Data Center Energy Demand Should be Contested

We can plan the grid for social good

By Winston Yau

Photo by Chad Davis on Wikimedia Commons

The AI data center boom is having a dramatic effect on energy bills, grid reliability, and pollution, and increasing attention is being paid to how to solve this problem. There are many proposals for how to best accommodate the massive electricity requirements of data centers by minimizing the negative impacts, including paying for their own generation and grid upgrade costs, as well as having grid operators require demand response, load flexibility and long-term contracts. Some have also started pointing to the need for public planning of the grid.

However, it is also necessary to question whether this data center load growth should exist in the first place, and to think through the political implications of merely responding to that exogenous demand over orienting the grid planning question around which demand should be accommodated in the first place.

Setting aside whether those proposals are sufficient to meet Big Tech’s power need estimates that continue to revise upwards, they tend to prioritize technical considerations over power dynamics that determine their ultimate effectiveness and uptake. There are real risks of creating overdependence on contributions from data centers, especially when “financial interests conflict with grid reliability.” If Big Tech views terms as too restrictive, they have the financial wherewithal to site their project with a different utility or region entirely that, say, offers looser curtailment or pollution control requirements. Along with the existing opaque and privately controlled utility landscape that limits public control over grid access terms, the financial, information and power asymmetry altogether make designing and implementing regulation over data centers’ grid participation incredibly challenging. Plus, there are limited resources on the grid as it stands: equipment delays up and down the supply chain, Trump’s attacks on renewable generation, slow interconnection queues, and sometimes dueling private utility and grid operator interests had already made existing tasks of electrification difficult. There is an opportunity cost to devoting resources to data centers over more socially useful grid demand.

Already, the push for data center abundance is providing license for a new wave of bipartisan “permitting reform” legislation that could remove environmental safeguards and spaces for substantive public input. And not questioning Big Tech’s urgency to power expansive electricity demand for these hyperscaler data centers facilitates the choice of continuing to run or even refurbishing coal and gas plants, which keeps prices and pollution high for working people.

Additionally, even though the gutting of the Inflation Reduction Act removed a powerful demand signal for renewable projects, attempting to shape public view of data centers as a substitute for that risks hitching clean energy onto a deeply unpopular political project, potentially undermining the political will and social license necessary for the grid expansion and improvement we desperately need for decarbonization and resilience. In Alabama, a Meta data center and renewable power proposal is giving an opening to renewable opponents to ban utility-scale solar in the state. And in California, people are pushing back against battery storage projects that could power future data centers. Any short-term calculus for clean energy advocates and developers to ride data centers’ coattails could backfire spectacularly for future electrification needs. It matters as a political choice—especially now—what we let tap into our collective electricity resource. Do we build out the grid by 5-8 times for home, industry and transportation electrification, or do we build out a grid that allows Big Tech to maximize profits?

There is also a risk of siloing the issue to an economic and physical question of how we accommodate data center load growth, when this accommodation itself undermines broader affordability and dignity by slowing hiring, intensifying work rather than reducing it, and disciplining worker power. You can’t pay a slightly lower electric bill if you lose your job or suffer from a wage cut from your employer introducing AI in your workplace (or at least citing it as a reason to). Put another way, the data center buildout and the AI takeover of our lives in consumer and labor realms are directly connected. No longer are these “asset-light” companies whose valuations soar without the scale of infrastructure expenditures like they are doing now. Data centers are strikingly unpopular, but AI itself is too! The public is not just asking for grid reliability and affordability, they are asking for agency over their own lives.

That is why the resistance to AI takeover of our lives has been running through data centers. The most exciting and viral clips on social media have not shown people asking for a renewable-powered, union-built data center that displaces jobs elsewhere anyway—they have demanded stopping them entirely. Their sentiments are correct: stopping data centers prevents new, non-electricity costs such as the imposition of massive physical footprint like clearcutting forests, localized air pollution, noise pollution (a very serious and underrated issue), and water usage, along with often betrayed promises of local benefits. The right question should be how many data centers should we have collectively, and what purposes they serve.

At present, grid access is what determines if companies can use this social good to expand slop and/or job displacement across different sectors, and we have an unprecedented opportunity to publicly link the grid, the AI boom, and the lack of democratic control over both to mainstream the conversation about the need for planning. Denying grid access to data centers slows the growth of the bubble and the foisting of AI tools across all aspects of our lives.

We need to put democratic planning of our grid—and economy—on the agenda. Bernie Sanders’s call for a national data center moratorium could set the stage to recalibrate our response to data centers and grid growth. Expanding public ownership of utilities that allow greater oversight and control over grid participation is another possible tool. Creative thinking on new or expanded FERC authority over interregional transmission that these increasingly large data center projects rely on could be yet another. When a denial of grid access to data centers is connected with an ambitious green industrial policy that ties in popular governance to deliver immediate and visible material benefits, grid buildout could be a durable political alternative. For example, we could orient towards electrification for energy-intensive industries like green steel production and EV battery manufacturing that employ significantly more workers than AI data centers and produce products that are socially useful.

The overwhelming scale of AI’s attack on our resources and dignity is being matched on the ground with higher political consciousness over the lack of ownership and control over what our resources are used for. Any proposals for centering planning in grid governance will have to evolve beyond being merely technical and reactive to corporate demand, and instead accommodate participatory forms of determining grid access. More democratic planning of our energy system and contesting what and who our grids are built for is the answer, not just letting Big Tech set the agenda.