The city that built the cloud has decided it is finished subsidizing the AI industrial complex. Amazon and Microsoft extract public electricity to run a private AI racket draining municipal power. It is true, and the developers had a point in the narrow sense property developers usually do, when they argued before the Seattle city council that the city’s grid has the physical capacity to meet the demand of five new large facilities. The trouble is that the capacity they are describing is a capacity between Amazon and Microsoft and the public grid, and not a contestability that offers a fair price to the ratepayers who are being asked to bear the cost of the buildout. There is no market at which the utility can refuse to sell power and the developers can turn to a competitor; there is only a monopoly buyer leveraging its size to extract a subsidy in kind. This is not a marginal addition to the industrial load; it is a capture of the base grid, and the city council committees unanimously passed a year-long moratorium last week to stop the meter from running while the terms of the extraction are redefined.

It is a common error to view datacenters as ethereal manifestations of “AI progress.” They are, in plain engineering terms, heavy industrial facilities that demand vast, baseload power, thousands of gallons of water, and significant land conversion. When a city pauses to consider whether these structures are a “good use of urban land,” it is simply applying the same land-use principles that govern every other industrial operator. The pattern here is chokepoint capitalism applied to physical infrastructure. Amazon and Microsoft are headquartered in the same metro area where the datacenters are proposed, and their combined AI capital expenditure is projected at $390 billion for 2026. They are extracting the value of the public grid to support an asset class that is currently generating returns measured in stock price rather than product delivery. The infrastructure is being pitched as the foundation for a new AI utility, but in practice it operates as a mechanism to extract value from existing labor structures: executives deploy capital to automate functions and slash headcount, turning engineers into depreciating assets until the workforce is rendered simultaneously indispensable and disposable, while the actual value flows upward to shareholders and downward into municipal strain.

The discipline of cryptographic-protocol verification—a field that trains engineers to distinguish systems that happen to work from systems whose security has been formally established—has a useful concept here. A system whose specification is not written down cannot be verified for safety or fairness. The utility impact statements for these datacenters are not specifications; they are marketing documents. There is no price attached to the draw on municipal power, no contract securing the public’s share of the value generated by the compute, and no mechanism for ratepayers to refuse until the meter is already running. The playbook is older than the cloud. When a hyperscaler positions itself between a municipal grid and a city’s ratepayers, extracting value in the form of cheap power and deferred infrastructure costs, it is running a racket that an older generation of Canadian steelworkers would have recognized immediately. The mechanism of the acquisition was different—the rolling mills in Selkirk were restructured in 1995, the datacenters in Seattle are proposed in 2026—but the political economy is the same: capture the chokepoint, extract the rent, and leave the community to repair the damage. An industrial plant historically tethered workers to a neighborhood and built a local tax base; an AI datacenter does the opposite, deploying automated capital that replaces the community’s labor, severs the connection between work and place, and leaves behind a carbon footprint and a bill for local utilities to cover.

The most telling aspect of the Seattle protest is not just the climate concern—though the reliance on fossil-fuel baseloads is a genuine hazard—but the source of the resistance. It is being led by a coalition that includes Amazon’s own employees, organized through internal channels and long-standing groups like Amazon Employees for Climate Justice. Tech workers, having endured a year of layoffs, argue that this new era is making them more productive but demonstrably more disposable. The sentiment captured at a recent city council meeting was precise: AI is now locally synonymous with making people more productive and more disposable at the same time. The workers understand what Doctorow names enshittification—the four-stage decay where a platform abuses its users to capture value for its shareholders—played out in the physical layer. First, the utility was good to the tech firms, offering cheap power and expedited permitting. Then, the value was clawed back in the form of unregulated buildouts and degraded grid reliability. The next stage is where the ratepayers and the workers are now: being asked to pay for the privilege of being extracted. The moratorium allows the city to impose a rate structure that is fixed and transparent, preventing the kind of twiddling that lets developers adjust their energy costs from moment to moment, hiding the extraction behind a billing statement that never shows the adjustment.

Regulatory capture thrives on the claim that cities are locked into a race to the bottom. Seattle is proving that even in the epicenter of the sector, that shackling is a choice. Mayor Katie Wilson’s admission that she was not even aware of these datacenter plans until they hit the Seattle Times speaks to the arrogance of a sector that assumed the grid was its private resource to drain. Seattle is joining a growing coalition of jurisdictions that have realized the federal government offers no leadership on the political economy of AI infrastructure. The trend began in earnest last year, when the Maine legislature approved the nation’s first statewide delay on new facilities, and it intensified with a Texas county’s decision to pause construction to study impacts and a permanent ban in Monterey Park, a US first, announced just yesterday. Before that, a Michigan utility blocked water access for twelve months, and a Washington state utility already imposed separate rates for large load customers. The framework is the same across these disparate cases: local governments are asserting the right to define datacenters as a use of urban land that must earn its keep for the community, rather than an immutable force of nature that the city must accommodate. They are reclaiming the role of the public utility as a steward of shared resources, not a passive landlord for hyperscaler capital. The pause gives the municipality time to consult with tribal lawyers on treaty water rights, to draft regulations tailored to the facility’s actual footprint, and to demand that the developers contribute to affordable housing and transit to offset the strain on local services.

By forcing these companies to justify their energy use, their land occupancy, and their public-benefit contributions, Seattle is asserting the most fundamental municipal right: the authority to decide who pays for the power, and who gets to breathe the air. The era of the “zoning-by-default” land-use regime is running headlong into the reality of the municipal grid, and for once, the grid is holding firm. There is a public consultation running alongside the moratorium on the design of the “large load” rate structure and the public benefit requirements. The deadline for the council vote is next Tuesday, June 9, 2026. The building permits are frozen. The work is to be done.