Elon Musk gave money an expiration date during a widely circulated interview with The Economist. Asked how his companies would earn returns in an age of artificial intelligence, he answered, “Money won't matter in 2036.” Musk believes robots and AI will produce more goods and services than any person could consume, making work optional and money largely irrelevant.
His forecast takes technological abundance seriously, which is one reason it deserves a serious response. AI and robotics will make many forms of intelligence, manufacturing, and service delivery radically cheaper. Most people cannot anticipate the radical changes technology is about to bring to society (changes that will be, on balance, for the better).
Yet money will remain necessary because abundance changes the location of scarcity without eliminating finite resources, competing preferences, ownership, or uncertainty. That is what money is for – it is a tool for coordinating scarce resources. Even in a world of abundance, scarcity will exist. An interesting question is whether the economy of 2036 may require better money technology than we have today – likely to comprise bitcoin-based payment infrastructure readily usable by AI agents – in order to facilitate the explosion of goods and services that Musk envisions.
Elon Musk’s Vision Of Abundance
Musk’s argument begins with a vision of extraordinary productive capacity. Early in the interview, he predicts an “age of amazing abundance” in which “anyone can have anything they can think of.” When the interviewer later asks why investors should buy shares if profits eventually cease to matter, Musk makes the 2036 prediction and explains that people want money for food, housing, transportation, and entertainment. If robots provide more of those goods and services than people can use, he asks, what purpose would money serve?
Later in the interview, he predicts that work will become optional. Musk compares future employment to gardening: people can buy pristine vegetables at a store, yet some still grow imperfect tomatoes because they enjoy the activity. In his account, productive labor follows the same path once machines satisfy material needs.
This is an optimistic forecast, and the direction is plausible. Mechanization already reduced the labor required to produce food, clothing, transportation, and information; AI simply extends that long term trend into cognitive work. A legal brief, software prototype, medical image analysis, or custom design that once required hours of scarce professional labor can increasingly be produced in minutes.
Lower production costs, however, do not establish that every desired good becomes freely available. Musk’s conclusion requires abundance to spread across the resources used to build and operate the machines, the ownership claims governing their output, and the human preferences that determine who gets what. Those constraints become more visible as the machines grow more capable.
Scarcity Moves Upstream
AI is often experienced as weightless software, but its physical demands are substantial. The International Energy Agency reports that a typical AI-focused data center consumes as much electricity as 100,000 households, while the largest facilities under construction may consume 20 times as much. Global data-center electricity demand is projected to more than double to roughly 945 terawatt-hours by 2030, slightly exceeding Japan's current consumption.
Those figures describe only one input. Advanced chips require specialized fabrication plants, critical minerals, water, land, transmission capacity, and long construction timelines. Humanoid robots add motors, batteries, sensors, maintenance, and physical space. Productivity can rise dramatically while these upstream resources remain finite and geographically constrained.
Scarcity also persists wherever two people want the same rival good. AI may build more houses, but it cannot reproduce a particular acre in Manhattan or a waterfront lot in California. Robots may provide excellent medical care, but access to a specific surgeon, researcher, artist, or leader remains limited by that person’s time. Perfect digital copies can even increase the value of authentic originals because abundance changes what people regard as rare.
Prices help resolve these competing claims by conveying how intensely people value goods relative to their supply. If an allocation system abandons money, it still needs another method for deciding who receives scarce land, electricity, computation cycles, or the output of a factory. Put another way, even if most of what people need money to pay for today becomes super-abundant by 2036, the construction of arcologies, spaceports, off-world settlements, and countless other wonders of the 21st century will need resources allocated among them, and money is how that gets done. You can try to use some other mechanism to coordinate the economy, like political rationing, waiting lists, lotteries, status, and personal influence, but we should all be able to agree that money is preferable to those.
Money Is Insurance Against An Uncertain Future
Money also serves a purpose beyond purchasing familiar goods. It preserves optionality when people do not yet know what they will want or need. Ludwig von Mises explained the connection in Human Action: “Only because there is change, and because the nature and extent of change are uncertain, must the individual hold cash.”
A person saves because tomorrow may bring illness, opportunity, disaster, invention, or a change of mind. Perfectly abundant groceries would reduce one category of expense, but they would not tell anyone how much future medical treatment, travel, education, energy, or computing power will be needed. Greater technological change can expand the range of possible futures, which gives people more reasons to preserve purchasing power across time.
AI does not eliminate that uncertainty because learning changes the decisions people and machines will make. A discovery made in 2035 may create desires and industries that no one can specify today. New capabilities also produce new bottlenecks. The internet made information abundant, but this led to a heretofore unappreciated scarcity of human attention, trusted curation, domain authority, and network position.
Money allows millions of people with incompatible plans to prepare without agreeing on one forecast. It carries purchasing power from the known present into an unknowable future. Any system that performs this function will behave like money, whatever name Musk’s world gives it.
Who Owns The Robots?
The Economist’s interviewer reaches the decisive issue moments after Musk’s prediction. If Tesla creates the humanoid robots, will it sell them or control them while selling the goods and services they produce? The question exposes the ownership layer beneath the abundance thesis.
If companies own the machines, their output belongs to shareholders until contracts transfer it. Wages, dividends, subscriptions, rents, and prices determine how the output is distributed. If households own robots, people still need a way to acquire the machines, purchase energy, replace components, and exchange one robot’s output for that of another. If governments own the productive system, allocation moves into politics and administration.
Robotic abundance can therefore reduce the marginal cost of production while increasing the importance of capital ownership. The people and institutions controlling data centers, energy infrastructure, fabs, robot fleets, land, and intellectual property decide how capacity is deployed. Money provides a common language for valuing those assets and directing scarce investment toward competing uses.
Following the ownership claims through Musk’s scenario leaves ample room to criticize today’s distribution of wealth and the failures of fiat finance. Production answers how goods are made. Money and property rights answer who may use them, when, and on what terms.
Machines Will Need Money Too
Musk’s forecast also arrives as AI systems become economic participants. Autonomous agents already purchase services, negotiate API access, and allocate computing resources. In a study I examined earlier this year, 36 frontier AI Models made monetary choices across 9,072 experiments. The models chose bitcoin in 48.3% of all scenarios, preferred it in 79.1% of store-of-value scenarios, and selected stablecoins most often for payments.
The experiments measure machine reasoning under designed conditions, and their relevance lies in how readily scarcity recreates familiar monetary functions. An agent with a compute budget must compare prices, reserve resources, pay counterparties, and decide whether an expected result justifies the cost. Multiply those decisions across billions of agents operating at machine speed and the demand for programmable economic coordination increases.
Bitcoin is relevant because it gives machines and humans access to a scarce digital bearer asset with predictable issuance and final settlement. AI can interact with bitcoin directly, without asking a bank to recognize a machine as a customer. As I argued in an earlier examination of bitcoin and the forces shaping the new economy, artificial intelligence may be deflationary for many goods while making credible monetary scarcity more valuable.
Musk may be right about the scale of the coming abundance and the declining role of compulsory work. Those changes would improve billions of lives. They still leave society with finite energy, finite locations, finite attention, ownership claims, and an unknowable future. Money remains the most adaptable tool humans have developed for coordinating those realities. In 2036, it will still carry that burden.