Featured Topics
Featured Products
Events
S&P Global Offerings
Featured Topics
Featured Products
Events
S&P Global Offerings
Featured Topics
Featured Products
Events
S&P Global Offerings
Featured Topics
Featured Products
Events
Financial and Market intelligence
Fundamental & Alternative Datasets
Government & Defense
Professional Services
Banking & Capital Markets
Economy & Finance
Energy & Commodities
Technology & Innovation
Podcasts & Newsletters
Financial and Market intelligence
Fundamental & Alternative Datasets
Government & Defense
Professional Services
Banking & Capital Markets
Economy & Finance
Energy & Commodities
Technology & Innovation
Podcasts & Newsletters
Research — May 5, 2026
By Dan Thompson and Tony Lenoir
Highlights
With hyperscalers chasing all the energy they can to fuel their AI ambitions, the rising levels of curtailed solar and wind power within the Electric Reliability Council of Texas represent an electricity surplus opportunity for the sector. Curtailment, the intentional reduction of electricity production, is often a symptom of a region's transmission constraints. Therefore, strategies to absorb curtailed energy should focus on colocation, adaptable load operations and siting in zones particularly affected by the phenomenon. Although not a silver bullet, data centers pursuing these strategies could lower their power expenditures, contribute to decarbonization and help sustain renewable energy development.
Curtailed renewable energy represents a power-sourcing opportunity for data centers. With transmission and distribution infrastructure struggling to keep up with the rapid deployment of solar and wind generation capacity in Texas, the Electric Reliability Council of Texas (ERCOT) stands out, curtailing 9.8 terawatt-hours of renewable energy that could have been used in 2025 — up 98% since 2020. This is a win-win for the technology and power sectors: Data centers can benefit from low electricity prices for at least part of their loads, and generators can generate revenue from output that would otherwise have been wasted. Not all data centers are equal within this paradigm, however. Facilities with adaptable workloads and without latency barriers, or those with other versatile operational and siting profiles, have the advantage. That said, even for this category of data centers, parameters beyond power supply must also be considered, including topology, fiber availability, water constraints and community sentiment toward data centers.
Curtailment is the intentional reduction of electricity production. This grid management tool can be activated due to a variety of factors, but the main driver is typically a surge in regional production that transmission capacity cannot fully accommodate. Since the fuel supplies for wind and solar power — air currents and sunlight — are outside the operator's control, these technologies are prime targets for curtailment. They often produce more electricity than the grid can absorb or is needed at a given time, particularly in areas where renewable penetration is high.
Solar is particularly challenging due to the mismatch between periods of maximum production and peak power demand. In comparison, wind theoretically can produce power anytime. While maximum solar production generally reaches its zenith early in the afternoon, peak power demand occurs in the evening hours, when populations — reflecting an average 9-to-5 work schedule — ramp up their home electricity usage, including increased air conditioning use in hot weather.
Curtailment is on a relatively steep upward trajectory in ERCOT, with curtailment of solar and wind, in aggregate, advancing at a 14.6% compound annual growth rate (CAGR) between 2020 and 2025, reaching 9.8 TWh. For perspective, 451 Research, in its base-case scenario for data center utilization rates, estimates that power demand from data centers within the ERCOT footprint stood at approximately 45 TWh in 2025.
A closer look at the data shows consistently higher levels of curtailment for wind in absolute terms between 2020 and 2025. The roles are reversed, however, when examining curtailment as a percentage of estimated total uncurtailed output, with solar oscillating in a 5.4%–8.0% range during the interval under consideration, versus 4.0%–4.9% for wind.
These trends reflect the history of renewable deployment in the region, where wind turbine deployment initially established a sizable lead before being caught up by a surge in new solar plants, with the mechanisms described above amplifying curtailment pressure on the solar side. For additional perspective, solar curtailment increased at a 40.9% CAGR in ERCOT between 2020 and 2025, compared with a 6.3% CAGR for wind. Solar accounted for less than 15% of total annual curtailed output in 2020; by 2025, this figure had risen to nearly 42%.
Western Texas and parts of the panhandle managed by ERCOT have been particularly affected by the rapid expansion of solar generation capacity in the state, with transmission constraints preventing excess generation in these regions from reaching demand centers in the east. A look at development pipelines suggests more of the same for the foreseeable future. With approximately 36 gigawatts in operation, ERCOT already leads all independent system operator regions in solar generation capacity. As of the end of the first quarter of 2026, the region had 101 GW of additional solar capacity in planning — more than 27% of the US pipeline for solar. For perspective, the California ISO has a solar pipeline of just 23 GW.
