Lessons from Tokyo: Why the AI Hype Risks Repeating the 1990s

Do remember how at the end of the eighties the world was absolutely convinced that the future spoke exclusively in Japanese? It was a time when the land beneath the Imperial Palace in Tokyo was worth more than all the real estate in California. But Japan's 1990 story is not merely an economic chronicle — it is a harsh reminder of what happens when a "grand narrative" suddenly collides with harsh reality. Today, as we gaze mesmerized at Nvidia's capitalization charts and OpenAI's recurring promises to achieve AGI in a couple of years, this Japanese ghost of the past seems more relevant than ever.

A grand narrative is not simply a belief in success; it is a collective hallucination in which growth is considered infinite and technology is deemed all-powerful. In 1990, the Japanese believed that their management model and semiconductor dominance would make them the world's leading economy. Today, our grand narrative is built around the idea that a large language model can solve any problem — from code writing to cancer treatment.

We are pouring hundreds of billions of dollars into infrastructure, building data centers the size of cities, and hoping that the exponential curve will never flatten into a plateau. But after each brilliant cherry blossom bloom, inevitably comes a period when petals begin to fall.

The problem with any technology bubble is not the technology itself, but the expectations it generates. In late eighties Japan, innovation was real — their electronics and automobiles truly were changing the world. But the price of these achievements was inflated to the point of absurdity.

In the AI industry, we see a similar picture. We have fantastic tools like GPT-4 or Claude 3, but the market values them as if they have already replaced all of humanity. When expectations outpace the actual monetization possibilities, that fragile dream is born, the one that Chinese analysts write about.

And when this dream is interrupted, it is not corporations that pay the price, but ordinary people whose careers and hopes were tied to this growth.

If we look carefully at how the bubble burst in 1990, we will see that the trigger was not the disappearance of technology, but the realization that it could not sustain the insane asset valuations attributed to it.

Right now, the AI industry is at a point where investors are beginning to ask uncomfortable questions about return on investment (ROI). Training each new model costs orders of magnitude more than the previous one, while the quality improvement becomes ever less obvious. This is a dangerous zone.

If the grand narrative about AI as the "new oil" wavers, we will see the same empty offices and lost decade that Japan went through.

Yet there is irony here. Those who survived the 1990 crash became the foundation for many modern technologies. A crash clears the market of random passengers and leaves behind those who truly create value.

Perhaps we too need the dust to settle a bit, and loud headlines to give way to boring, but functioning business models. The era of grandiose promises always ends the same way — with a return to the numbers in accounting reports.

The key here is not to fear the crash, but to understand: behind every "revolutionary" leap stands someone's willingness to foot the bill when the music stops.

The bottom line: Are we ready for AI to be simply a very good tool, rather than a deus ex machina capable of justifying trillion-dollar valuations?