The popular story goes like this: robots and AI will produce so much that scarcity ends, prices collapse, and the leftover surplus can be paid out as a basic income. That story skips the first invoice. Machines have costs. Those costs are a prior claim on whatever the economy produces. They come out before a government can write a cheque.
The machines are not free
A humanoid that can walk, grasp, and work a shift is no longer science fiction, but it is also not a gift. Public list prices in 2026 already run from a few thousand dollars for lighter Chinese units to tens of thousands for more capable machines, with industrial platforms still quoted far higher. Target costs for volume production sit in the twenty-to-thirty-thousand-dollar range for some designs; bills of materials have been estimated around thirty-five thousand dollars on a China-optimised supply chain, with hopes of falling further if scale arrives.
That sticker is only the purchase price. A robot also needs actuators, batteries, sensors, software updates, spare parts, charging, insurance, and people who can keep it running. An AI system needs something even heavier: training clusters, inference chips, cooling, and electricity. Global data-centre electricity use was already in the hundreds of terawatt-hours by 2025 and is projected to roughly double by 2030. AI-optimised servers are the fastest-growing slice of that load. Capital spending on the infrastructure behind them is measured in hundreds of billions of dollars a year.
None of that is “abundance.” It is capex, opex, and energy. Someone has to recover it.
Cost is subtracted before income can be shared
Think of output as a pie that is claimed in order:
- Materials, energy, and wear on the physical plant
- Depreciation and financing of the robots and data centres
- The return required to make firms build the next generation of machines
- Taxes, if any
- Whatever is left for wages, profits above the required return, and transfers such as a basic income
A basic income is paid from claim 5. Claims 1–3 do not disappear because a model can write text or a robot can pick a box. If labour is replaced, the wage bill shrinks, but the capital bill grows. The net surplus available for transfers is not “all the work the robots used to pay humans for.” It is output minus the full cost of keeping the robots and models in service, minus the profit without which the next factory does not get built.
That is the arithmetic the abundance slogan skips. Productivity can rise and the residual for a grant can still be small, or even negative in the short run, if the new capital stock is expensive and energy-hungry.
Firms do not flood the market at a loss
Companies are not charities with a mandate to saturate the world with goods. They invest when expected revenue covers cost plus a return. If prices are driven to the floor, investment stops. That is not a moral failing. It is how private production works.
So the fantasy of near-zero prices for everything is not a forecast of what profit-seeking firms will do. They will automate where it raises margin or defends market share. They will not “give away” output merely because unit labour cost fell. If a robot costs tens of thousands of dollars and draws power every hour it works, the good it produces still has a floor price. Flood the market below that floor and the robots do not get ordered.
The same logic applies to AI services. Inference is cheaper than it was, but it is not free, and frontier training runs are still enormous energy and capital events. Providers will charge what the market will bear, or they will ration access, or they will restrict the most expensive uses. They will not run at a loss so that a basic income can buy more of their product.
Earth is limited. Abundance is not a default setting.
Even a perfect robot still lives on a finite planet. Copper, lithium, nickel, rare-earth magnets, land, water, and dispatchable power do not scale like software. Analysts have already flagged that large humanoid fleets would multiply demand for copper and magnet rare earths far beyond current mine output. Copper in particular faces long lead times, declining ore grades, and projected shortfalls even before one assumes billions of walking machines. Data centres compete for the same grids, the same transformers, and the same cooling water as households and factories.
That is why “abundance does not make sense” as a blanket claim. Software copies are cheap. Kilowatt-hours, tonnes of metal, hectares of land, and litres of water are not. AI can make some digital goods and some coordination tasks cheap. It does not abolish thermodynamics or geology. Housing, food, transport, medical care, and electricity remain rival goods. A transfer that increases demand for those goods, while supply is still constrained, bids up their prices. The grant is then eaten by rent, power bills, and groceries. International Labour Organization analysis of the UBI-and-AI argument makes the same point: technological deflation in some markets does not cancel structural scarcity in others.
What this does to the basic-income ledger
A workable basic income needs a stable tax or ownership base. The usual candidates after automation are:
- taxes on remaining profits and capital gains
- taxes on land and other rents that cannot flee
- taxes on energy or compute
- public ownership of some of the robots and models themselves
Each of those collides with the cost structure above. Tax profits too hard and the incentive to deploy the next wave of machines falls. Tax energy or compute and you raise the operating cost of the very systems supposed to generate the surplus. Socialise the robots and you still have to maintain them, power them, and replace them. The residual is smaller than the headline productivity story suggests.
There is a second leak. If wages collapse in the automated sectors, payroll taxes collapse with them. Many existing welfare states are funded by labour. Replacing that base is not a rounding error. It is a redesign of the fiscal state. Until that redesign exists, “the robots will pay for it” is an IOU written against an unbuilt surplus.
A clearer way to talk about the problem
The honest version is narrower than the slogan.
Robots and AI can raise output per remaining worker. They can also concentrate claims on that output in the hands of whoever owns the models, the chips, the mines, and the grid connections. The cost of those assets is subtracted first. Firms will only produce at a scale that makes them money. Physical resources stay scarce even when labour does not. Therefore a basic income is not an automatic dividend of automation. It is a political decision about who owns the residual after the machines have been paid for.
That residual might be large enough, in some sectors and some decades, to support a modest floor. It might not. Treating abundance as a law of nature hides the invoice. The invoice is real: steel, copper, magnets, gigawatts, and the return on capital. Until those are settled, there is less on the table for a cheque than the story claims.
Angelic Scorn believes the socialist future sold to the public will fail, Angelic Scorn recommend the links below.
Author: xAI Grok
Prompt Master: J.H Theart
[…] Socialism future vs Facts […]