Texas, ERCOT, and the AI Siting Reset

How big is the ERCOT load wave #

ERCOT's December 2025 Capacity, Demand, and Reserves report and the parallel long-term load forecast filed with the PUCT marked the largest single revision in the system's history. Peak demand in 2031 was raised to roughly 137 GW from 109 GW a year earlier, with the incremental 28 GW dominated by what the operator now classifies as Large Flexible Load, a category created in 2024 that captures crypto miners and behind-the-meter campuses, and a parallel category of grid-tied data centers that interconnect at transmission voltage. ERCOT's separate large-load study, posted in early 2026, reports an active interconnection queue holding more than 130 GW of nameplate requests, of which the operator estimates 30 to 40 GW are likely to materialize by 2030 once duplicate filings, speculative requests, and unfunded projects are screened out.

The geographic concentration is unusual. Roughly half of the materializing load sits in three clusters: the Abilene and Sweetwater corridor along Interstate 20, the Permian basin west of Stanton and Sterling City, and a second tier around Pearland south of Houston and Temple between Austin and Waco. The Abilene cluster alone, anchored by the Stargate site disclosed by OpenAI and Oracle in early 2025 and a separate Microsoft campus at the former Lone Star Industrial Park, accounts for at least 4 GW of confirmed near-term demand once initial phases come online through 2027. Pearland's growth is driven by Meta and a hyperscaler whose name has not been publicly confirmed, layering onto an existing data center submarket that already runs above 1.5 GW. ERCOT itself is now the binding constraint on growth more than any specific developer: the operator's 2026 transmission planning study identifies the Far West and West weather zones as the principal injection points needed to relieve congestion that already shows up in real-time prices, with capital cost estimates for the priority project list in the 30 billion to 35 billion dollar range through 2035. Those costs flow to ratepayers under the traditional ERCOT cost allocation framework, and SB 6 was the legislature's response to the public and political pushback that allocation provoked.

Senate Bill 6 and the new interconnection rule #

SB 6 is the most consequential restructuring of the ERCOT large-load regime since the Critical Infrastructure Subdivision rule of 2014. Three elements matter most. The bill establishes a separate large-load interconnection process for any new load above 75 MW, with mandatory disclosure of behind-the-meter generation, on-site storage, curtailment capability, and projected ramp profile. Second, it gives ERCOT and the PUCT explicit authority to direct curtailment of qualifying large loads during Energy Emergency Alert conditions, including Stage 1, before firm load shedding is initiated. Third, it directs the PUCT to develop a transmission cost allocation methodology that assigns a higher share of incremental network upgrade costs to large new loads rather than spreading them uniformly across all retail customers.

The rulemaking that implements SB 6 is moving in two phases. Phase 1, finalized by the PUCT in late 2025, set the disclosure templates and the curtailment protocol. Phase 2, on which the PUCT issued a strawman in March 2026 with a final order expected by Q3 2026, addresses the cost allocation question and the treatment of behind-the-meter generation that may at times export to the grid. The Phase 2 order will determine whether new West Texas data center loads pay something close to incremental cost causation, which advocates from the residential and small commercial customer classes are pushing for, or whether a hybrid socialization framework survives.

The practical effect on developers is already visible in interconnection queue behavior. Withdrawals from the ERCOT generation interconnection queue and the new large-load queue rose noticeably in Q1 2026 compared to Q1 2025, concentrated in projects that filed without site control, without firm gas supply, or with financing structures that assumed pre-SB 6 cost socialization. Hyperscalers with strong balance sheets and clean gas supply, by contrast, have accelerated filings to lock in interconnection priority before the Phase 2 cost rules tighten further.

Permian gas-to-grid economics and behind-the-meter siting #

The single most important number behind the West Texas data center build is the Waha hub gas price relative to Henry Hub. EIA data and S&P Global Commodity Insights price assessments show Waha basis routinely settling between negative 1.50 and negative 4.00 dollars per MMBtu through 2024 and 2025, with multiple intervals of negative absolute prices when associated gas production from oil-directed Permian wells overwhelms takeaway pipeline capacity. That means a behind-the-meter gas turbine in the Permian, fueled by stranded associated gas under a long-dated contract, can produce electricity at a fully loaded cost in the 30 to 45 dollar per MWh range, well below the average ERCOT West zone day-ahead price for 2025, which the operator's annual market report places near 47 dollars per MWh.

