Every morning across financial media, the narrative gets scrubbed, polished, and piped into your feed with the same breathless urgency. The message is always identical. Artificial intelligence is an unstoppable economic tide lifting every boat, a Star Wars-level technological force destined to rewrite capitalism overnight, and if you are not aggressively bidding up every stock touching silicon or server racks, you are financially illiterate.
It is a clean, comforting story for people who collect management fees. It is also a multi-trillion-dollar distraction from basic economic gravity.
I have spent the last two decades watching institutional capital chase shiny objects right off a cliff. I watched telecom giants bury oceans of fiber-optic cable in the late nineties under the exact same religious fervor. I watched corporate balance sheets incinerate capital during the enterprise software bloat cycles. Every time, the cheerleaders on television promised that this era was different. This time, productivity would outrun arithmetic.
Let us look past the marketing departments and examine the cold, unvarnished P&L statements. The current market narrative rests on a fundamental falsehood: that surging infrastructure expenditure equals guaranteed enterprise value creation. Wall Street loves total addressable market projections because they let analysts pull massive numbers out of thin air and discount them back to the present day to justify absurd forward multiples. They treat capital expenditure by hyperscalers like a sacred covenant.
Here is what the consensus refuses to say out loud. Building multi-billion-dollar data centers powered by nuclear plants or dedicated natural gas turbines is not a sign of triumphant demand. It is an act of desperate economic consumption. When a tech titan spends fifty billion dollars on specialized chips in a single fiscal year, that is not an investment with a clear, linear path to margin expansion. That is an enormous fixed cost that must be amortized against software margins that are already beginning to compress.
The lazy consensus tells you that artificial intelligence will eliminate corporate headcount and send profit margins straight into the stratosphere. The reality on the ground is infinitely messier. I have sat in boardrooms over the past year watching Fortune 500 executives throw millions at enterprise software licenses because they were terrified of looking like luddites to their boards. Six months later, those same executives are staring at bloated software budgets, integration bottlenecks, and employees who spend half their day prompt-engineering emails that used to take thirty seconds to write.
Productivity is not measured by how many tokens an enterprise consumes. Productivity is measured by output per unit of capital input. When you factor in the staggering cost of compute, power consumption, talent acquisition, and continuous model retraining, the return on investment equation starts looking deeply unromantic. For the vast majority of traditional enterprises, these tools are turning into expensive marginal improvements wrapped in heavy licensing fees.
The market has priced in a utopian future where every legacy business successfully transforms into an algorithmic powerhouse. The historical playbook suggests a different outcome. A tiny handful of infrastructure gatekeepers will capture the majority of the economic rents, while thousands of software buyers will discover they bought a very expensive utility bill that did not move their top line by a single percentage point.
Admitting this truth does not make me a techno-pessimist. I use these tools daily. They are remarkable pattern-matching engines. They are phenomenal at writing boilerplate code, summarizing messy transcripts, and accelerating initial research phases. But a better autocomplete does not justify a thirty-times sales multiple on companies whose core business models are getting commoditized by open-source alternatives.
The real trade right now is not blindly buying the hype cycle. The real trade is identifying which companies are treating these technologies as operational efficiencies rather than magical growth fairy dust.
Stop asking how artificial intelligence is going to change the world. Start asking who is going to pay the electricity bill when the venture capital subsidies run dry.
