We Are Taxing the Wrong Thing
The AI revolution demands a total rewrite of the global tax code.
The social contract of the 20th century was built on a simple trade: labor for revenue. In the vast majority of the world, governments fund their existence—justice, infrastructure, and defense—by taking a "cut" of human effort. If you sweat, code, or consult, the state gets paid.
But we are currently witnessing the rapid decoupling of productivity from human labor. As Artificial Intelligence moves from simple chatbots to autonomous agents, the foundation of our tax system is beginning to crumble.
The Fiscal Death Spiral
We are entering a paradoxical era. On one hand, automation is driving productivity to record highs. On the other, the traditional tax base—human income—is being hollowed out. From assembly lines and software engineering to medicine and legal research, machines are assuming the roles of high-income earners.
This creates a "fiscal pincer movement":
1. Revenue dwindles: As humans are replaced by silicon, income tax receipts shrink.
2. Expenses skyrocket: Governments face mounting pressure to fund social safety nets (and potentially Universal Basic Income) for a displaced workforce.
If we continue to tax "inputs" (human hours) rather than "outputs" (machine productivity), we aren't just facing a deficit; we’re facing a systemic collapse of the state’s ability to function.
The "Lumen" Logic: Taxing Utility, Not Effort
Think of the transition from incandescent to LED lightbulbs. We used to buy bulbs based on wattage—a measure of how much energy the bulb consumed. When LEDs arrived, they produced the same amount of light using 90% less energy. To help consumers understand the value, manufacturers labeled them as "100-watt equivalent."
If the government taxed bulbs based on energy consumption (the input), their revenue would have fallen by 90%. If they instead taxed the lumens (the output/brightness), the revenue would remain stable regardless of how efficient the technology became.
Taxes are not meant to punish innovation; they are the "subscription fee" we pay for a functional society. Just as we moved from taxing wattage to valuing light, we must move from taxing the effort of the worker to taxing the value created by the machine. A car that once required 100 humans and now requires 30 robots should still yield the same "labor-equivalent" tax revenue for the state.
The Case for a Post-Income VAT
The most efficient mechanism for this shift is a sophisticated Value Added Tax (VAT) system. While the U.S. remains the only OECD country without a federal VAT, the AI era makes it an inevitability.
A VAT taxes products at every stage of production where value is added. By shifting the tax burden to the point of sale and the point of production:
1. We stop punishing the working class: Eliminating or drastically reducing income tax removes the "success penalty" on human workers.
2. We capture AI "surplus": When a company replaces a $150k-a-year engineer with a $20-a-month AI subscription, the massive increase in profit margin is captured via the VAT rather than disappearing into a corporate tax haven.
3. We regulate the "Velocity of Automation": Currently, employers have a massive tax incentive to automate; they don't have to pay payroll tax, health insurance, or social security for a GPT-5 agent. A productivity-based tax levels the playing field, ensuring automation happens because it's better, not just because it’s a tax dodge.
Getting Ahead of the Curve
Our governing bodies—be they Congress, Parliaments, or specialized agencies—need to stop viewing AI purely through the lens of "safety" and start viewing it through the lens of solvency.
This shift serves three vital purposes:
1. Securing the State: It ensures revenue remains decoupled from the shrinking pool of human work hours.
2. Protecting the Laborer: It removes the artificial "tax discount" businesses get when they fire a human.
3. Economic Self-Regulation: A VAT-driven system creates a more direct feedback loop between supply and demand. As the cost of production drops through AI, the tax revenue scales with the volume of goods and services actually consumed by the public.
If we don't rewrite the code now, we will be left with 21st-century technology trying to fund a 20th-century government on a 19th-century tax base. The math simply doesn't add up.
I resonate with this. The math simply isn't adding up otherwize.