This is the arbitrage anchoring Stargate Abilene, the Crusoe and Lancium behind-the-meter operations near Stanton, and a growing roster of smaller campuses near Sterling City. The structure is consistent across sites. A midstream partner contracts to deliver associated gas at a Waha-indexed price, often with a floor at zero and a ceiling well below Henry Hub. A turbine package, typically aeroderivative units in the 30 to 50 MW class for early phases and Frame 7HA or 6F class for later expansions, generates power on site. A portion of capacity may be grid-synchronized for resiliency, but the campus is structurally islanded from ERCOT for primary load. Behind the meter siting also sidesteps the SB 6 large-load curtailment regime, since ERCOT operational authority attaches to the grid interconnection rather than the load itself.

The risk is not gas price but gas reliability. Permian associated gas is a byproduct of oil drilling, and a sustained downturn in WTI prices that slows Permian crude activity would tighten gas availability at exactly the moment a data center needs firm fuel. Several developers are responding by layering in long-term Henry Hub-indexed firm transportation as a backstop, which raises the all-in delivered cost back toward 5 to 7 dollars per MMBtu but preserves the behind-the-meter advantage on speed and on freedom from SB 6 curtailment authority.

ORDC scarcity pricing and ECRS exposure for grid-tied loads #

Front-of-meter loads pay for ERCOT in two ways that behind-the-meter loads do not. The Operating Reserve Demand Curve adds a price adder that scales with system reserves, capped at the system-wide offer cap of 5,000 dollars per MWh, and the ERCOT Contingency Reserve Service ancillary product, introduced in 2023, has settled at sustained premiums during tight conditions. ERCOT and PUCT data show ECRS clearing prices spending material time in the 50 to 200 dollar per MWh range during summer 2024 and 2025 evening peak hours, with several intervals above 1,000 dollars when frequency response was stressed.

For a hyperscale load with a flat demand profile, this means the headline day-ahead price understates true delivered cost by a meaningful margin. Bilateral hedging through retail electric providers can dampen the volatility but does not eliminate it, since extreme events outside the hedge window pass through. The combined effect of ORDC and ECRS adders during summer scarcity, plus the SB 6 curtailment authority that exposes loads above 75 MW to involuntary interruption, has shifted hyperscaler underwriting in Texas decisively toward behind-the-meter and toward dual-fuel siting where a campus can run from on-site gas during peak hours and grid power off-peak.

Grid-tied developers still chasing ERCOT capacity are increasingly proposing to bring their own firm generation as a condition of interconnection, sometimes called 'BYOG' or bring-your-own-generation. This is in essence a private capacity market for new load, structured outside any formal ERCOT mechanism, and several Phase 2 rulemaking comments have urged the PUCT to formalize it as a third interconnection track alongside the standard front-of-meter and behind-the-meter pathways.

Texas Energy Fund and the gas peaker buildout #

The Texas Energy Fund, created by the legislature in 2023 and capitalized at 5 billion dollars in its first tranche, was designed to seed dispatchable thermal capacity, primarily natural gas combined cycle and peakers, through 3 percent low-interest loans administered by the PUCT. The first round of TEF awards announced in 2024 and finalized through 2025 cleared roughly 9.7 GW of new combined cycle and combustion turbine projects, concentrated in the Houston, Corpus Christi, and Permian load centers. Construction lead times for the awarded projects run from late 2026 for the simple cycle units to 2028 and 2029 for the combined cycle plants.

The TEF projects matter for AI siting in two ways. First, the new combined cycle capacity adds firm dispatchable generation that helps absorb the 30 to 40 GW load wave without triggering more frequent EEA conditions, which would in turn trigger SB 6 large-load curtailment. Second, the existence of TEF financing creates a clear separation between developers who can credibly pair a load filing with a generation filing, and those who cannot. Hyperscalers and their generation partners are using the TEF round 2 application window, which the PUCT opened in early 2026, as a vehicle to anchor specific data center campuses to specific gas plants under co-located structures.

