Sun, Wind, and Stranded Risk: The Structural Limits of the Southwestern Data Center Model
The previous articles in this series have examined two distinct geographies in the data center market: Northern Virginia, where network…

Sun, Wind, and Stranded Risk: The Structural Limits of the Southwestern Data Center Model
The previous articles in this series have examined two distinct geographies in the data center market: Northern Virginia, where network agglomeration, federal security demand, and decades of targeted infrastructure investment produced a structural lock that commercial economics alone could not have created; and the Rust Belt industrial corridor, where inherited grid stiffness, brownfield permitting advantages, and communities actively seeking reinvestment are positioning legacy industrial districts as the rational geography for the next generation of distributed AI inference infrastructure. Each of those geographies has a coherent logic rooted in the physical and administrative characteristics of the underlying infrastructure.
Texas and Arizona represent a third model. It merits careful examination because the logic that built the Southwestern data center market was sound in its origins but is increasingly strained at its edges. These markets did not grow by accident or by proximity to federal demand. They were chosen deliberately based on deregulated power economics, abundant land, renewable energy procurement advantages, and favorable tax regimes. Those advantages were real, but the Southwestern model also carried structural vulnerabilities that were underpriced during the build-out phase and that are now becoming visible in ways that should inform both operator decisions and regulatory frameworks.
What Built the Southwestern Market
The initial attraction of Texas and Arizona for data center development was straightforward. Land was available at prices that made large campus footprints economically viable, without the parcel-assembly complexity that characterizes infill development in established markets. Power was relatively inexpensive, particularly in Texas, where the deregulated ERCOT market offered operators the ability to procure electricity through competitive contracts rather than regulated utility tariffs. The Southwestern solar resource is among the strongest in the country, and both states offered favorable conditions for long-term renewable energy procurement agreements, allowing hyperscale operators to meet internal sustainability commitments while managing exposure to power cost escalation.
Arizona added a workforce advantage rooted in the growth of Phoenix as a major metropolitan market, with a labor pool large enough to support the technical and facilities-management workforce required by data center operations. The state’s regulatory environment was accommodating, and local governments in the Phoenix metropolitan area developed permitting frameworks that reduced entitlement friction for large industrial projects. Maricopa County, which encompasses most of the Phoenix market, became one of the fastest-growing data center geographies in the country, attracting investment from every major hyperscale operator.
Texas offered a different but complementary set of advantages. The Dallas-Fort Worth metroplex provided a major population center and an existing fiber ecosystem that met connectivity requirements without the full exchange-point density of Northern Virginia. The state’s business climate, tax structure, and political posture toward large industrial investment created a consistently favorable regulatory environment operators could rely on over project timelines measured in years.
The ERCOT Isolation Problem
Texas operates its primary electrical grid, ERCOT, as an island. Unlike the PJM Interconnection serving the Rust Belt and Mid-Atlantic markets, or the Western Interconnection serving Arizona and the broader Mountain West, ERCOT has minimal interconnection with neighboring grids. This design was intentional: it allows Texas to avoid federal jurisdiction under the Federal Power Act by keeping its transmission infrastructure entirely within state boundaries. The practical consequence is that when ERCOT experiences a supply shortfall, it cannot draw on surplus capacity from neighboring systems at scale. The grid must balance itself with the resources it has.
February 2021 demonstrated what that means at the extreme. A sustained cold-weather event simultaneously increased heating demand and caused widespread generator failures, including at natural gas facilities whose fuel-supply infrastructure was not winterized for extended periods of subfreezing temperatures. ERCOT could not import sufficient power from adjacent systems to compensate. The result was rotating outages that affected millions of customers and caused deaths, property damage, and economic losses estimated in the hundreds of billions of dollars. Data centers in the affected areas experienced conditions that their uptime commitments were not designed to accommodate.
The February 2021 event was not an anomaly that has since been fully resolved. Subsequent summer heat events have repeatedly strained ERCOT’s capacity margins, and the fundamental constraint, the inability to draw on external surplus during stress events, persists. The grid has added generation capacity since 2021, but the structural isolation that created the vulnerability remains. For data center operators with contractual uptime obligations and no tolerance for extended outages, this risk profile requires a response that the Northern Virginia or Rust Belt geographies do not demand to the same extent.
