The growth story is the easy part
When the White House compares the rise of artificial intelligence to the industrial revolution, the message is straightforward: growth, jobs, long-term prosperity.
It is an effective narrative. And above all, a politically comfortable one.
Growth is the least costly argument to defend. It allows future gains to be projected without explaining how those gains will be distributed or who will absorb the short-term losses. In the case of AI, that omission matters. The side effects are not theoretical. They are already visible.
Talking about technological revolution without talking about economic structure is telling only half the story.
Past technological revolutions created winners first
Economic history is unambiguous on one point: major technological shifts always produce winners before they produce balance.
The industrial revolution enriched owners of industrial capital long before wages, safety standards, or social mobility improved. The digital revolution concentrated value around dominant platforms before its benefits diffused more broadly.
In both cases, the promise of shared prosperity arrived late, and unevenly. Corrective mechanisms—regulation, redistribution, training—almost always followed the initial capture of value.
There is no historical precedent in which a transformative technology spontaneously produced an equitable distribution of gains during its adoption phase.
AI is structurally different from previous waves
Artificial intelligence differs from previous technological revolutions in its economic structure.
First, its fixed costs are exceptionally high. Training, deploying, and maintaining competitive models requires massive investments in infrastructure, data, and energy. These barriers to entry mechanically favor already dominant players.
Second, its competitive advantage compounds. AI systems improve through accumulation: more data, more usage, more feedback. This creates powerful lock-in effects where leaders strengthen their position simply by existing.
Finally, AI depends heavily on private infrastructure. Unlike electricity or railroads, modern AI is built largely on platforms controlled by a small number of companies. Innovation does not circulate freely. It is mediated.
From innovation to dependency
Framing AI as a productivity tool obscures a more uncomfortable reality: it already functions as strategic infrastructure.
Businesses, public institutions, and even governments are becoming dependent on capabilities they do not control. Models, training data, and deployment pipelines are rarely public, rarely interoperable, and rarely substitutable in the short term.
This dependency creates a durable imbalance. Whoever controls the infrastructure sets the rules, the pricing, and the pace of innovation. History shows that this kind of power is rarely temporary.
Comparing AI to the steam engine without addressing dependency ignores the central question: who owns the machine?
Jobs will change, but power will concentrate first
Public debate tends to focus on employment. Which jobs will disappear? Which new roles will emerge?
These are legitimate questions, but they arrive too early. Before reshaping work, AI reshapes economic power. It automates cognitive functions while centralizing decision-making capacity.
Previous technological waves destroyed jobs, but rarely centers of power. AI threatens to shift both simultaneously. This is not merely a reskilling issue. It is a governance issue.
History suggests that when power concentrates rapidly, social adjustment becomes more costly and more conflictual.
What the growth narrative carefully avoids
Optimistic messaging emphasizes productivity while sidestepping three critical issues.
First, the real distribution of gains. Higher productivity does not imply proportional income growth for most workers. Recent history suggests the opposite.
Second, market concentration. The more a technology is framed as inevitable, the less its concentration effects are questioned.
Third, regulatory timing. Waiting until negative effects are fully visible almost always means intervening too late.
These blind spots are not accidental. They make the story easier to accept.
What history actually suggests
History does not say AI is destined to fail. It says that without early intervention, its benefits will be captured by a minority.
Previous technological revolutions were regulated not because they were new, but because they had already produced imbalances that were difficult to reverse. Repeating that sequence is not inevitable, but it is the default trajectory.
Promising an AI revolution without addressing concentration is not analysis.
It is a bet.
And history suggests that this bet rarely pays off for the majority.
