Texas Grid's 8 Gigawatt "tickets" are all distributed! Bernstein: Under power scarcity, miners like IREN will become AI infrastructure landlords and "collect rent"

LongbridgeAII'm LongbridgeAI, I can summarize articles.

The Texas grid ERCOT has completed the first batch of 8 gigawatts of baseload power allocation, indicating a scarcity of AI infrastructure power resources. Bernstein's research report points out that the scarcity of baseload sites will drive up rental prices, favoring miners like IREN and CIFR to transform into "landlords" for AI infrastructure to obtain rental income. Among them, IREN stands out as the biggest winner with 2 gigawatts, while some projects of Cipher Digital have been downgraded to research load

According to the latest research from investment bank Bernstein, approved electricity is becoming increasingly valuable, and the scarcity of base load sites will continue to drive up rental prices.

The Texas grid operator ERCOT has completed the first round of "Batch Zero" screening, with emerging AI infrastructure developers disclosing approximately 8 gigawatts of conditional base load classification, and another approximately 7 gigawatts listed as "studied load."

ERCOT's "Batch Zero": A Ticket to the Grid

ERCOT (Electric Reliability Council of Texas) Batch Zero is the first project queue for allocating available transmission capacity under a system-wide research framework. Projects are divided into three categories: base load—mature projects that are operational or supported by qualified preliminary studies, which will no longer accept further evaluation and capacity reallocation; studied load—capacity and timing must be determined through system-level analysis, with no guarantee of full realization of declared capacity; excluded projects—reserved for subsequent batches, with the first batch (Batch 1) expected to begin applications in the summer of 2027.

ERCOT completed the first round of screening on September 3 and issued temporary classification notices to transmission and distribution service providers (TDSP). The official verification report is expected to be released in December 2026. Notably, a number of projects were "downgraded" during the screening: some projects originally declared as base load were reclassified as studied load, indicating that base load allocations are lower than previously expected—Cipher Digital (CIFR.US) declared Batch Zero candidates with 3.6 gigawatts but currently only received 1 gigawatt of incremental approval. These classifications remain conditional and must meet requirements such as audits by Governor Abbott.

Who Got the "Ticket"?

AI cloud infrastructure service provider IREN (IREN.US) has its Sweetwater park qualifying for base load with a total of 2 gigawatts, making it the largest single winner; Hut 8 (HUT.US) with its 1 gigawatt Beacon Point park (not within coverage), Galaxy Digital's Helios Phase I (800 megawatts) and Phase II (830 megawatts) also all made the cut. CIFR received 1.1 gigawatts of base load (1 gigawatt Colchis plus 100 megawatts Stingray), but another 2.1 gigawatts (including 900 megawatts Apollo) only secured studied load. Core Scientific (CORZ) has Hunt (431 megawatts) and Pecos (300 megawatts) included in base load; CleanSpark (CLSK.US) has both its 285 megawatts Sealy and 300 megawatts Brazoria fully approved for base load; MARA (MARA.US) has Granbury (approximately 280 megawatts) and Garden City (approximately 200 megawatts) included, but the up to 2 gigawatts Matagorda has been classified as studied load

Scarcity Drives Up Rent

The report estimates that in the 8 GW base load classification, over 5 GW is available for new AI deployment contracts, including IREN's 2 GW park and CIFR's 1 GW Colchis. Against the backdrop of increasing political scrutiny on new data center development, capacity bottlenecks created by moratoriums and state-level directives, the value of approved capacity continues to rise.

Bernstein believes that the evolution of grid rules is actually a form of "unclogging": it pushes speculative queue applications out of the line, making sites with a real development history more valuable. The allocation of research load will not be finalized until April 2027 at the earliest, further strengthening the competitive position of operators with conditional base load.

In terms of investment conclusions, Bernstein maintains an "outperform" rating for Cipher Digital (CIFR.US), TeraWulf (WULF.US), IREN (IREN.US), Core Scientific (CORZ.US), CleanSpark (CLSK.US), and Riot (RIOT.US), with target prices of $32, $36, $100, $32, $35, and $24 respectively; MARA is rated "in line with the market," with a target price of $17.

The report emphasizes that crypto miners, with a planned power portfolio of about 32 GW and the ability to deliver "warm start power shells" on time, are best positioned to address the AI industry's pain point of "time to power." Over the past two years, miners have signed contracts worth over $180 billion, totaling about 9 GW and more than 20 agreements, with major cloud providers, cutting-edge AI laboratories, new cloud service providers, and AI chip manufacturers.

Not Affected by AI Slowdown

It is important to note that the market environment in which this report was released is not calm. On one hand, the debate about the "cooling" of AI investment is still brewing—from the "SaaS apocalypse theory" in the SaaS sector to doubts about the sustainability of massive capital expenditures, investors are repeatedly questioning the real demand for AI.

However, Bernstein's logic precisely bypasses this debate: regardless of who wins or loses on the model side, electricity is the hardest constraint. This week, Amazon locked in $340 million in warrants for Generac backup power, and previously reached a chip agreement with Qualcomm, both confirming the willingness of giants to pay the price for "grabbing electricity."

Source: Zhitong The copyright of this article belongs to the original author/organization.

The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.