Sometime in late 2022, a friend who works in accounting forwarded me an email from her firm's leadership. The subject line announced a new "AI initiative." The body was vague — something about exploring transformative potential and staying ahead of the curve. She asked me if I understood what it meant. I told her honestly that I didn't think they did either.
That email has since multiplied into thousands of variations. Across industries, from hospitals to hedge funds, from public schools to podcast networks, organizations are announcing AI strategies, hiring AI officers, and rearranging budgets around a technology that most decision-makers still struggle to define clearly. This is what a boom looks like from the inside: not a clean narrative of progress, but a messy, uneven, sometimes confusing rush toward something people feel they cannot afford to miss.
The current AI boom is real. The technology has genuinely crossed thresholds that seemed distant just five years ago. Large language models produce coherent text. Image generators create visuals that would have required skilled professionals a decade ago. Code assistants write functional software. These are not illusions or parlor tricks. But a boom is not the same thing as a settled transformation, and understanding the difference matters enormously for anyone trying to make sense of this moment.
What Makes This Boom Different
Technology booms follow recognizable patterns. The railroad mania of the 1840s, the radio craze of the 1920s, the dot-com bubble of the late 1990s — each shared certain features: rapid capital inflow, speculative overbuilding, breathless media coverage, and a gap between what the technology could actually do and what people imagined it would do.
The AI boom shares these features, but it also carries something previous booms did not: the technology at its center is directly relevant to how people think, write, communicate, and create. Railroads moved goods. Radio moved sound. The internet moved information. AI moves into the territory of cognition itself. This makes the current moment feel more personal, more unsettling, and more difficult to evaluate calmly.
When a new tool changes how fast a train travels, the implications are mostly economic. When a new tool changes how a paragraph gets written, the implications touch identity, employment, education, and self-understanding. People are not just asking whether AI is useful. They are asking whether their particular skills still matter, whether their children should study certain subjects, whether the career paths they imagined still exist. These questions deserve honest answers, not marketing slogans.
The Gap Between Capacity and Application
Right now, the most striking feature of the AI landscape is the distance between what the technology can demonstrably do and how it is being deployed in practice.
The capacity is impressive. Models can summarize documents, translate between languages, generate code, analyze data patterns, and produce creative content at a level that would have seemed like science fiction to most people in 2015. Research labs continue to push boundaries, and each new model release brings genuine improvements.
But application is another matter. Most organizations adopting AI today are not reimagining their core operations. They are layering AI tools onto existing workflows, often with modest results. A law firm might use AI to scan documents for key terms, saving associates hours of tedious reading. A marketing team might use it to draft social media posts, then spend nearly as much time editing the output as they would have spent writing from scratch. A customer service department might deploy a chatbot that handles simple queries well but frustrates customers with anything complex.
These are real improvements, sometimes meaningful ones. But they are incremental, not revolutionary. The gap between the revolutionary rhetoric surrounding AI and the incremental reality of most current deployments is one of the defining features of this boom. It does not mean the revolution will never arrive. It means we are in an awkward middle phase, and recognizing that phase for what it is helps prevent both premature despair and unwarranted euphoria.
Who Benefits, Who Waits, Who Worries
Booms do not distribute their benefits evenly. The AI boom is no exception.
Large technology companies — the ones building foundation models and controlling cloud infrastructure — are positioned to capture enormous value. They have the capital, the data, the talent, and the computing resources required to train and deploy these systems at scale. Their stock prices reflect this positioning.
Startups that build genuinely useful applications on top of existing models can also thrive, though many face an uncomfortable dependence on the platforms they rely on. When OpenAI changes its pricing or Google adjusts its API terms, downstream businesses feel the impact immediately.
Meanwhile, workers in certain fields face genuine uncertainty. Not all jobs are equally vulnerable — roles requiring physical dexterity, complex human interaction, or deep contextual judgment remain difficult to automate. But positions centered on routine information processing, basic content generation, and predictable pattern recognition are already shifting. Freelance translators, entry-level copywriters, junior analysts — people in these roles are not imagining the pressure. They are living it.
The worry is not that AI will suddenly replace millions of workers overnight. It is that the transition will be gradual enough to feel unremarkable to people not directly affected, yet fast enough to leave those who are affected with little time to adapt. Booms tend to move wealth and opportunity toward those who already have capital and technical literacy. The AI boom appears to be following this pattern.
The Infrastructure Nobody Talks About
Beneath the visible products — the chatbots, the image generators, the code assistants — lies a physical infrastructure that rarely makes headlines but shapes the boom's trajectory.
