Elon Musk has made one of the boldest economic predictions of the artificial-intelligence age: by 2036, money may no longer matter. The reasoning is clean. Money exists because things are scarce. If AI and robots produce more food, housing, transport and healthcare than human beings can consume, scarcity fades, and with it, the point of money.
It is a compelling technological argument built on a single unexamined assumption: that money is fundamentally a rationing device for consumer goods.
It was never only that.
Money does not merely distribute goods. It distributes power. It records ownership, settles debts, finances states, gives its holders claims on the future output of others and in the dollar’s case, can determine who retains access to the most important channels of the global financial system.
Look at the numbers. In the first quarter of 2026, the world's central banks held $13.1 trillion in foreign-exchange reserves, of which the dollar accounted for 57.13%, against 20.03% for the euro and 1.99% for the renminbi. Global foreign-exchange turnover reached $9.6 trillion a day in April 2025, with the dollar on one side of 89% of all trades. Roughly 55% of international banking claims and 60% of liabilities are dollar-denominated. The Federal Reserve's composite index of international currency usage scores the dollar at 64.9 against 23.9 for the euro and 3.1 for the renminbi, while the United States produces just 26.1% of world GDP.
That gap between 64.9 and 26.1 is the whole story. The dollar is not powerful because America is large. It is powerful because the world is organized around it.
And the sharpest edge of that power has nothing to do with rationing anything. Dollar clearing runs through institutions under American jurisdiction, which is what makes sanctions, secondary sanctions and correspondent-banking exclusion function. That is money operating as coercion, not as a queue for televisions. Cheaper robots do not automatically weaken a single strand of it.
There is a deeper anchor still. What ultimately floors demand for a state's currency is that the state accepts nothing else in settlement of taxes. Taxation places a durable floor under demand for sovereign currency. States rarely surrender that monetary authority willingly, and no major power would abandon it merely because technology had made society richer.
So the institutions built on money will not try to stop AI abundance. They will try to own it.
They are already doing so at a scale once reserved for national infrastructure. The International Energy Agency reports that the expenditure on electricity, advanced chips, suitable land, cooling systems and transmission capacity by five technology companies exceeded $400 billion in 2025 and is set to rise a further 75% in 2026, more than global investment in oil and gas production. Data-centre electricity consumption is projected to double from 485 terawatt-hours in 2025 to around 950 by 2030.
Which exposes the first contradiction in the post-scarcity story. Producing abundance consumes things that cannot themselves be made abundant: electricity, advanced chips, land, cooling, transmission capacity, capital. Someone will own those. And when one scarcity dissolves, competition simply migrates to whatever remains, beachfront property, land in New Yark or central London, university places, political authority, environmental capacity, human attention, trust. Machines can copy a design a billion times. They cannot manufacture a second Manhattan.
Then comes the harder point. Abundance of production does not guarantee abundance of access.
We know this already from the digital economy. Software copies at zero marginal cost, and people still pay license fees. Films duplicate instantly, and copyright still gates them. Information is technically abundant and economically scarce, because ownership law surrounds it with tollbooths. Intelligence may follow exactly the same path: cheap to generate, expensive to reach.
Nor does this require conspiracy. A company does not maximize output; it maximizes sustainable profit. If unlimited production collapses prices, owners restrict supply, tier access, differentiate products and lobby for rules that keep competitors out. Scarcity can be produced by nature. It can also be manufactured by ownership.
Even the supposed challenger reinforces the incumbent. Stablecoin market capitalization stood at approximately $320 billion at the end of May 2026, with about 98% of its value denominated in dollars. The BIS estimates that stablecoins recorded approximately $28 trillion in transactions during 2025, but much of that reflected trading and movements within the crypto system. BCG estimates that only $350 billion to $550 billion represented real-economy payments; McKinsey’s separate estimate was approximately $390 billion.
Musk may well be right that AI may sharply reduce the cost of education, analysis, medical administration, diagnostics, transport coordination and many manufactured goods. That would be an extraordinary achievement. But it is a claim about production, and he offers it as an answer to a question about claims on production.
Machines decide what a society can make. Law and politics decide who receives it. The technology in both futures is identical; the ownership arrangements are not.
AI may reduce the scarcity of goods. It will not abolish the human appetite for power, and money remains one of the most durable ways that appetite is recorded, transferred and exercised.
