The glow of the terminal didn’t flicker, but to David, it felt like a failing pulse.
It was 3:15 AM. In the silent sprawl of his tech startup’s office, the only sound was the low, rhythmic hum of the HVAC system. David sat staring at a spreadsheet that felt heavier than lead. Six months ago, his company had pivot-shifted its entire infrastructure toward generative artificial intelligence. They bought the chips. They licensed the massive models. They promised investors a revolution.
Tonight, the math refused to lie anymore. The cost to answer a single user query was ten times higher than traditional computing, while the revenue generated was a trickle of trial subscriptions. The runway was vanishing.
David’s story isn’t lonely. It is repeating across the globe, hidden behind the sleek glass facades of Silicon Valley, London, and Tokyo.
We have entered the era of the great AI exuberance. It feels intoxicating. It feels inevitable. But a quiet, devastating warning issued by the Bank for International Settlements (BIS)—the central bank of central banks—suggests that this blinding light is creating a massive economic shadow. The risk is no longer just about whether technology works. The risk is a multi-trillion-dollar investment bust that could freeze innovation for a decade.
The Ghost in the Machine
To understand how we arrived at this precipice, we have to look past the marketing videos of talking avatars and automated coding assistants. We have to look at the plumbing.
Think of the current AI boom like building a continent-spanning railroad system before anyone has invented a locomotive that can carry freight profitably. Investors are pouring hundreds of billions of dollars into data centers, specialized semiconductors, and massive electrical grids. They are laying the tracks at a frantic, desperate pace. Every venture capitalist is terrified of being left at the station.
But railroads require enormous upkeep. The BIS report highlights a fundamental asymmetry: the capital expenditure required to build and maintain AI infrastructure is staggeringly high, while the commercial use cases remain largely experimental.
Let’s look at the cold reality of the balance sheet. A traditional software business enjoys high margins because once the code is written, distributing it costs almost nothing. Artificial intelligence upends this economic law. Every single prompt requires massive computational power. Every interaction burns electricity, strains cooling systems, and demands expensive chip time.
When a company relies on a technology where the operational cost scales almost linearly with use, the traditional tech playbook breaks.
The BIS warns that if the revenue generated by these tools fails to scale rapidly enough to justify the eye-watering infrastructure costs, the market face-plants. Investors will pull back. Capital will dry up. The result won't just be a few bankrupt startups; it will be a systemic chilling effect across the global financial sector.
A History of Bright Horizons
We have seen this movie before. The script rarely changes.
In the late 1990s, the world discovered the commercial internet. The excitement was entirely justified; the internet did change everything. But the timeline was profoundly wrong. Investors poured billions into telecom companies to lay millions of miles of fiber-optic cables beneath the oceans. They assumed the demand would explode overnight.
It didn't.
The demand took another seven to ten years to catch up with the infrastructure. In the meantime, the Dot-Com crash wiped out trillions of dollars in wealth. Telecom giants went belly-up. The fiber-optic cables sat dark in the deep ocean, referred to by economists as "dark fiber." It took a decade of painful restructuring before the world actually began to utilize the infrastructure that caused the bust.
The BIS suggests we are making the exact same temporal error with artificial intelligence.
The technology is impressive. The potential is real. However, the gap between hype and commercial utility is widening into a canyon. Companies are buying sophisticated tools to do tasks that could be accomplished with a simple spreadsheet or a well-trained human worker. They are paying a premium for novelty, and novelty has a notoriously short shelf life.
Consider a hypothetical bank trying to automate its customer service. It spends $50 million integrating a cutting-edge large language model. The system is fast, conversational, and occasionally brilliant. But it also hallucinates financial advice, requiring a team of human supervisors to audit its outputs constantly. The bank hasn’t saved money; it has merely shifted its labor costs from customer service agents to expensive data engineers.
When the CFO realizes the return on investment is negative, the project gets canceled. Multiply that realization across thousands of enterprises, and the tech sector faces a sudden, catastrophic drop in demand.
The Real Stakeholders
The danger of an investment bust isn't confined to wealthy venture capitalists losing their shirts. The stakes are deeply human.
When an industry experiences an artificial boom, it distorts the entire economy. Talent shifts. Thousands of young engineers, data scientists, and researchers are currently abandoning crucial fields—like cybersecurity, hardware reliability, and basic systems engineering—to chase the AI gold rush. They are building models that generate corporate slide decks instead of fixing foundational vulnerabilities in our digital infrastructure.
If the bust occurs, these professionals face sudden displacement. Entire ecosystems of secondary businesses—the construction companies building data centers, the green energy providers scaling up grids, the local economies built around tech hubs—will feel the whiplash.
Central banks are watching this closely because the financial system is deeply interconnected. The companies manufacturing the chips and building the data centers are heavily leveraged. They borrow billions based on the assumption that demand will grow exponentially. If that demand flattens, the debt service fails. The ripples hit commercial banks, asset managers, and pension funds.
The BIS isn't telling us that the technology is a fraud. They are telling us that our impatience is dangerous.
The Balance of the Scale
We are trapped in a cycle of extreme narratives. On one side, the techno-optimists claim that we are months away from human-level artificial intelligence that will solve every human crisis. On the other side, the cynics dismiss it as a glorified autocomplete machine.
The truth is far more mundane, and far more complicated.
Artificial intelligence is a powerful new tier of computing. It is an evolutionary step, not a magical one. It requires an immense amount of energy, physical space, and human oversight. Treat it like an infinite money machine, and you guarantee a crisis. Treat it like an expensive, specialized tool that requires careful cost-benefit analysis, and you build sustainable progress.
Back in the quiet office, David closed his laptop. He looked out the window at the city skyline, where the logos of major tech firms glowed against the dark sky. He knew what he had to do. The pivot to AI was an illusion designed to please a manic market. Survival meant scaling back, focusing on core software that actually solved real problems for real people, and ignoring the siren song of the hype cycle.
The infrastructure of the world cannot be built on enthusiasm alone. Concrete requires cement, data centers require profitable electricity, and investments require real returns. If we refuse to ground our expectations in economic reality, the market will eventually do it for us, and the landing will not be soft.
The screen went dark. The hum of the room remained, a cold reminder that the machines don't care about our expectations; they only consume what we feed them, until there is nothing left to give.