The industry’s response has been instructive. Major operators in the Texas market have moved aggressively toward on-site generation and large-scale battery storage, precisely because they cannot rely on ERCOT as a sole source of power during stress events. This capital investment in backup and supplemental generation capacity represents a cost premium that does not appear in the headline power price but is real and substantial. It also previews a structural question the series will address in a later installment: whether small modular reactor deployment, as a source of firm, on-site, carbon-free baseload power, offers a longer-term resolution to the grid reliability problem that battery storage addresses only partially and at high recurring cost.
Water in the Desert
Arizona’s data center market was built, in significant part, on evaporative cooling: the use of water to remove heat from computing equipment through evaporation. In a dry climate with a strong solar resource, evaporative cooling is highly efficient in terms of power consumption. It requires less electricity than mechanical refrigeration and provides effective cooling across a wide range of ambient temperatures. For operators optimizing for power usage effectiveness, the metric that measures how efficiently a data center uses its total power supply, evaporative cooling in an arid climate offered genuine advantages during the initial build-out phase.
The water consumption implications of that choice are now generating regulatory and political pressure that was not fully anticipated when the campuses were designed. The Phoenix metropolitan area draws its water from the Colorado River via the Central Arizona Project, a system of canals and pumping stations that delivers water from Lake Mead. Lake Mead has been at historically low levels for most of the past decade, a consequence of sustained drought across the Colorado River basin and decades of over-allocation that distributed more water rights than the river reliably produces. The federal government has imposed mandatory cutbacks on Colorado River deliveries to Arizona, Nevada, and California, and further reductions are likely as the basin’s water balance remains under stress.
Data centers that drew water from municipal supplies connected to this system have found themselves operating in a political environment where large-scale industrial water consumption is increasingly difficult to defend publicly. Several municipalities in the Phoenix metropolitan area have restricted or conditioned new data center water hookups, and scrutiny of existing operations has intensified. The industry has responded by developing more water-efficient cooling technologies, including closed-loop systems that recirculate rather than evaporate, and air-side economization that uses ambient air rather than water when outdoor temperatures permit. These technologies substantially reduce water consumption but require capital investment and, in some cases, impose power-consumption penalties that affect the economics that made the Arizona market attractive in the first place.
The water constraint is not unique to Arizona. Texas data centers in the Dallas-Fort Worth market draw on water systems that, while not subject to the same federal compact constraints as the Colorado River, face their own supply pressures from a combination of population growth and periodic drought. The broader pattern is that the Southwestern markets that built data center capacity on the assumption of abundant water access are now managing a constraint that was externalized during the build-out phase and is being internalized through regulatory pressure and shifting public tolerance.
The Renewable Energy Procurement Advantage, and Its Limits
Texas’s and Arizona’s genuine advantages for renewable energy procurement remain relevant and should not be understated. The solar resource in both states is exceptional by national standards, and the wind resource in Texas, particularly in the Panhandle and West Texas, is among the best in the country. Long-term power purchase agreements with solar and wind developers have enabled hyperscale operators to meet sustainability commitments at costs competitive with conventional power procurement in other markets. The renewable energy procurement story that attracted initial investment is not fiction.
The limitation is intermittency. Solar generation follows a diurnal cycle and is unavailable at night. Wind generation is variable and does not necessarily peak when demand peaks. Data centers require firm power, available at all hours and seasons, at the scale and reliability computing infrastructure demands. Renewable procurement addresses the carbon accounting question but does not resolve the firm power requirement. The gap between renewable generation and firm demand must be filled by battery storage, natural gas backup generation, grid imports where available, or on-site dispatchable generation.
In the ERCOT market, grid imports are constrained by the isolation described above. Battery storage at the scale needed for extended outage coverage remains expensive and has not yet been deployed at the capacity that would make it a reliable sole backup source. Natural gas backup generation reintroduces carbon emissions and fuel supply exposure. The renewable procurement advantage is real at the portfolio level but does not eliminate the firm power problem that ERCOT’s isolation and intermittency create at the facility level. Operators who built their power strategy on the combination of renewable procurement and grid reliability have found that ERCOT’s reliability assumptions can no longer be supported.
