There’s a fatal flaw in Elon Musk’s abundance utopia

In the tycoon’s telling, humanity is on the cusp of “sustainable abundance”. Work is optional. Poverty is abolished. It is a beguiling vision, but also naive.

Elon Musk is selling the world a seductive story.

In the tycoon’s telling, humanity is on the cusp of “sustainable abundance” – billions of humanoid robots, powered by artificial intelligence, will manufacture almost anything, unleashing a supply shock so vast that the output of goods and services massively outpaces growth in the money supply.

Everyone can have a penthouse. It is a beguiling vision. It is also, on close inspection, economically naive and conveniently self-serving for a man whose corporate empire is valued on precisely these promises.

Start with the most basic problem. You cannot conjure billions of robots out of thin air. Each humanoid is a dense package of advanced semiconductors, memory, sensors, actuators, batteries, rare earths and exotic materials, all wrapped around staggering quantities of AI compute for training and inference. Before Musk’s supply shock arrives, the world must first absorb a demand shock of historic proportions across every one of these inputs.

We do not need to speculate about what that looks like. It is happening now. The AI data centre build-out has sent memory and chip prices soaring, blown out lead times on turbines, transformers and switchgear, and pushed electricity prices higher across grids from Virginia to Victoria.

That is the inflationary consequence of building the computational substrate for AI alone. Now scale that up to a global manufacturing program for billions of physical robots, each one competing for the same silicon, the same lithium, the same copper, the same megawatts.

The road to Musk’s deflationary paradise runs through years, plausibly decades, of intense inflation in the very commodities and capital goods his revolution requires. The abundance thesis has the sequencing exactly backwards: the demand shock comes first, and it is enormous.

The second flaw is the fantasy of competitive supply. Musk’s deflation story implicitly assumes robots become cheap, ubiquitous commodities. But the economics of the robot race point in precisely the opposite direction. Like search engines, smartphone operating systems, social networks, rockets and satellite internet, humanoid robotics is an industry defined by colossal fixed costs, steep learning curves, data network effects and winner-take-most dynamics. The firm that trains the best brain on the largest fleet of deployed machines compounds its advantage with every hour of operation.

That is a recipe for monopoly or, at best, a cosy oligopoly. And monopolists do not price at marginal cost – they extract supernormal rents. There is no reason to expect the dominant robot maker to hand its productivity miracle to consumers via collapsing prices rather than to shareholders via fat margins. Humanoids are likely to be inherently expensive to manufacture for the foreseeable future and rationed by price, available first to corporations and the wealthy, not showered upon the masses. Scarcity does not disappear – it migrates from labour to the ownership of robot capital.

Nor is it obvious the robots will be deployed making cheap commodities at all. A rational monopolist directs its scarce robot fleet towards the highest-margin activities available, namely novel, value-added goods and services we have not yet imagined, where pricing power is greatest, rather than driving the price of toasters to zero. Consumer preferences will shift toward these new offerings, as they always do, creating fresh scarcities and fresh rents.

Two centuries of productivity growth have never found the ceiling on human wants. Demand simply rotates into new forms of status, novelty and positional consumption. Musk’s claim that humans will “run out of things to want” is contradicted by the entire history of capitalism, including the history of his own product launches.

Then there is energy. Musk waves this away by invoking the sun: harness a sliver of solar irradiance and power is effectively free. Perhaps, eventually. But that remains unproven, and in the here and now the marginal megawatt for AI is being supplied by gas peakers and deferred coal retirements at rising prices.

Add the near-certainty that governments will heavily regulate humanoids walking among us, through safety standards, liability regimes, licensing and insurance, and the cost stack rises further. Regulation of this kind is inflationary for both the robots and everything they produce.

If you want a preview of how Musk’s companies behave once they achieve dominance, look no further than Starlink.

Having built a near-monopoly in satellite broadband, SpaceX this year raised prices across almost every consumer plan, hitting both new and existing customers, while quietly removing the option for residential customers to buy their dishes outright in favour of rentals. This is textbook monopoly conduct: subsidise adoption, capture the market, then harvest the rents.

It is also a rather awkward data point for a man promising that his machines will usher in the end of scarcity. The Starlink precedent suggests the robot economy will feature abundant pricing power, not abundant penthouses.

This brings us to the grandest claim of all, that superintelligence will exceed the collective cognitive capacity of humanity within a few years, rendering our economic institutions obsolete. Here the rhetoric has raced far ahead of the evidence. For all the genuinely impressive capabilities of large language models, and I use them daily, we have precisely zero examples of autonomous AI – no AI that is self-aware. No AI with consciousness, agency, or a self-preservation instinct. No AI operating as an independent entity pursuing its own goals in competition with others. Passing a Turing test is not autonomy – it is mimicry at scale.

Today’s AI is critically and completely dependent on human resources, human infrastructure and human direction. It runs on chips humans fabricate, in data centres humans build, drawing power from grids humans operate, executing objectives humans specify. Strip away the anthropomorphic marketing and AI, at this juncture, is nothing more than a robotic slave executing the whims of humanity – an extraordinarily capable tool, but a tool nonetheless. Tools raise productivity. They do not abolish economics.

None of this is to deny that AI and robotics will be transformative, or that reproducible goods will get cheaper over time. They will. But the honest forecast is not Musk’s frictionless glide into deflationary utopia. It is a decade or more of input-cost inflation as the robot economy is built; the concentration of extraordinary pricing power in a handful of AI, compute and humanoid quasi-monopolies, quite possibly Musk’s own; widening inequality between those who own the machines and those displaced by them; and wrenching societal disruption as industries are eviscerated.

One should always ask cui bono. A narrative in which Tesla’s Optimus robots and Musk’s AI ventures are the panacea for the human condition is, at minimum, a magnificent piece of pre-IPO marketing.

Abundance may well come, eventually, in some form. But the path there will be inflationary, monopolistic and politically explosive, and the man promising you a free penthouse is the one most likely to end up owning the building.