[ghoblin]

Energy is just one component of cost per unit of work, and in most modern economies it's a very small one. Industry has always traded energy efficiency for productivity. Excavators burn huge amounts of fuel compared to humans with shovels, data centers consume megawatts compared to the brain's ~20W, and forklifts are nowhere near biologically efficient. They still win because the economic cost per unit of work is lower.

In the case of AI the key difference is how the systems scale. Human labor scales linearly: if you want 10x more output, you hire 10x more workers, and the cost scales roughly the same. A human brain might run on ~20W, but each worker is still €20–€50+ per hour and can only do one task at a time. AI systems scale more like software infrastructure. Once the model and servers exist, the marginal cost of additional tasks is mostly compute and electricity. A data center might burn far more energy than a human brain per task, but it can handle thousands or millions of tasks in parallel and run 24/7. The cost per task can end up being cents even if the system is much less energy efficient "per brain".

Labor is expensive because you're not just paying for the task. You're paying for the worker's entire life infrastructure. Wages have to cover housing, food, healthcare, transportation, retirement, taxes, etc.., So the price of labor largely reflects the cost of sustaining a human being in society, not just the marginal cost of performing the work