Everybody’s talking about rising electricity prices. And just about everyone is blaming those increases on the massive growth in data centers powering “AI.” But the problem is deeper than mere increased demand. The neoliberal wholesale electricity market is in crisis because its underlying economic structures treat guaranteeing utility profits as the only mechanism for ensuring an adequate supply of electricity.
As the cliche goes, however, every crisis is also an opportunity. So far, the political response to rising prices has been to double down on privatization, attempting to protect electric ratepayers from tech companies’ demands on the power grid by insisting data center developers build their own power. It’s an admirable, but misguided strategy because, as I explain below, the entity that controls the grid most in demand by data centers is not accountable to ratepayers at all. The inchoate but widespread movement against data centers has created an opening to change how our grid is managed; rather than patch up the current system, state officials must insist on new, democratically accountable, institutions for the governance of the grid. The economics of wholesale electricity must reflect its role as an essential and collective good.
Tech Brings Its Own Power—and Cuts the Line
As tech companies and third-party developers line up billions of dollars in promised data center construction to run large language models (LLMs), utilities and grid operators are projecting astronomical increases in the need for energy in the next decade, disproportionately concentrated around Northern Virginia. Electricity users in the Mid-Atlantic and parts of the Midwest, where a non-profit “Regional Transmission Organization” called PJM runs the grid, will be particularly hard hit, paying an additional $7.3 billion more to keep the lights on from June 1 of this year to June 1, 2027 than they did during the previous year.
Elected officials in these states broadly agree that this trend is unsustainable. But the response even among “blue state” governors has been to avoid direct confrontation with data center hyperscalers by facilitating their access to electricity while attempting to mitigate the cost impact to customers. Governor JB Pritzker of Illinois has pushed for data center developers to “bring their own power,” that is, to secure their own new generation as a condition of connecting to the grid, and Governor Shapiro of Pennsylvania recently issued an executive order that, similarly, requires data centers to secure their own generation to move forward with the permitting process.
The goal of such requirements is a good one: protect the rest of us electric customers from the costs associated with huge new data center loads and ensure demand does not outpace supply. But in practice, this approach provides political cover for PJM—an organization run by the utilities themselves—to grant data center developers access to new electric generation projects ahead of other competing priorities, including the transition away from fossil fuels.
Environmental groups have long criticized PJM for delaying the process for adding new generation to the grid. For years, wind and solar farms patiently awaited required studies, while incumbent fossil fuel power plants benefitted from the artificial shortage PJM created by constraining supply. When tech walked in the door, PJM suddenly found the need for speed: in late 2024, PJM allowed fifty-one projects—overwhelmingly gas and nuclear—to jump the line and get connected. This summer, PJM allowed yet more line-jumping, this time allowing ten large (>250MW) projects to secure interconnection rights outside of the queue process entirely. Wind and solar projects, which are smaller but more numerous, were effectively ineligible for this special treatment. Even these accelerated pathways for large-scale gas and nuclear have not been fast enough for some hyperscalers, like Amazon, who is buying electric generation capacity paid for by past ratepayers and taking it off the grid entirely, instead using the energy to directly power data centers (in what is known as a co-location agreement where the power plant delivers electricity “behind the meter” to a single customer rather than feed into the grid from which all electricity users draw power).
In this context, a policy of “bring your own power” accelerates the tech privatization of the grid already underway. Rather than assert their democratic authority to evaluate and limit the growth of data centers at the state level, elected governors have attempted to avoid increased rates for their state residents while simply accepting that a such rapid increase in generation is necessary and hoping that by compelling data centers to procure this generation directly, they can protect ratepayers from paying for it. For instance, among the recent “principles” adopted by thirteen governors and the Trump Department of Energy is an additional auction meant to facilitate long-term “revenue certainty” for new generators outside of PJM’s usual process. As William Boyd has explained, these shortcuts overwhelmingly benefit data center developers—and gas plants. This fall, PJM will serve as a matchmaker between large loads (that is, data centers) and generators, enabling tech companies to obtain capacity outside of a centralized process. Tech companies will then “bring [their own] capacity” when seeking interconnection rights, avoiding the threat of interruption if the system becomes overloaded.