What the Southwestern Model Tells the Series
The reasoning supporting the development of the Texas and Arizona markets was not faulty. It was built on a set of economic and physical conditions that produced genuine advantages, and those advantages attracted investment that would not have gone to the Rust Belt during the same period, regardless of the stiff-point grid argument. The economics of renewable energy, land availability, favorable tax environments, and the workforce characteristics of major Southwestern metropolitan areas are real competitive factors that will sustain investment in these markets for the foreseeable future.
The Southwestern model, as a complement to the series’ broader argument, illustrates that siting frameworks built on single-variable optimization tend to produce attendant vulnerabilities. In Texas, the industry optimized for power costs and deregulated market flexibility without fully pricing the grid-isolation risk. Arizona optimized for cooling efficiency and access to renewable energy without fully pricing the water constraint. These were rational choices at the time they were made. They were choices that externalized costs and risks that have since been internalized through regulatory pressure, extreme weather events, and resource constraints that the original projections did not adequately weigh.
The stiff-point framework proposed in this series is a response to a similar pattern in the broader national market: the externalization of grid stress costs onto ratepayers who had no role in the siting decision. The Southwestern experience adds another dimension to that argument. Siting decisions that impose costs on shared infrastructure, whether electrical grids, water systems, or regional transmission networks, create political and regulatory backlash that eventually constrains the operators who made those decisions. The industry’s long-term interest is in siting frameworks that account for these costs in advance, matching facility requirements to the infrastructure that can absorb them without generating the kind of community and regulatory conflict that is now reshaping the Southwestern markets.
The Transition Already Underway
The industry is adjusting, and the adjustment is reflected in investment patterns. Major operators with large Southwestern footprints are diversifying their development pipelines toward markets with more favorable water profiles, more reliable grid interconnection, and lower exposure to the regulatory backlash that has emerged in Maricopa County and other congested Southwestern submarkets. The Rust Belt markets discussed in the previous article are among the beneficiaries of that diversification. So are markets in the Pacific Northwest and Upper Midwest that offer water abundance, grid interconnection to larger regional systems, and the combination of renewable energy access and firm hydro or other dispatchable generation that the purely solar-and-wind-dependent Southwestern model lacks.
The ERCOT market will not be abandoned; its scale, established fiber topography, specialized workforce, and favorable renewable energy economics ensure that Texas retains its status as a primary data center geography. However, the next phase of Texas market development will be decisively shaped by the legacy grid-vulnerability lessons from Winter Storm Uri and subsequent thermal stress events. This shifting landscape will increasingly mandate on-site generation, utility-scale battery storage, and a serious evaluation of alternative firm baseload infrastructure. Arizona will similarly maintain its investment velocity, but intensifying water-constrained permitting environments and regulatory scrutiny will inevitably restrict future development to facilities designed from the outset around closed-loop air cooling and minimized freshwater withdrawal.
The series will turn, in the next installment, from the question of where to build toward the question of how to govern what gets built: the RTO and ISO interconnection frameworks, demand cap mechanisms, and the administrative pathway by which the stiff-point qualification standard becomes enforceable policy rather than analytical argument. The Southwestern experience is directly relevant to that discussion, because the grid stress and resource conflicts that Texas and Arizona are managing are precisely the conditions that a more rigorous interconnection governance framework would have moderated during the build-out phase. The policy question and the geographic question are not separate. They are the same argument approached from different directions.
메타데이터
- post_id
- c16d4f9f99cd
- slug
- sun-wind-and-stranded-risk-the-structural-limits-of-the-southwestern-data-center-model-c16d4f9f99cd
- url
- https://medium.com/@jvanncunningham/sun-wind-and-stranded-risk-the-structural-limits-of-the-southwestern-data-center-model-c16d4f9f99cd
- canonical_url
- https://medium.com/@jvanncunningham/sun-wind-and-stranded-risk-the-structural-limits-of-the-southwestern-data-center-model-c16d4f9f99cd
- author_url
- https://medium.com/@jvanncunningham
- status
- ok
- fetched_at
- 2026-06-15 20:49:13