Introduction
A 1.2 GW Load for a Single Data Center
The 27th of July, 2026, Meta submitted to the Arizona Corporation Commission (ACC) a connection request for a new data center in Mesa, with a nominal power of 1.2 gigawatts. The project, which involves the installation of over 30,000 high-density servers, was preliminarily approved despite the average load of the city of Phoenix being approximately 600 megawatts. The local electrical infrastructure — already under pressure due to the expansion of the residual solar energy market — must now handle an instantaneous increase equal to 3.2% of the maximum regional capacity.
The critical point is not the theoretically available power, but the interconnection time: obtaining authorization from the ACC takes an average of 18 months. The speed at which permits are granted therefore becomes a determining factor in attracting strategic investments. In practice, every six-month delay results in an estimated loss of $42 million in potential operating revenue for the data center.
The Internet as a Barrier to Technological Progress
The entire Arizona region consumes approximately 10 gigawatts solely to power the computational loads related to artificial intelligence. This level of consumption, which includes not only data centers but also research centers and high-frequency financial institutions, has exceeded the value predicted in the 2025 energy plan by Arizona Public Service (APS). The data indicates a structural divergence between strategic planning and the exponential growth of synthetic systems.
Local networks — designed for peak seasonal loads, not for continuous 24/7 use — are showing early signs of saturation. The most critical physical node is the Phoenix South substation, which operates at 96% of its nominal capacity. Each new applicant must therefore undergo a network impact assessment: if the expected load exceeds 0.5% of the local total, a review process is triggered by the ACC.
The Paradox of Growth Without Capacity
According to an internal report from Meta, the decision to locate the new center in Arizona was motivated not only by lower energy costs compared to the West Coast, but also by the availability of industrial land and the absence of regional moratoriums. However, the collateral effect is a compression of resources for the rest of the system: local utilities are already exploring options such as restructuring the 138 kV grid or purchasing small modular thermal power plants.
“The expansion of data centers in Arizona poses a challenge for electrical grids, but also an opportunity to renew the governance of thermodynamic flow. The problem is not consumption, but the strategic allocation of resources.” — Vittorio Quaglione, Project Syndicate
The data indicates that technology location decisions are now determining the redistribution of regional logistical power. Those who control access to thermodynamic flow also control the pace of innovation.
The Cost of Delay in Systemic Restructuring
Today, the Arizona power grid operates with an average operating margin of 0.8%. To maintain this threshold in the presence of an average annual IA load growth of 41%, at least $3 billion must be invested by 2027 to expand the transmission network. Without intervention, within three years, the entropy dissipated by the system will increase by 58% compared to 2024.
The real trade-off is not between development and sustainability, but between regional competitiveness and logistical control. Those who decide to build a data center today pay an infrastructure cost that will be distributed across the entire local energy population. The limit is not physical: it is political. And the operating margin for new initiatives decreases every day.
Monitor the network saturation index
If you are considering an investment in AI infrastructure, the key metric to monitor is the saturation index of the main substation in the Mesa district. Any value exceeding 95% indicates that the network cannot accommodate additional loads without structural interventions. The current value is 94.2%. The critical threshold will be reached within the next semester.