The Fight Over Data Centers Is Dividing the Labor Movement - Jacobin

TEXT START: Blue-collar and white-collar workers increasingly find themselves on opposite sides of the groundswell against AI data centers.

THE DISSECTION

This is not really an article about water, zoning, or inter-union etiquette. It is an exposure map of the automation pipeline. Building-trades unions are being paid now to construct the physical substrate that enables future cognitive replacement, while nurses, professors, teachers, communications workers, and other unions face the downstream threat.

The labor movement is splitting by position in the sequence: some workers sell labor to the buildout; others are threatened by what the buildout produces. The article correctly identifies data centers as the visible battlefield and the proxy target for a population that lacks control over AI firms and federal policy. It also documents the immediate jobs and local environmental disputes without pretending every criticism is accurate.

But this remains coalition management, not a full structural diagnosis. The blue-collar/white-collar divide is a lag artifact. Construction, energy, logistics, and maintenance may create survival niches around AI capital; they do not restore the mass wage-consumption circuit.

THE CORE FALLACY

The article identifies AI as the root problem, then treats moratoria and enforceable standards as if they could solve it. Under Discontinuity Thesis mechanics, a data-center pause is a lag defense. It can delay capacity, raise costs, and buy time. It does not alter P1, ownership of AI capital, or the competitive pressure to automate. Existing facilities continue operating, and capital can relocate or return when political resistance weakens.

The article also mistakes data centers for neutral infrastructure. A hospital is built to provide care. An AI data center is built to scale machine substitution. Comparing their construction jobs ignores the function of the asset. Current union wages are not evidence of a durable new middle class; they are temporary payments for assembling the machinery that makes much of the old middle class economically unnecessary.

HIDDEN ASSUMPTIONS

• Regulation can outrun AI capital and prevent firms from routing around local restrictions.
• Local moratoria can coordinate at national scale despite jurisdictions competing for jobs and investment.
• Guardrails can preserve mass employment while leaving AI’s cost and performance advantage intact.
• “Good projects” can be separated from bad ones without changing who owns compute and captures the output.
• Union bargaining over AI in workplaces can preserve productive participation rather than merely delay displacement.
• Temporary construction employment is an adequate answer to permanent downstream labor substitution.
• Public opposition can become durable class power instead of remaining a coalition of local externality fights.
• Environmental misinformation weakens the case against data centers; even perfect water accounting would not resolve the labor and ownership problem.

SOCIAL FUNCTION

Classification: partial truth, transition management, and ideological anesthetic.

The article tells the truth about the labor split, political capture, local impacts, and absence of alternatives for the trades. It performs transition management by proposing pauses, standards, and negotiated boundaries. It becomes ideological anesthesia when it implies that regulation can reconcile an automation regime with the old wage economy. A regulated AI system can still terminate productive participation if the underlying ownership and competitive mechanics remain unchanged.

THE VERDICT

The labor movement is not primarily divided by principle. It is being sorted by exposure. Building trades are extracting near-term income from the same capital formation that threatens the rest of labor; opponents are attacking the only visible bottleneck because ownership and federal power are out of reach.

The article’s strongest insight is that data centers are a proxy battlefield. Its failure is stopping there. A moratorium may buy time and create bargaining leverage, but without control of AI capital, compute, energy, logistics, and deployment, P1 survives, P2 defeats stable human-only domains, and P3 remains the endpoint. The construction boom is not a replacement for the old order. It is paid assembly work for the machine that will make that order obsolete.