This rush has environmental consequences, as data center developers are satisfying the demand to “bring their own power” by going on a gas plant building spree. Data centers need constant electricity, and while a renewable grid can meet that demand, it requires time and scale to do so. The purpose of massive data center buildout, however, is not merely to meet surging customer demand, but to create it, by embedding LLMs in existing technological infrastructure; more immediately, the purpose of the rapid buildout is to win a capital expenditure race in which the prize is a higher stock price. Under these conditions, tech companies do not want to wait for the development of a renewables, so instead they build gas plants. Google, for instance, has abandoned its 2030 net zero pledge and is building a massive gas plant in Texas to power a new data center. Microsoft appears likely to do the same. All told, Global Energy Monitor has identified 250 GW of new gas projects planned in the United States. If built, these would double the United States’ current generation capacity.
It gets worse: through a quirk of statutory language, data center developers who build their own power plants may evade certain elements of federal Clean Air Act regulation altogether. The Acid Rain Program, which along with other measures reduced sulfur dioxide emissions by 95%, applies to any power plant that “produces electricity for sale.” The Trump administration has already seized on this language to argue that data center developers who build their own gas plants would not be subject to the requirements of the Acid Rain rule.
Corporate abandonment of environmental pledges and right-wing environmental deregulation are hardly remarkable, but that is precisely why the turn to private power will not solve our current energy crisis. Positing “bring-your-own power” as a solution to the problem of insufficient power to meet demand concedes the notion that electricity is not a shared resource but one that goes to highest bidder. “Bring your own power” thus risks facilitating a privatization of the grid akin to the tech sector’s attack on hotels and taxis through AirBnB and Uber: hyperscalers can leverage their access to astronomical amounts of capital to build a parallel set of resources and then force governments to accommodate it. This structural similarity is not a coincidence—hotels, taxis, and utilities all fall within the category of businesses that LPE scholars have termed “networks, platforms, and utilities,” businesses recognized for hundreds of years as “affected by the public interest.” And electric utility regulation drew extensively and explicitly from the regulation of common carriers.
This notion of electricity as a fundamentally and irreducible public good is now under threat. Data center developers have begun the very material process of establishing a parallel power infrastructure that will ultimately compete with what is now the world’s largest machine, the North American grid, which was built through the collective contributions of generations of electricity customers. It is not a point of surprising political convergence that Trump administration, unabashed cheerleaders for maximizing data center (and gas plant) build, has seized on bring-your-own-power as a “solution” to the crisis; it is a tell that an invitation to tech companies to bring their own power cedes our own.
Creating a Democratic Grid
If that is the crisis, what is the opportunity? Ironically, the watershed federal policy that “deregulated” the U.S. grid—facilitating the replacement of regulation with privately-run markets—has created the conditions for the kind of large-scale, interconnected grid that the energy transition demands. It also provides elected officials with tools to prevent, rather than facilitate, the privatization of the grid.
In 1996, the Federal Energy Regulatory Commission (FERC) adopted arguably the most consequential federal administrative rule for the electric power sector in its history: Order No. 888. At the time, the wires used to transmit electricity within and between states were controlled by vertically integrated utilities; these utilities protected their monopoly on generation and delivery by limiting access to transmission. With the express goal of introducing competition into the wholesale electrical market, FERC adopted Order No. 888, which mandated that long-distance transmission lines allow other utilities to sell electricity over those lines and thus create a “market” for electricity independent of the structures that transmitted it. Doing so, the Commission believed, would deliver “the benefits of competition” to customers, benefits that FERC estimated could be as much as $5.4 billion in cost savings per year.
But markets are neither self-creating nor self-enforcing, and so four years later, in 2000, FERC encouraged the voluntary formation of Regional Transmission Organizations, non-profit entities that would set the rules for both the physical operation and economic superstructure across vast segments of the country’s grid. These rules now include what steps must be undertaken before a new power plant can connect to the grid, how prices for wholesale electricity are set, how much wholesale electricity retail utilities are required to buy, and how much of a power plant’s nameplate capacity “counts” toward the capacity it sells into that same market. Regional Transmission Organizations effectively make energy policy for about two-thirds of U.S. residents. PJM is the largest of these entities.