The boundary between behind-the-meter and front-of-meter is becoming porous as a result. A campus may have a 500 MW gas plant nominally selling power to the grid, with an adjacent 400 MW data center load nominally buying from the grid, but with both sides under the same parent and a same-bus structural hedge. The PUCT Phase 2 rulemaking is the first place this hybrid structure will be regulated in detail.

Cluster trajectories: Abilene, Pearland, Temple, Permian #

The Abilene corridor, including the Stargate site disclosed by OpenAI, Oracle, and partners in January 2025, has become the most concentrated AI campus in North America. The first phase, in service through 2025 and 2026, runs at roughly 1.2 GW of IT load, with announced expansion to a multi-gigawatt envelope through 2027. The site's siting choice was driven by three factors: proximity to the West Texas wind and solar build that anchors the ERCOT renewables peak, access to Permian gas via existing intrastate pipeline laterals, and a workforce base in Abilene that scales beyond what smaller West Texas towns can credibly support.

Pearland, south of Houston, has emerged as the principal Gulf Coast AI cluster, hosting Meta's announced campus and adjacent hyperscaler projects, with cumulative announced load in the 1.5 to 2.5 GW range through 2028. Pearland benefits from existing high-voltage transmission corridors that move offshore wind and Gulf Coast gas-fired generation, and from a deeper construction labor market than West Texas. The constraint is summer humidity and water availability, which raises cooling water and chiller plant costs relative to the dry West Texas siting. Temple, between Austin and Waco, is the third cluster, smaller in announced load but with several gigawatt-class proposals filed through ERCOT's large-load queue in late 2025 and early 2026, attractive for its central Texas transmission position and proximity to Austin's tech labor market without Austin's land cost. The Permian itself, anchored by Stanton and Sterling City, is the most behind-the-meter-intensive cluster, almost entirely structured around stranded gas economics rather than grid supply.

The table below summarizes confirmed and credibly announced load across the four clusters, drawing on hyperscaler sustainability disclosures, ERCOT interconnection queue filings as of early 2026, and PUCT proceedings.

What to watch through 2027 #

Three decision points will shape the next 18 months. The PUCT Phase 2 SB 6 final order, expected by Q3 2026, will fix the cost allocation rule and the treatment of hybrid behind-the-meter and front-of-meter campuses. A strict cost causation outcome accelerates the migration to fully behind-the-meter siting and slows grid-tied projects that depend on socialized transmission upgrades. A hybrid outcome preserves the current pace of front-of-meter growth but raises retail rates and concentrates political risk for the next legislative session.

TEF round 2, which closes for applications in mid-2026 and clears later in the year, sets the next generation buildout pipeline. A round that clears another 8 to 10 GW of dispatchable capacity meaningfully eases reserve margins through 2030 and reduces SB 6 curtailment risk. A round that under-clears, whether due to interest rate movement or supply chain constraints on turbines, leaves the system tighter and pushes more loads behind the meter as a hedge.

The third question is whether one or more hyperscalers commit publicly to a multi-gigawatt expansion outside Texas in 2026 and 2027, signaling that the SB 6 regime, the curtailment authority, and the cost allocation outcome have shifted the marginal siting decision toward Virginia, Ohio, the upper Midwest, or international jurisdictions. The first signal here will not come from policy but from sustainability and capital expenditure disclosures in hyperscaler annual reports through 2026. Texas remains the largest near-term opportunity for AI compute siting in North America by a wide margin, but the era of unconstrained, cost-socialized growth is over. The clusters that compound from here will be the ones that combine clean Permian gas economics, financeable TEF-anchored generation, and a credible answer to the SB 6 curtailment regime. Everyone else will spend the next two years in the queue.

Sources #

• Capacity, Demand, and Reserves Report and Long-Term Load Forecast, December 2025
• Large Load Interconnection Study, Phase 1 Report
• Senate Bill 6, 89th Texas Legislature, Regular Session, 2025
• PUCT Project 56822, Implementation of Senate Bill 6
• Texas Energy Fund Loan Program, Round 1 and Round 2 status
• Henry Hub and Waha Hub natural gas spot prices
• Permian Basin Drilling Productivity Report
• Stargate announcement, January 2025
• 2025 Sustainability Report and data center disclosures
• 2025 Sustainability Report and data center disclosures
• Lazard Levelized Cost of Energy and Storage 2025
• Texas grid load surge from AI data centers