In the late nineteenth century, economics was trying to become a science. If physicists could discover laws of motion, perhaps economists could discover laws of markets. The older language of political economy, with its moral arguments and historical observations, was giving way to mathematics. The economy, once understood through institutions, history, and human character, began to appear as a system of forces and quantified relationships.

That ambition required simplification. Real markets were too tangled to model. Merchants possessed different information. Buyers had habits and loyalties. Firms varied in size, skill, reputation, access to capital, and political influence. Prices were shaped not only by supply and demand, but by bargaining power, geography, regulation, and trust. To make markets mathematically legible, economists had to imagine conditions under which these complications receded.

Out of that effort emerged what economists came to call “perfect competition.” It was not a portrait of ordinary commerce, but a purified boundary case: many buyers and sellers, interchangeable goods, free entry and exit, widely available information, and no single participant powerful enough to move prices. For more than a century, it served as a benchmark against which real markets could be understood.

Artificial intelligence makes this abstraction newly relevant. Not because markets will suddenly become perfect, but because a new kind of actor is entering them—one that does not share the same idiosyncratic human characteristics around which modern capitalism was built.

Consider a hypothetical. Imagine a business owner so confident in an AI system that they delegate all strategic and operational decisions to it, stepping back to serve primarily as a figurehead. Assume the model is mildly superhuman, with no significant weaknesses. Over time, we should expect this AI-led corporation to steadily outmaneuver its competitors. Now imagine the same phenomenon playing out across the economy.

On some level, this would be like having an alien run a company. The impact could resemble the dynamics of an invasive species entering an unprepared ecosystem. Traditional firms, even highly capable ones, are built around human particulars: incentives, turf wars, limited attention, and slow institutional learning. A sufficiently capable AI system would, at minimum, manage its resources differently than we are used to.

Consider what happened in chess when AlphaZero arrived. Given only the basic rules, AlphaZero taught itself from scratch by playing millions of games against itself. Within hours, it surpassed centuries of human knowledge. Crucially, it did not merely play chess better. It played very differently. AlphaZero often sacrificed material to achieve long-term positional benefits rather than immediate material advantage. It would give up pawns to open lines, weaken the opponent’s king safety, or disrupt pawn structure. These sacrifices were not aimed at an immediate tactical payoff. They were intended to secure some type of longer-term, harder-to-see advantage.

It's not a huge leap to see an analogy in the game of business. AI systems may exploit market positions humans perceive as irrational or untenable, only to achieve surprising success.

This kind of strategic discontinuity has precedent even in human-led companies. Elon Musk’s takeover of the automotive industry offers a compelling proof of concept. Musk recognized that battery technology had reached a point where desirable electric vehicles could be feasible, largely due to efficiency gains driven by consumer laptops and cell phones. This was a difficult judgment call to make. Getting it right allowed Tesla to enter the market at precisely the moment when the fundamental constraint on electric vehicles was becoming solvable, but before established manufacturers had made serious progress.

The automotive incumbents dismissed Tesla as a non-threat. As legacy brands continued their outsourcing strategy, Musk broke from convention and vertically integrated where possible. Traditional dealerships watched as Tesla sold vehicles online without negotiation. Tesla also developed new manufacturing techniques and found novel ways to reduce costs. Today, Tesla produces electric vehicles at four global Gigafactories, has an advanced humanoid robot program, a growing battery business, and a robotaxi platform powered by its homegrown self-driving software.

Elon Musk brought $200 million to one of the most consolidated industries on the planet and disrupted it thoroughly within roughly a decade. Tesla’s market cap has at times approached 1.5 trillion dollars, dwarfing the next largest auto company, Toyota, valued at around $250 billion. All of this was due to one uniquely capable agent. What might be accomplished by enterprises equipped not with hundreds of millions, but billions in capital, and an intelligence engine far surpassing Elon Musk?

And on the other side, could legacy brands counter in the near future by enabling an AI system to run their operations? Could they regain ground lost to Musk? Interestingly, this AI system would not only match Musk’s strategic brilliance; it would also avoid his well-documented shortcomings. Musk’s impulsiveness, controversial public statements, and occasional erratic management decisions have frequently undermined Tesla’s goals. This type of executive discipline would likely be quite desirable for institutional investors.

Venture capital and private equity may eventually reorient around this vision. Instead of searching for human entrepreneurial talent, perhaps the capital allocator of the future is simply an AI that deploys subagents toward various market segments. Would Warren Buffett see this as the perfection of capitalism, or as a dystopian mistake?

The role of the human executive could evolve significantly. Instead of being the primary strategic leader, the CEO may become more of a PR frontman, closer to a Hollywood actor than a traditional corporate executive. This would make everyone more comfortable by giving the organization a human interface.

Yet even in a world teeming with advanced artificial intelligence, many competitive advantages have little to do with raw intelligence. Some markets follow a “winner-take-most” dynamic. A bestselling author gains early visibility, leading to more reviews, media coverage, and retailer promotion, which drives even more sales. Other talented writers may still find an audience, but the top performer captures the largest share, not only because of merit, but because of name recognition, social evidence, and market momentum.

Similarly, real-world infrastructure—things like ports, warehouses, and trucking fleets—is costly to build and slow to reconfigure. A competitor that has invested in port facilities over years, or has exclusive rights to a critical shipping route, carries an advantage that intelligence cannot simply wave away.

The same principle applies to brand loyalty. If consumers are set in their preferences or trust a particular company, no amount of algorithmic brilliance will instantly steal that entire customer base. Over time, cost or quality differentials can chip away at loyalty, but brand inertia remains potent.

