(Yicai) Dec. 5 -- The artificial intelligence boom, which is the fastest technological boom ever, has created the strangest economics in history.

The AI boom is driven by something strange: reality itself is running ahead of economics. This is the contradiction at the heart of AI, the only gold rush where the gold keeps getting discovered while prices keep collapsing.

AI is not a bubble in the classical sense, as the technology, demand, and adoption are all real. What is not yet real is the business model, and this creates an unfamiliar phenomenon: a valuation wave driven not by hype, but by the speed of technological progress itself.

There are six structural forces making this AI boom unlike any bubble in economic history. And more importantly, these forces reinforce one another, forming a self-sustaining macro loop.

The first is time mismatch. Technological cycles usually follow the sequence: breakthrough, product, monetization, profit, and then scale. However, AI compressed this sequence into a continuous breakthrough loop.

Corporate budgets still follow annual cycles, but capabilities now advance on three-to-six-month intervals. This mismatch is structural, as monetization cannot keep pace.

The second is value chain mismatch, as capital flows upward while value flows downward. As capability accelerates, capital floods toward model training, graphic processing unit clusters, alignment teams, and compute-heavy research, which are essential but structurally low-margin layers.

Meanwhile, enterprise budgets concentrate on workflow automation, vertical AI, industry-specific solutions, and physical AI and robotics.

A decade ago, the cloud industry saw value migrate upward into infrastructure. But AI flips this logic entirely. Time mismatch reinforces value chain mismatch, with layers improving fastest attracting the most capital, even though they capture the least value.

Much of the disagreement around an "AI bubble" comes from these divergent analytical lenses, as traditional investors evaluate capital cycles, frontier technologists focus on capability trajectories, while enterprise operators anchor on workflow value. The debate is less about facts than about frameworks, as different observers are looking at different layers of the same system.

The third is the capability-price paradox. Under normal economics, better technology yields higher pricing power, but AI does the opposite. Each capability jump expands substitutes, accelerates open-source diffusion, weakens differentiation, and makes intelligence feel like a utility.

Over the past 18 to 24 months, token prices have fallen between 20 percent and 70 percent across major providers, even as model capabilities rose by orders of magnitude in reasoning, multimodality, and context. No prior technology cycle saw deflation caused directly by progress itself.

The fourth is infinite supply. Classic bubbles depend on scarcity, while AI is built on abundance: infinite model replication, zero marginal costs, limitless substitutes, and frictionless distribution.

Scarcity collapses faster than business models adapt. Economically, this dynamic resembles commoditized energy markets more than past tech cycles. Infinite supply is good for humanity, but it is much less kind to margins.

The fifth is mandatory spending, a competitive trap no one can escape. Infinite supply fuels competition, which accelerates capability, which, in turn, accelerates infrastructure spending. At scale, this creates the strangest feature of the AI economy.

If an enterprise advances, the others must follow; if one cuts prices, everyone must recalibrate their economics; and if the operating system and agent races escalate, the GPU fleet must expand.

In past technology races -- mobile, cloud, and semiconductors -- capex was heavy but discretionary. But AI changes this, as standing still is equivalent to conceding the platform's future. AI may be the first industry where the cost of stopping exceeds that of continuing, even when short-term returns are unclear.

The sixth is inverted economics, with costs concentrated at the top and profits at the bottom. AI is the first industry where the highest-cost layer sits above the highest-margin layers.

At the top of the stack, model training is capital-intensive and commoditizing, and cloud inference is expensive and increasingly price-pressured. At the lower layers, workflows, vertical AI, decision and automation platforms, and physical AI and robotics are relatively cheaper to build, harder to replicate, and structurally higher-margin.

This inversion is not a market failure but a logical outcome of rapid capability expansion, misaligned capital flow, pricing collapse, infinite substitutes, and mandatory escalation. Economic gravity has flipped.

So, is AI a bubble?

If "bubble" means fake technology, imaginary demand, and purely speculative narratives, then no. But if "bubble" means valuations ahead of sustainable cash flow, capital structurally locked in cost centers, pricing collapsing as capability rises, infinite supply crushing margins, forced spending without optionality, and economics inverted across the stack, then yes.

However, AI is not a speculative bubble. It is a structural bubble created not by illusion but by the contradictions of its own success.

This is the first cycle in economic history where technology outruns business models, value creation and value capture diverge, capability undermines price, supply expands indefinitely, competition enforces mandatory spending, and economics invert structurally.

It is not a bubble that bursts, it is a bubble that resolves, as economics eventually catch up to capability and value consolidates where scarcity and defensibility can survive.

The bubble is not only in the numbers, but also in the frameworks we use to understand a technology evolving faster than our mental models can adapt.

The author of this article is Lu Duowei, an investor, researcher, and founder of FJ Insights. He previously worked at international financial institutions and sovereign wealth funds, focusing on global structural trends, AI economics, and frontier technology innovation.

Editor: Futura Costaglione