The High Price of Free: How AI Broke the Model of Technological Progress
The great paradox of the AI revolution is that it began by ignoring one of the most reliable laws of technological development. New technologies start expensive, scarce, and restricted. The first mobile phones were not handed out freely to millions. They were costly tools used initially only where they created unmistakable economic value. Early adopters were businesses, governments, and professionals who could justify the cost because the devices made them more productive. Profitability emerged first, mass adoption later. Technology proved itself in the marketplace before it was democratized.
Large language models took the opposite path. They were introduced not as premium tools, but as free, abundant public goods. Billions of dollars of compute, research, and engineering were poured into products that generated no revenue and could not pay for themselves. This was not a deliberate business strategy. It was the by-product of a monetary era in which capital was so cheap, so plentiful, and so desperate for returns that it flowed into projects without requiring a credible path to profit.
The result is an industry built upside down, mass adoption first, economic logic second. And such an inversion comes with consequences.
The Early Abundance
By making AI free at the start, companies removed the natural filter that once determined whether a new technology had real-world usefulness. In the era of early mobile phones, cost itself acted as a constraint, only productive users could justify the expense. That filter ensured the technology grew along sustainable lines. Every new investment had to pass a simple test, does someone pay for this because it creates measurable value?
AI skipped this test. With free access, use exploded, but usage is not the same as value. Billions of queries prove only curiosity, not economic viability. And because the dominant platforms were subsidized by investors rather than customers, no one actually knows which applications genuinely matter. The market cannot reveal the answer because the market has not been allowed to price the service. This is not innovation. It is opacity.
When Free Means Unprofitable
AI can only become sustainably profitable when it replaces human labour in specific, economically meaningful tasks. That is the harsh criterion. And replacing labour is far more difficult than the narrative suggests. Most work involves judgment, context, emotion, or interaction, areas where AI remains shallow and brittle. Some jobs will be automated, eventually. But the industry’s profitability depends not on ideology, hype, or raw model power, but on whether a business can actually fire workers and keep quality stable.
That threshold has not been crossed for most firms. It will take time. It may take decades. And during that period, the companies that have given away their product will bleed money while searching for the profitable applications that should have been discovered first.
A correction is inevitable. It is simple arithmetic. No industry can survive on investor patience indefinitely. Especially not one whose cost structure involves hyperscale data centres consuming power like small nations.
The Distortions of Cheap Money
The decision to give AI away for free was not a stroke of genius. It was the product of a monetary boom. Years of ultra-low interest rates and aggressive money creation overwhelmed productive investment opportunities. When capital cannot find profitable uses, it floods into speculative ones. AI became the perfect outlet, futuristic, exciting, and, crucially, so expensive to build that only firms flush with investor cash could participate.
In a healthier monetary environment, such models would have been far more expensive for users. Firms would have been forced to charge early, and only the economically valuable applications would have survived. Instead, we have a technology that is both ubiquitous and untested, popular and unprofitable, celebrated and structurally unsustainable.
This is the danger, an innovation model detached from economic reality produces illusions of progress, not progress itself.
A Future Decided by the Bust
We now live in a technological twilight where nothing can be priced honestly. We cannot know which AI applications are truly valuable, because everything is subsidized. We cannot know which companies deserve to survive, because investor money protects them from market reckoning. And we cannot know which uses of AI will matter long-term, because abundance hides inefficiency.
Only a major correction, a withdrawal of easy money, a collapse of speculative valuations, a return to financial gravity, can reveal the truth. When the subsidy ends, only profitable uses remain. Only then will AI find its natural place in the economy.
The irony is that the future winners may not be the companies with the largest models, the flashiest demos, or the most users. The winners will be those who can survive the free era long enough to reach the paid era. Whenever that arrives.
Technological progress is not the problem. The problem is the environment that distorted its path. We are not witnessing a revolution. We are witnessing the consequences of excess. And when the excess ends, we will finally learn what AI is actually worth.
