The four groups are distributed across two regions: two are located in North and North-Central Texas, and two are in West Texas. None currently operates above 3,200 MW on an individual basis, but their automated protection systems are the source of concern. Unlike traditional industrial customers that ramp consumption gradually, data centers and crypto mining operations rely on automated protection systems that instantly disconnect upon detecting voltage disturbances. That protects the equipment but can strip thousands of megawatts from the grid in fractions of a second, effectively amplifying a fault rather than absorbing it, in the framing of the working group's analysis.

ERCOT reviewed approximately 20 gigawatts of large customers seeking grid connection as part of the same working group review. Eight projects totaling about 3.9 GW are targeting connection before July 1, right at the start of peak summer demand. The grid operator is forecasting a record summer peak of 92,211 MW this year, nearly 10 percent above last summer's figure and well above the all-time record of 85,464 MW set in 2023. Since September 2025, Texas has added roughly 0.5 GW of new crypto mining capacity to the grid, and an additional 2 GW of large industrial load, including data centers and crypto miners, is projected to connect between May and September 2026.

A December 2022 incident is shaping how grid planners read the current risk. A single transformer failure that month caused nearly 400 facilities to disconnect simultaneously without warning. ERCOT has also documented, in prior grid reliability reports, that crypto miners in particular "exhibited inconsistent behavior during resource scarcity events that have brought the grid perilously close to failure," a pattern tied to how miners respond to Bitcoin price movements. When BTC prices are high relative to power costs, miners resist curtailment requests. When prices drop, they disconnect abruptly and in volume. Not all experts share the same level of concern about overall summer preparedness, however. Ed Hirs, an energy fellow at the University of Houston, offered a more measured read on seasonal capacity: "ERCOT has plenty of available capacity to get us through the summer. The huge growth in wind and solar generation on the Texas grid over the past five years is really the reason."

The regulatory apparatus around large loads is being rebuilt in real time. Texas Senate Bill 6, signed in June 2025, gives ERCOT authority to forcibly disconnect data centers and other large loads of 75 MW or more during emergency events. It also requires them to pay a non-refundable interconnection fee of $50,000 per megawatt of peak demand and to fund 100 percent of their own direct interconnection infrastructure. The Public Utility Commission of Texas published a draft rule, TAC §25.194, in March 2026 to implement SB6's interconnection standards, with the public comment period closing April 17, 2026. On the standards side, ERCOT is proposing to replace the category "Large Electronic Load" with "Large Computational Load" under draft rule PGRR144, reflecting new dynamic modeling requirements. At the federal level, NERC launched Project 2026-02 in parallel, a dedicated initiative to set registration criteria and reliability standards for what it is calling "Computational Load Entities," covering data sharing, interconnection studies, protection coordination, and operational communication. ERCOT has also launched its Batch Zero interconnection review, described by Jeff Billo, ERCOT's Vice President of Interconnection and Grid Analysis, as "one study to rule them all." Rather than evaluating projects one at a time, Batch Zero conducts a system-wide batch evaluation every six months, a mechanism designed to address the cumulative grid stress that sequential individual studies can obscure. One concern on record is that projects which have been pending in the queue for 18 months or more, including large loads backed by companies such as Meta, may not qualify for Batch Zero and would continue under legacy study processes. ERCOT's interconnection queue now holds roughly 410 GW of pending large-load requests; as recently as February 2026 that figure stood at over 250 GW, a pace of growth that itself reflects the urgency of the challenge. About 87 percent of the queue is tied to data center projects, many of them AI infrastructure. AI companies now account for roughly 73 percent of new large-load applications in Texas, up sharply from the crypto-dominated queue of 2022 through 2024.

International Parallels

Pakistan's government announced plans in 2025 to redirect 2,000 MW of idle electricity toward Bitcoin mining and AI infrastructure, a scale that mirrors what Texas is managing but with far thinner grid margins and without ERCOT-equivalent voltage ride-through requirements. The IMF blocked Pakistan's attempt to offer subsidized electricity to miners, citing market distortion. Pakistan's Virtual Asset Act 2026 addresses financial compliance, requiring licensed miners to declare electricity sources, comply with anti-money-laundering rules, and meet cybersecurity standards, but analysts note the legislation does not yet establish technical grid stability standards for mining loads. In East Africa, Ethiopia draws about 600 MW for Bitcoin mining from Grand Ethiopian Renaissance Dam surplus, and operators like Gridless run hydro-microgrid mining sites across Kenya, Malawi, and Zambia. Those grids are too small to face destabilization at current mining scale, but the Texas lesson is relevant: waiting until failures occur to build a modeling framework is far more expensive than requiring it at the licensing stage. India presents a third case. The country has pursued heavy crypto taxation without equivalent regulatory clarity on mining energy usage, leaving large-scale operators in legal grey zones. Indian grid operators have not yet developed ERCOT-equivalent large-load modeling standards, even as data center demand surges alongside AI infrastructure investment.

The network-level risk extends beyond any single country. Texas hosts a substantial share of global Bitcoin hashrate. A winter storm in February 2026 already demonstrated the exposure: Foundry USA lost roughly 60 percent of its capacity at peak, pulling the global hashrate down to 826 EH/s from an all-time high above 1,000 EH/s reached in January. Hash price, as of the Q1 2026 close, was near multi-year lows at approximately $23.90 per petahash per day. A summer grid failure in Texas triggered by cascading disconnections of the kind the May 21 report describes could trigger a negative mining difficulty adjustment, compressing margins for operators worldwide, from Kazakhstan to Ethiopia and across Southeast Asia. The regulatory framework ERCOT has assembled, including Batch Zero, is designed to get ahead of the problem. The voltage test failures suggest the timeline is tight.