Identifying and connecting directly with solar and wind generators heavily impacted by curtailment is the most straightforward way for data centers to make use of these available power surpluses. This can be done via a behind-the-meter setup within a power purchase agreement framework, allowing the load to consume the excess generation that the grid cannot absorb, which would otherwise have been lost. With this type of solution, data centers can also circumvent interconnection delays and reduce reliance on constrained infrastructure during congestion hours. That said, unless truly islanded, most data centers still need a connection to the grid to supplement a deficit in on-site generation or export excess.
Soluna Holdings Inc. already relies on the colocation option in Texas. Soluna's 166-MW Project Kati cryptocurrency-mining data center, located in the southern tip of the Lone Star State, is colocated with the 273-MW Las Majadas Wind Project. In May 2025, Soluna announced a deal to colocate its upcoming Project Annie data center with an undisclosed 114-MW solar farm in northeast Texas. Finally, in western Texas, the company's Project Dorothy data center uses behind-the-meter power from an undisclosed 150-MW wind farm, with the setup explicitly designed to capture excess energy that would otherwise be curtailed.
For data centers with "batchable" workloads, such as AI training and cryptocurrency mining, operations can be adjusted to align with solar and wind energy production dynamics, including periods of potential curtailment. By ramping up compute during high production periods, data centers with these profiles not only tap into carbon-free power that would otherwise be lost, but also benefit from low electricity prices that regularly dip into negative territory. This fits perfectly for cryptocurrency mining, which is extremely power-price-sensitive and would be less concerned about periods of downtime for the sake of higher profit margins.
The benefit to AI training workloads, while theoretically flexible, is more controversial. The hyperscalers are investing so heavily in AI, and particularly training, that downtime potential could be a major drawback. Colocated energy storage batteries could be leveraged in this case, with the storage systems charging during periods of excess renewable generation and discharging later; however, the most likely scenario is that the providers would want a grid-as-backup design solution, rather than being totally reliant on wind, solar and batteries — at least for now.
Data companies with an adaptable profile could distribute their workloads geographically, targeting excess renewable energy wherever it may be. Of note, Texas' Senate Bill 6, which was signed into law in June 2025, requires load flexibility for new large loads of at least 75 MW during emergencies; data centers with flexible load operations relying on power that would otherwise be curtailed would arguably be in a better position than most to meet this criterion.
Ultimately, curtailment is a localized phenomenon. Maximizing excess solar and wind energy opportunities entails targeting areas that combine high renewable energy penetration and congested transmission infrastructure. By siting facilities in such territories — often remote and rural — to tap into curtailed solar and wind energy, data center companies can also maximize the degree of autonomy that renewable power provides, unlike natural gas or coal power generation, which requires a continuous flow of fuel to operate. Finally, leveraging cheap would-be-curtailed energy insulates against the fluctuations of international markets, which geopolitical tensions can amplify.
That said, workload profiles matter. Such areas may not be a good fit for data center loads that require maximum uptime and low latency — such as AI inference server-housing facilities — which typically necessitate proximity to urban centers. Beyond electrons, it is possible that topology, fiber availability, water constraints and the physical attributes of existing renewable power plant sites could get in the way of colocation, regardless of the compute workload profile. Finally, stranded investments are a risk if server farms are built around curtailment economics only to see grid congestion relieved.
While adaptable workloads can help sidestep this obstacle in daily operations planning, it is important to keep in mind that intermittency is also often seasonal. Solar power generation, in particular, follows the Earth's orbital revolution around the sun. Combined with varying demand throughout the year — such as low heating or cooling needs in the spring versus intense air conditioning usage in the summer — this results in curtailment peaks and valleys between January and December each year.
ERCOT is no exception; in fact, it may amplify the acuteness of these swings, given Texas's sometimes harsh winters and typically sweltering summers. Examining the 9.8 TWh of total solar and wind power curtailed in 2025 underscores the magnitude of this phenomenon in the region, with curtailment cresting at 1.9 TWh in March and bottoming out at 145 GWh in August, when air conditioning units are taxed to the max. Perhaps surprisingly, wind shows the larger differential between peak curtailment and its lowest point, with the former reaching 22 times the latter in 2025, compared with an elevenfold difference for solar.