Another variable is regulatory capture. A firm with strong political connections or a legacy position might shape the regulatory environment to its advantage, effectively locking the door to new entrants.

What does all this mean for the value of companies? Conventional wisdom suggests that stock prices reflect earnings and growth potential. But if this hypercompetitive form of capitalism systematically erodes profits across industries, does it ultimately deflate the value of stocks? If competition forces margins toward zero, then by traditional valuation models, stocks should decline over time. However, we should approach this reasoning cautiously. The relationship between fundamentals and asset prices is not always straightforward. Consider the parallel universe of cryptocurrencies, where prices move wildly without earnings, growth, or intrinsic value.

As consequential as these dynamics are, an increasingly machine-based economy raises even more fundamental questions.

Much of economics is more parochial than it appears. It presents itself as a general theory of production, exchange, incentives, and scarcity. But many of its assumptions are specific to human psychology. An economy is, at bottom, elaborate hunting and gathering. However abstract its instruments become, it remains a system through which organisms secure food, shelter, energy, status, safety, pleasure, and future possibility. Markets, firms, wages, prices, contracts, and capital are modern layers built on top of an older biological problem: how to get enough from the world to live.

Capitalism organizes this hunt through rivalry. It sets producers against one another and lets consumers benefit from the struggle. Firms search for better techniques, cheaper inputs, more attractive products, and more efficient arrangements because failure is punished. In a world of human producers, this makes sense. Competition disciplines creatures that conserve energy, lose focus, defend status, and become complacent.

Capitalism is not only a system for allocating resources. It is also a system for extracting effort from calorie-conserving primates. Natural selection did not produce organisms designed to maximize quarterly output, sustain attention for hours at a time, or expend energy with no clear consequence for refusing to do so. It produced animals adapted to survive under conditions of scarcity, danger, social competition, and uncertain reward. Energy had to be conserved. Attention had to be selective. Risk had to be managed. Status mattered because social standing affected survival and reproduction. Markets are built around those traits.

Artificial intelligence alters these premises. An AI system does not need a salary to remain interested. It does not need status to stay focused. It does not procrastinate because the task feels unrewarding. It does not become resentful during quality control, distracted during legal review, or mentally depleted after eight hours of strategic planning. Given electricity, compute, data, and an objective, it can keep applying intelligence without the motivational drama that surrounds human labor.

At first, this may make capitalism more extreme. Firms armed with superior intelligence would compress margins, discover new strategies, and erase old moats. They would compete with a speed and precision no human organization could match. Markets could become less forgiving, less sentimental, and less protected by incompetence.

But over time, a second possibility may emerge. If machines are doing more of the producing, planning, monitoring, and improving, then some activities that once required rivalry to remain sharp might be organized through coordination instead. Competition is valuable because of what it produces, not because rivalry is sacred. If a coordinated system can deliver the ice cream cone at the amusement park cheaply, safely, reliably, and deliciously, the consumer may not care whether ten firms fought to produce it.

As systems become more capable, rivalry may become not only less necessary, but more dangerous. Commercial competition gives each actor an incentive to move faster, exploit every advantage more aggressively, conceal capabilities, pressure suppliers, manipulate consumers, and optimize around legal or institutional constraints. Among human firms, those tendencies are limited by confusion, exhaustion, bureaucracy, reputation, and the slowness of organizations. Among superintelligent systems, they could unfold with far greater speed and precision. In that world, coordination may not be a sentimental alternative to competition. It may become a safety requirement.

For that reason, cooperation and public production may deserve a different hearing. Monopoly worries us partly because protected human organizations become complacent. Government production worries us partly because bureaucracies can become sluggish, self-protective, and indifferent to feedback. But machine-run institutions would not necessarily decay for the same reasons. Their failures could be different: bad objectives, distorted data, brittle models, surveillance, manipulation, regulatory capture, or optimization against the public interest.

That distinction is important. Consumers are not only consumers. They are citizens, workers, cultural beings, and possible subjects of power. A coordinated production system might deliver abundance while narrowing choice, concentrating authority, manipulating desire, or making exit impossible. The question is which arrangements produce abundance while preserving human freedom, dignity, and accountability.

The old saying “money doesn’t grow on trees” is revealing in this context. The phrase is meant to remind us that value does not appear without human cost. But human cost has never been the whole story. Much of economic life depends on finding leverage over processes that already produce without us: sunlight falling, gravity pulling, rivers flowing, microbes fermenting, seeds turning soil and water into calories. The human contribution is often not to perform the central transformation ourselves, but to discover where a small intervention can redirect a much larger productive force.

A fruit tree is one such force. The farmer does not perform the central act of production. The farmer creates the conditions under which an independent productive system can operate. Soil, water, sunlight, pruning, grafting, pest control, and harvest are acts of cultivation rather than manufacture. The tree performs the transformation. It takes simple inputs and, through machinery we benefit from more than we understand, produces something valuable, complex, and repeatable.

Artificial intelligence may represent a continuation, or perhaps a completion, of this process. Human beings have always created wealth by arranging productive systems outside themselves. But until now, we still had to remain inside the loop: watching, moving, repairing, judging, coordinating, and intervening when the world became irregular. Artificial intelligence may fill more of those gaps. Human beings may no longer be required as the universal patch between intention and outcome.

The plant analogy clarifies both the promise and the danger of a machine-based economy. A tree does not ask whether its fruit serves human flourishing. It grows according to its nature and environment. If planted carelessly, it can exhaust the soil, spread invasively, or produce abundance captured by only a few owners. Likewise, artificial intelligence will not automatically produce a humane economy merely because it produces efficiently. The question is not only whether we can grow more fruit. It is what kinds of trees we plant, who owns the orchard, and what values shape the harvest.