US employers spend more than $1.5B a year to fight labor unions, report finds

B. TEXT ANALYSIS: The $1.5B Union Suppression Machine

1. THE DISSECTION

This is a symptom catalog without a diagnosis. The EPI has assembled rigorous empirical data—the $1.5B annual expenditure, the $442M specifically on consultants, Amazon's $26.6M in 2025 alone, the 41.5% violation rate, the 465-day average to first contract—then narrates it as a story of power asymmetry that can be corrected through stronger labor law, better enforcement, or more aggressive organizing. The article performs moral indictment of employer behavior. It does not interrogate the structural logic that makes that behavior inevitable.

2. THE CORE FALLACY

The article assumes the problem is political rather than structural. The entire EPI framing—that employers exploit loopholes, that weak reporting requirements enable this, that workers are fighting an uphill battle against well-resourced adversaries—implies the solution is to close the loopholes. Fix the law. Shutter Littler Mendelson's Workplace Policy Institute. Pass PRO Act. Mandatory card check neutrality.

This is the fallacy. Under the Discontinuity Thesis, the $1.5B is not a bug; it is rational capital behavior in a system where AI is progressively eliminating the mass employment substrate that made union power structurally viable in the first place. Employers are not suppressing unions out of malice. They are suppressing unions because the labor market they depend on is being automated away, and union power was always a function of labor scarcity leverage—which AI erodes mechanically. The $1.5B is a rearguard investment against a retreating employment model, not an expression of arbitrary corporate evil.

The article documents the symptom, moralizes the symptom, and offers a prescription that addresses nothing structural.

3. HIDDEN ASSUMPTIONS

• Assumption 1: Higher union density is a viable salvation path. The article treats the drop from 20.3% to 10% union density as the core problem. But under the DT, union density decline tracks the erosion of mass employment itself—not primarily because consultants are good at their jobs, but because AI is automating the jobs unions were designed to protect. More unions at 10% density does not solve the structural problem; it rearranges deck chairs on a deck that's being dismantled by automation.

• Assumption 2: The employment relationship is stable and permanent. The EPI framework implicitly assumes the traditional wage labor model is the baseline against which deviation is harm. It is not a stable baseline. It is the entity being automated out of existence.

• Assumption 3: Collective bargaining can capture value from AI-driven productivity. The assumption is that workers organizing can extract higher wages, better conditions, and shared prosperity from a growing economy. Under the DT, AI-driven productivity gains are structurally disconnected from human labor compensation—the gains accrue to capital owners, not labor, regardless of union density. This is the thesis's core claim.

• Assumption 4: The Starbucks 465-day delay problem is a legal failure. The article frames long delays as employer weaponization of process. Under the DT, this is symptomatic of institutional lag—the legal system's procedural architecture is slower than the speed of structural economic change. The delay isn't the problem; the problem is that even winning 465 days from now solves nothing structurally.

4. SOCIAL FUNCTION

This article serves as progressive institutional output: rigorous documentation of a real power asymmetry, delivered with moral clarity and policy-adjacent recommendations, targeted at an audience that wants to believe the system can be reformed through better rules. It is:

• Partially true: The $1.5B figure is real. The power asymmetry is real. The 465-day delay is real.
• Ideologically anesthetic: It channels legitimate worker grievance into a framework (labor law reform) that does not address the structural cause.
• Transition management: It manages worker organizing energy by directing it toward goals (union density restoration) that are politically achievable in the short term but structurally inadequate in the medium term.
• Prestige signaling: EPI and LaborLab produce this research to maintain institutional relevance in a policy ecosystem that still rewards labor-law-framed analysis.

The Amazon spokesperson's response ("teammates should understand what it actually means to have an external party take their voice") is Orwellian and correctly identified as such in the article. But notice: Amazon is spending $26.6M to suppress unions in a labor market where its own workforce is increasingly automated. This is not contradiction—it is foresight. Amazon is spending anti-union money today to prevent organizing friction tomorrow, when the workforce is smaller, more precarious, and more easily replaced by automation.

5. THE VERDICT

The EPI report is technically excellent and structurally useless. It documents the war with precision while the underlying terrain shifts underneath all combatants. The Discontinuity Thesis predicts that union organizing—whether successful or suppressed—becomes increasingly structurally irrelevant as AI severs the mass employment -> wage -> consumption circuit. The workers organizing at Amazon and Starbucks are fighting for rights in an employment model that is being mechanically dissolved by automation.

The $1.5B is a rounding error against the AI capital transition. Even if workers won every election, secured mandatory neutrality agreements, and achieved 30% union density—a political near-impossibility—the structural displacement of human labor by AI would continue, and the wage share of GDP would continue its secular decline, because the mechanism linking human labor to productivity gains is being replaced at the capital level, not the workplace level.

The workers organizing are not wrong to organize. The DT explicitly endorses the Hyena's Gambit and Transition Intermediation. But they are organizing inside the wrong frame. The fight cannot be won inside the employment model. The fight must be for ownership stake in the AI capital itself—profit sharing, equity participation, data rights, sovereign wealth funds for displaced workers, direct distribution of AI productivity gains. These are not in the EPI's framework. They are not in the article. And until they are, articles like this are elaborate, well-sourced eulogies for a labor movement that has not yet grasped what it is actually losing.