new rules for data centers, but it’s the primary example of a purely legislative approach.
A recent Latitude Intelligence study analyzed 25 data center tariffs around the country, and found wide variation in protections — but a clear trend toward upfront payments and demand commitments. In general, these tariffs are highly focused on ratepayer protection and require more upfront cash than the current SB6 proposals. Dominion Energy’s tariff in Virginia, for example, demands $1.5 million per MW in collateral.
These tariffs also tend to have higher billing floors (the amount of contracted peak demand a data center must pay for, regardless of how much energy it consumes) than SB6. What remains to be seen, wrote analyst Nick Zenkin, is whether competitive markets like ERCOT — where deal terms are usually confidential — can deliver the ratepayer protections that regulated tariffs provide.
At the market level, SPP has managed to avoid the drawn-out rulemaking process in which Texas currently finds itself by skipping the legislative route entirely and doing it all through FERC filings. The region’s High Impact Large Load process takes a similar approach to SB6, including leveraging clustered studies and requiring deposits to deter speculation.
Perhaps the most visible comparison to ERCOT’s work, however, is PJM, home to “Data Center Alley,” which has been considering large load interconnection via its Critical Interconnection Fast Path initiative since last summer.
“Everybody is looking to PJM on the one hand and ERCOT on the other hand, as two areas where this has already come to a head, to try to come up with some sort of solution,” explained Julia Hoos, Aurora’s head of USA East. ERCOT’s advantage lies in its isolation, she added. Because it is overseen by a single state PUC, it can synchronize interconnection standards, cost allocation, and demand response in one go. PJM, meanwhile, must simultaneously solve for capacity and deliverability, under much stricter federal standards.
PJM is juggling individual, and often conflicting, orders from FERC, the White House, and a coalition of state governors, she said. “This is part of the current challenge — there are so many questions that are actually deeply interrelated,” Hoos explained. “We are getting individual orders to solve them one at a time, and while that might speed things up in a vacuum, it adds massive inefficiency to the system as a whole.”
That said, the rest of the country is likely to look more like PJM than ERCOT, Hoos said, layering large-load rules on top of capacity markets and FERC mandates — in order to make the physics and the politics work.
Transition-AI 2026, Latitude Media’s flagship event at the center of the AI-energy infrastructure buildout, will convene developers, utilities, regulators, hyperscalers, and capital providers to move towards coordinated, sustainable solutions for AI-driven demand. Register now to secure the lowest rates and join us in San Francisco April 13-14.
