A 330-megawatt compute facility is difficult to describe as merely a building with servers inside it. Its computing capacity depends on an electrical system capable of delivering enormous power, backup generation that can sustain operations, batteries that can stabilize or support that supply, and cooling and water infrastructure that keeps the equipment operational. That physical reality has now collided with a regulatory question in Imperial County, California.
A tentative ruling from Imperial County Superior Court Judge L. Brooks Anderholt found that county officials could not treat the proposed Imperial Valley Computer Manufacturing project as a collection of largely mechanical approvals while relying on an older environmental review. The ruling calls for the county to set aside key approvals and conduct a full environmental review under the California Environmental Quality Act, or CEQA. The important issue is not simply whether this particular project moves forward. The more consequential question concerns how the industry defines a data center when the infrastructure surrounding the computing floor makes the computing floor possible.
The 330 MW Number Is Only the Beginning
The proposed project illustrates why that distinction matters. The development covers about 75 acres and includes an approximately 950,000-square-foot data center with a 330 MW load. But the project also includes a 330 MW substation, an 862 megawatt-hour battery energy storage system, cooling and water treatment facilities, four 500,000-gallon water tanks, a six-acre retention basin and a 330 MW emergency generation facility equipped with 132 natural gas generators. Those components do not operate as independent businesses that happen to sit near one another. They form an infrastructure chain.
The substation exists to serve the computing operation. The generators provide backup capability for the same operation. The batteries support its electrical architecture. Cooling systems protect the computing equipment from heat. Water infrastructure supports those systems. The physical development therefore creates a useful test of whether environmental and permitting frameworks should examine each component separately or evaluate the infrastructure as a single operational system. The court’s reasoning points toward the latter approach. It found that grading, electrical facilities, water systems, lot changes, road changes and related infrastructure serve one cohesive construction and operation of the data center. That is a more significant proposition for the compute industry than another dispute over a permit.
Hyperscale Projects Have Become Infrastructure Packages
The data center industry has spent years making development faster by dividing complex projects into specialized workstreams. Utilities handle grid connections. Developers negotiate land and entitlements. Energy specialists arrange generation. Battery providers supply storage. Mechanical contractors build cooling systems. Water specialists handle treatment and supply. The resulting project can look like a set of parallel infrastructure transactions. Operationally, however, the separation becomes harder to defend as facilities grow.
California’s Energy Commission already recognizes that data centers can involve substantial supporting energy infrastructure. The agency notes that most data centers rely on the electric grid as their primary power source but commonly require on-site generation for backup. It also regulates certain thermal generating facilities of 50 MW or more and uses the Small Power Plant Exemption process for qualifying facilities up to 100 MW. The Imperial County dispute pushes that concept into a broader question: Can permitting frameworks continue to treat supporting infrastructure as peripheral when the compute facility cannot function without it? At 330 MW, the answer becomes increasingly difficult to separate from the physical facts.
The Bigger Risk May Be Definition, Not Delay
A full environmental review adds another regulatory step to a project and, in this case, the developer has raised concerns that the process could affect the project’s timeline. That consequence matters in a market where developers compete for power capacity, construction windows and access to increasingly constrained infrastructure. But the more interesting consequence could involve project definition. If regulators increasingly evaluate the compute facility and its supporting infrastructure together, developers may need to establish the environmental footprint of a project earlier and more comprehensively. A power system that once appeared as a separate infrastructure decision could become part of the same environmental analysis. The same could happen with water systems, storage and on-site generation.
That does not necessarily mean every hyperscale development will face the same outcome. The California case turns on its specific facts, approvals and regulatory record. The court also found substantial evidence supporting potential environmental effects involving energy and electrical infrastructure, water supply and wastewater, air quality, greenhouse-gas emissions, hazardous materials, fire safety, traffic, public services and cumulative effects. Still, the reasoning gives developers and local authorities another project-specific example to consider when structuring and reviewing large compute developments. The question shifts from What permits does the building require? to What physical system is actually being approved?
Compute Capacity Could Become an Environmental Unit
That shift could prove particularly important as AI infrastructure projects continue to require substantial amounts of electrical capacity. A megawatt figure describes more than the electrical demand of servers. It can determine the scale of substations, transmission connections, backup generation, battery storage and cooling systems required to deliver reliable compute capacity. That makes power capacity increasingly inseparable from physical development.
The California Energy Commission’s own planning work reflects the growing importance of data center demand to the state’s energy outlook. The commission works with utilities to assess the likelihood that proposed projects will actually reach completion and incorporates that information into its broader electricity demand forecasting. The Imperial County case adds another layer: If a project consumes hundreds of megawatts and requires an extensive private energy architecture to support that consumption, the environmental footprint cannot easily stop at the walls of the computing building. This could eventually influence how developers structure applications, how counties evaluate infrastructure and how investors assess development risk.
The Industry May Need to Stop Thinking in Components
The most disruptive part of the ruling is therefore not that California may slow one data center. It is that the decision challenges a convenient mental model. The traditional development model can treat the data center as the central asset and everything else as supporting infrastructure. But AI compute increasingly reverse that relationship. The computing operation can drive the electrical architecture, which in turn can shape generation, storage and cooling requirements, while the resulting infrastructure can affect a project’s water and land requirements. The components remain legally and commercially distinct in many cases. They are not necessarily operationally distinct.
If environmental review follows the operational reality rather than the contractual structure of a project, developers may have less room to present a hyperscale facility as a collection of individually manageable approvals. The project could instead emerge as one infrastructure system whose impacts become visible only when the pieces appear together. That would not necessarily stop the expansion of AI infrastructure. It could, however, force the industry to define what it is actually building before it asks regulators to approve it. And that may prove to be the more consequential question raised by California’s 330 MW fight.