Training large AI models requires enormous computing power. That power requires data centers, which require land, electricity, water for cooling, and specialized chips. The demand for these resources has grown so rapidly that it is straining electrical grids in some regions, driving up chip prices globally, and creating bottlenecks that no amount of software innovation can bypass.
Nvidia's extraordinary revenue growth tells this story clearly. The company designs the graphics processing units that have become essential for AI training, and its market valuation reflects the reality that right now, computing hardware is the scarce resource in this ecosystem. Software moves fast. Hardware takes time to manufacture, install, and power.
This infrastructure layer also raises questions about environmental impact and resource allocation that the boom's enthusiasts often gloss over. Data centers consume significant energy. AI workloads are computationally intensive. As deployment scales, the environmental footprint scales with it. These are not reasons to abandon the technology, but they are factors that honest accounting must include.
Learning From Previous Booms
History does not repeat itself, but it offers useful echoes.
During the dot-com boom, the rhetoric was similarly grand. The internet would change everything. Many companies with flimsy business models and no path to profitability attracted enormous investment because they had a website and a story about the future. When the bubble burst in 2000, billions of dollars vanished, companies collapsed, and thousands of workers lost jobs.
But the internet did change everything. It just took longer than the optimists predicted, and the companies that ultimately defined the digital era — Google, Amazon, Facebook — were either just beginning or had not yet been founded when the bubble peaked. The technology was real. The timeline was wrong. The winners were not who most people expected.
A similar dynamic may be unfolding with AI. The technology is real. Some of today's most prominent AI companies may not survive. Many of the applications currently attracting investment may prove unprofitable or unnecessary. The most transformative uses of AI might not yet exist. The boom will almost certainly cool — perhaps sharply — but the underlying capabilities will persist and continue improving.
The practical lesson is not to dismiss AI because the hype exceeds reality. It is to approach the technology with clear eyes: experiment seriously, invest cautiously, remain skeptical of grand promises, and pay attention to what actually works in practice rather than what sounds impressive in a pitch deck.
What Actually Matters Right Now
For individuals trying to navigate this moment, several things seem genuinely important.
First, develop a working understanding of what current AI systems can and cannot do. This requires direct experience, not just reading about the technology. Use the tools. Push them to their limits. Notice where they fail. This kind of hands-on literacy is far more valuable than abstract opinions about whether AI is overrated or revolutionary.
Second, think carefully about which of your skills are rooted in routine information processing and which draw on deeper capabilities — judgment, relationships, creativity, physical presence, contextual understanding. The former are under pressure. The latter remain durable, though they may evolve as AI handles more of the routine layer.
Third, be wary of anyone selling certainty. The trajectory of AI development is genuinely uncertain. Capabilities may plateau for years. Unexpected breakthroughs may accelerate progress in directions no one anticipated. Regulatory responses may reshape the landscape. Anyone who tells you they know exactly how this will unfold is selling something.
Finally, remember that technology adoption is a social process, not just a technical one. Organizations change slowly. Regulations take time. Cultural norms shift gradually. Even when a technology is technically capable of transforming an industry, the actual transformation often takes a decade or more because the surrounding systems — training, standards, legal frameworks, trust — need time to catch up.
The Boom Will End. The Technology Won't.
Every boom ends. The excitement fades, the capital tightens, the breathless headlines slow. This is normal and even healthy. Booms attract speculation and overextension. Corrections prune the unrealistic projects and leave the viable ones with more room to grow.
The AI boom will likely follow this pattern. Within the next few years, we will probably see a cooling period — perhaps triggered by disappointing financial returns, perhaps by a regulatory shock, perhaps simply by exhaustion with the hype cycle. Some high-profile AI companies will struggle. Many startups will close. The phrase "AI-powered" will stop being a selling point and become a background assumption, the way "internet-enabled" did after the dot-com era.
But the underlying technology will not disappear. The models will continue to improve. Useful applications will continue to emerge. The infrastructure will continue to expand, though perhaps more deliberately. The capabilities that exist today will become cheaper and more accessible, and new capabilities will emerge from research that is quietly proceeding regardless of the hype cycle.
The people who benefit most from this boom will not be those who bought the most extreme promises or those who dismissed the entire phenomenon as a fad. They will be those who paid close attention, experimented honestly, and built real understanding of what the technology can do today — not what someone promises it will do tomorrow.
Living through a boom is disorienting. The noise is loud, the signals are unclear, and the pressure to act or ignore is constant. The best response is neither to chase nor to retreat, but to observe carefully, think clearly, and engage with the technology on your own terms rather than on the terms of whoever is shouting loudest.