PJM’s management system is complex. While independent of any individual utility companies, it overwhelmingly reflects the interests and attitudes of the utilities themselves. PJM’s Board is elected and major decisions are made by an elaborate “stakeholder” voting scheme in which most members are for-profit utilities and the vote share of “end-use customers” is limited to twenty percent. It is not surprising that the organization would be inclined to facilitate, rather than manage, the potential profit-generating data center race. Meanwhile, PJM’s wholesale capacity market, which was created pursuant to Order No. 888 to secure the benefits of competition for customers, has become highly concentrated. In 2024, PJM’s Independent Market Monitor, a team of professional energy economists, found that “structural market power” by massive electric generator holding companies is “endemic” to PJM.
While the data center crisis has intensified and accelerated longstanding issues within PJM, the organization has always suffered from a fundamental paradox. PJM, like any Regional Transmission Organization, is tasked with ensuring the grid’s reliability, including guaranteeing enough generation to meet demand, and this is understood as its highest priority. At the same time, PJM is ideologically committed to “markets” and profit incentives as the means of achieving this guarantee. The result is that PJM itself decides how much electricity is needed, decides how much consumers are willing to pay for it, and facilitates a “market” in which various generators offer electricity—nominally competing on price—and PJM purchases enough to meet estimated demand. But in creating this demand curve, PJM assumes that the only way to secure sufficient generation is to offer prices that guarantee utilities a return on their investment; in the words of the Trump-state “Statement of Principles on PJM,” “price certainty” is presumed to be the only guarantor of sufficient supply. Consumers can never benefit from low prices, because, according to PJM, utilities will not build unless they can be assured of profit. Prices are set by market ideology (that is, the belief that profits are necessary to secure generation), rather than an actual market.
There is no legal reason why wholesale electricity has to be managed by utility-dominated Regional Transmission Organizations (California’s grid is managed by an organization whose Board is appointed by the governor and confirmed by that state’s legislature) or adequate generation secured through a market in which profits are always guaranteed. Beneath the economic superstructure, the interstate network of electric generation and transmission is fundamentally shared and therefore amenable to being managed as a collective good. Transmission lines are a shared resource that cross boundaries between utility footprints, the necessary physical means by which we can collectively use the wind and solar resources to provide cheaper energy at a fraction of the human cost of fossil fuels. The mandate to open up those lines for general use was not just a means of introducing wholesale energy markets selling electricity across them; it has now also become a necessary and powerful step toward grid decarbonization and creates genuine possibilities for democratic governance.
A grid built on renewable generation requires interconnection and flexibility, finding needed energy across long distances, shifting time-of-use in accordance with battery capacity and daylight hours, and distributing generation to increase resilience. While the transition away from fossil fuel generation increases the need for planning and infrastructure to improve these qualities, the grid itself is a machine built on our interdependence. This is not a metaphor: to remain operative, electricity must cycle through the system at a frequency of exactly 60 Hz; if generation and use diverge, this frequency falls, and without intervention, the whole system collapses. The need for democratic governance of the grid arises not just from free-standing political principles, but from the physical nature of the thing itself.
Utilities and tech companies alike will continue to exploit this shared resource at the expense of the global climate and electric customers unless and until elected leaders insist on politically accountable governance that refuses to allow the privatization of electric generation or transmission. Rather than attempt to manage the data center demand crisis in a way that moves the grid—like so much else in the United States—toward a two-tier system of corporate access and collective dependency, elected officials should seize the opportunity to challenge our current system of grid governance. The solution is not a regression toward monopolistic ownership of transmission lines, but a system directly accountable to ratepayers and empowered to incorporate environmental and social values in its decisions about what gets built and who it serves. Whether by federal legislation under a new Congress, or state compacts among aligned governors, elected officials need to wrest management over our grid back from utilities and build a new model of democratic wholesale electricity oversight that will insist that data centers do not get to buy their way of interdependency or burn methane at the expense of our power bills and climate.
