Written by: Chaoxiang Research

The Democratic candidate won a surprise victory in the Texas gubernatorial election, putting a pause on the largest AI power tap in the U.S.

Morgan Stanley analyst Stephen Byrd pointed out that the power to restrict data center expansion is almost entirely in the hands of the states. The newly elected governor has committed to halting new construction approvals, meaning that 474GW of grid connection applications in the Texas grid face extensive delays or even cancellations.

This local election shattered the illusion of unlimited supply of AI power. Computing power projects without readily available electricity face standstill, while existing power plants that have secured electrical capacity are seeing unprecedented price premiums.

Local executive orders directly block grid connection channels

The power of the White House and Congress does not reach local transformers.

The report points out that the real leverage that constrains data centers from being established lies not in Washington, but within state utility commissions, legislatures, and county-level administrative agencies. The new Texas Governor Gina Hinojosa explicitly advocated for a moratorium on new data center approvals during her campaign.

Once in office, the Governor can independently utilize executive orders to require state-level agencies to suspend the issuance of unapproved permits without the need for approval from the Republican-controlled legislature.

This directly locks down the Texas Electric Reliability Council's (ERCOT) heavy load grid connection process. Previously, Texas had already paused some connection approvals to clear the queue of the power grid, with the applications in the queue amounting to as much as 474GW. This figure is more than five times the historical peak load of Texas, with data centers accounting for 90% of the new electricity demand.

Following the executive order, the ongoing Batch Zero transmission planning review may face a complete halt. The timeline for unstarted projects to obtain power connection permits will extend indefinitely from the originally expected months.

The computing power landscape is splitting into two camps

Policy risk is dividing the U.S. into two distinctly different regions: states welcoming computing power and states driving it away.

In Ohio, the election of Democratic candidate Amy Acton as governor has brought the same friction on ground. The most notable obstructed project in the state is a 10GW giant computing power cluster planned by NVIDIA in collaboration with a leading U.S. AI lab. Amazon's 1.5GW project under construction in Fayette County also faces the impact of tightening tax incentive policies.

However, in Pennsylvania and Louisiana, policy resistance is comparatively manageable. Pennsylvania Governor Shapiro implements a standard certification mechanism based on the affordability of the power grid and environmental protection, without directly halting project approvals.

This sharp divergence in state policies is completely changing the site selection logic for AI computing power construction. Developers who previously valued Texas for its low electricity prices and loose regulations are now finding that the uncertainty of the approval process has overshadowed simple cost considerations. The overall macro scale of AI capital expenditures across the U.S. may not shrink, but it is being forced to restructure geographically.

Existing assets are on average 30% more expensive than stock prices

With new projects stuck in approval limbo, independent power producers (IPPs) holding certain electrical assets are seeing a sudden rise in intrinsic value.

Morgan Stanley has reassessed the existing assets of core power stocks. This assessment only takes into account the cash flows from existing assets, signed contracts, and forward electricity price curves, completely excluding any potential premium from future contracts signed by data centers.

The results show that the intrinsic value of existing assets is far higher than the closing prices on November 3:

Vistra's (VST) existing assets have an intrinsic value of about $180 per share, which is 28.57% higher than the closing price of $140.

NRG Energy's (NRG) existing assets have an intrinsic value of $129 per share, 34.38% higher than the closing price of $96.

Talen Energy's (TLN) existing assets have an intrinsic value of $430 per share, 20.45% higher than the closing price of $357.

Constellation Energy's (CEG) existing asset value is $249, slightly below the closing price of $263, but this valuation entirely excludes potential data center nuclear power direct connection contracts in Pennsylvania's PJM market. If we account for expectations of future nuclear power contracts, the firm maintains its target price of $378.

The market is currently extremely pessimistic about the pricing of independent power producers, having not only failed to give any premium valuation for future computing power supply contracts but also discounting the basic cash flows of existing power plants. As long as the pricing power of existing units is realized, these targets have solid value support.

The computing power infrastructure chain is accelerating its divergence

Targets also riding the AI power wave are affected to varying degrees by policies.

Data center service providers (PSPs) that purely rely on the Texas grid for expansion face the brunt of these effects. Galaxy Digital's (GLXY) planned Helios II project (830MW), which originally had a 95% probability of timely production, now faces reevaluation. Cipher Mining's (CIFR) new site leases, such as Colchis, also require significant downward adjustments in probability, with a pessimistic scenario in their stock price valuation model without new projects being just $5 per share.

In contrast, the advantages of behind-the-meter (BTM) power supply models, which are not restricted by the Texas grid, are becoming more pronounced.

Williams Company (WMB) utilizes its nationwide natural gas pipeline resources to provide gas turbine self-generation power plant solutions for hyperscale cloud service providers. By pre-booking gas turbines, WMB can deliver electricity directly within 18 to 24 months after contract signing, completely bypassing the public grid connection process that has been waiting for years.

Bloom Energy's (BE) solid oxide fuel cells also feature zero emissions, no water consumption, and low noise characteristics. In environmentally strict states led by Democrats, fuel cells are becoming a direct alternative to large gas turbines.

TeraWulf's (WULF) main planned project is located in Kentucky, where it previously successfully responded to construction limitations in New York. The management's experience in cross-regional navigating of administrative approvals keeps it from being dragged down by Texas policies.

Institutional accounts

Chaoxiang interpretation

The weakest assumption of this report is equating the governor of Texas' construction limits with a permanent cessation of actual approvals. The Republican-led state legislature in Texas is deeply intertwined with powerful oil, gas, and industrial interest groups. A complete blockage of computing power investments would severely impact the state’s finances and employment.

If the new governor cannot establish legislation to normalize the construction limitation rules within 12 to 24 months, many backlogged projects will still be restarted through additional minimum electricity price clauses after the executive order freeze period ends. Readers do not need to focus on national AI computing power sentiment; they only need to monitor the Texas Public Utility Commission (PUCT) for appointments to elected board positions in 2027, as well as the final release list after audit of ERCOT's first batch of heavy load transmission planning (Batch Zero).

Disclaimer

This article is a compilation and interpretation of third-party brokerage research reports by Chaoxiang Research. The ratings, target prices, earnings forecasts, and related judgments cited in the text are solely the views of analysts from that brokerage and do not represent the views of Chaoxiang Research, nor do they constitute any investment advice.

Markets are risky; decisions should be made independently. This article should not be used as a basis for buying or selling any securities.