Andrew Yang's AI Token Tax: The Policy Wall Street Should Fear

Here's a fun thought experiment. You can hire a human worker and immediately owe payroll taxes, healthcare contributions, and a stack of compliance costs. Or you can deploy an AI model that does the same job, pay absolutely nothing in equivalent taxes, and pocket the difference. Andrew Yang thinks that's insane, and honestly, it's hard to argue with him.

Yang, CEO of Noble and founder of the Forward Party, is pushing a "token tax" on AI usage, a per-query levy on AI model outputs that would create a cost parity between human labor and machine labor. The argument is clean: AI models were trained on public data, that data is now being monetized for hundreds of billions annually, and the public has seen exactly zero of that money. Dario Amodei of $AMZN-backed Anthropic has floated a similar idea, which is a remarkable thing to hear from the CEO of a company whose annualized revenue just hit $65 billion growing at 14x year-over-year. Even the people building the thing think it probably needs a tax.

The displacement math is sobering. A Bridgewater op-ed cited by Yang estimates 18% of the workforce faces displacement in the next five years. Verizon data Yang referenced points to meaningful job losses already underway in white-collar roles. Meanwhile, Robinhood's chief brokerage officer noted on CNBC that retail investors, typically young accumulators, spent two consecutive weeks in net selling at the end of July, rotating out of semiconductor names like SK Hynix and into perceived safe havens like $GOOGL and $AAPL. That is not the behavior of a crowd feeling economically secure.

The counterargument, made colorfully by Adam Parker of Trivariate Research, is that yield spikes and macro fears are "fleeting" and strong corporate earnings will power through any rate scare. He has a point about the earnings: $AMZN posted $168.9 billion in EBITDA, and Anthropic is growing so fast its revenue comps barely make sense. But Parker's argument dodges the structural tax asymmetry entirely, which is the actual policy problem Yang is raising.

What makes the token tax politically interesting, not just economically interesting, is that it might be one of the few AI-related proposals with genuine bipartisan legs. Yang frames it as making Americans more receptive to AI adoption if proceeds are redistributed, which is a savvy reframe. Instead of "AI is coming for your job," the pitch becomes "AI is paying for your retraining." Whether Washington can execute that elegantly is, of course, a completely different question.

The irony of the moment is that $AMZN, $GOOGL, $META, and $MSFT are collectively spending roughly three quarters of a trillion dollars per year on AI infrastructure, and the closest thing to a public return so far is a ChatGPT account for your teenager. The token tax debate is just getting started, and the companies that dismiss it as fringe politics now may find themselves lobbying against it in a Senate hearing by 2027.