A CNBC segment this morning featured the mayor of Chandler, Arizona, discussing the city council’s unanimous rejection of a proposed AI data center campus. Residents voiced strong opposition to the data center, citing potential impacts on the grid, massive power and water use, limited local concerns, and fears that AI facilities could displace jobs and drive up already-high energy bills. 

The mayor said residents are already struggling with high energy bills and are worried prices will go even higher. He also noted that Chandler has already hosted more than its fair share of traditional data centers. But those data centers differ in that they draw large, but relatively steady and predictable, loads that utilities can accommodate. AI-data center demand is fundamentally different. Training clusters produce extreme, intermittent power spikes. Meanwhile, inference workloads have high, variable baselines, pushing overall consumption and grid stress beyond historical levels. 

Despite this, many proposed AI facilities are still designed as passive energy consumers. In doing so, they prioritize uptime and grid connectivity but remain exposed to transmission constraints, curtailment risks, tariff volatility, and supply disruptions. Onsite generation fundamentally changes that dynamic. 

With integrated onsite power generation and storage, AI campuses transform into dispatchable, energy facilities that can support the health of the overall grid. Onsite generation enables primary self-supply, dynamically allocating power, scheduling intensive training during optimal windows, deferring non-critical tasks, and optimizing in real time for cost, reliability, or emissions. Done right, that means dramatically lower incremental stress on the grid, eliminating the burden on community power availability, and one less reason for communities to say no.