Andrew Yang: More Money is Going into Machines, Not Workers
Former presidential candidate Andrew Yang posted that less money is flowing to workers and more is entering machines. His recent communication states that the labor share of GDP has been declining for decades, approaching a historical low of about 50%, and claims that artificial intelligence will send more value to the cloud. He cited over $30 trillion in AI spending that will siphon off the share previously paid to workers, with nearly half of recent graduates underemployed, and research institutions hiring fewer assistants due to models.
The 2020 campaign focused on automation and universal basic income. The 2026 communication advocates taxing AI or computing power while reducing taxes on people, citing Dario Amodei from Anthropic discussing token taxes, Sam Altman from OpenAI discussing a national wealth fund, Vinod Khosla advising against taxing the bottom 80% of workers, and John Arnold discussing taxing computing to enhance hiring competitiveness. He noted that model companies are still losing money and thus do not contribute income tax, resulting in a combination of "fewer hires and no new tax base." Office communications previously estimated that about 70 million white-collar jobs in the U.S. would significantly shrink in the coming years, with publicly traded companies opting for layoffs to boost stock prices.
Money entering machines refers to data centers, chips, and model inference bills, not just robotic arms in workshops. Wage bills are being rewritten as cloud bills, and the tax base follows the contracts. He described universal basic income as a minor arithmetic solution to return the technological dividend to households.
The labor share is a macro accounting entry. The machine share is a micro invoice of capital expenditures and cloud services. When both increase in the same year, the jobs in between are squeezed out.
Market mechanisms involve factor substitution. The buy side consists of companies that aim to complete the same reports and codes with fewer people; the sell side involves models and cloud vendors selling inference by tokens. Funds are shifting from wage accounts to capital expenditures and operational cloud costs. Beneficiaries are chip, power, and model equity; those under pressure are junior and mid-level white-collar workers who can be replaced by prompts. Event-driven analysis is a judgment, not a new labor statistics revision.
Invoices are being re-titled. The title changes from "wages" to "computing power." Household accounts will not automatically revert.
Source: Public Information
ABAB AI Insight
Yang is re-invoicing the warnings from 2020. The decline in labor share predates generative models, with models providing an observable accelerator: assistant positions disappearing can be pointed out. The $30 trillion is a supply-side gamble that must be recouped from the demand side of wages or profits. Loss-making model companies cannot temporarily contribute income tax, leading to a fiscal timing mismatch. Token taxes and computing power taxes are nailing the tax base to machines, while UBI nails cash to individuals; he writes both legs, yet no legislation has been enacted.
The capital path is cloud contracts replacing employment contracts. Companies are converting research assistants into subscriptions, and stock prices reward fewer heads. Chip and power vendors take the construction phase money, while model vendors rely on operational phase money. Workers lose hourly rates, while shareholders gain operational leverage. Recent graduates are the first group to be underquoted because their jobs resemble tasks that prompts can complete.
The analogy is automation's impact on manufacturing employment share and software's impact on travel agencies. The white-collar phase is controlling cognitive tasks that can be packaged into models. Those who can write tasks as APIs will draw from the wage pool, while those who can only sell time face a shrinking pool. Universal basic income is a post-compensation, not a preservation of original positions.
Structural change is technological substitution. The mechanism is when marginal cognitive costs are lower than marginal labor costs, the labor demand curve shifts inward, and capital expenditures shift outward. Pricing power shifts from job grades to tokens and electricity prices. Machines do not need vacations but do require substations. Ultimately, the "entry into machines" on the books is entering a few companies that can issue invoices.
ABAB News · Cognitive Laws
- In years when cloud bills lengthen, wage bills often shorten first.
- Machines can create valuations first and only later create tax bases.
- What is replaced by prompts is not the meaning of work, but its unit price.