I remember the height of the dot-com boom when lobbyist Tony Podesta changed his firm’s name to Podesta.com. He was not alone in his irrational exuberance. Most investors and many entrepreneurs went along. They were not wrong; just too early. A few years later we got Google, Facebook, the iPhone, and the Internet boom.
It’s impossible to predict the future, but it’s likely AI could follow a similar path: a few hiccups now, followed by robust, sustained growth.
To be sure, AI hype is the order of the day. In 2025, we were bombarded with announcements that artificial general intelligence (AGI) that could do almost anything was right around the corner. And forget jobs; they were a buggy whip relic.
But new technology hype ahead of reality is pretty common. Originally coined by Gartner, the technology hype cycle holds that most new technologies go through phases. After a new technology is introduced, there is often a phase of “Inflated Expectations” when the technology is seen as game-changing, or in the case of AI, universe changing. But most technologies follow a predictable path to the “Trough of Disillusionment” when original expectations don’t manifest themselves immediately. But many technologies keep going to the “Plateau of Productivity,” where the technology is widely used and has important benefits to society, just less than the inflated expectations.
After a new technology is introduced, there is often a phase of “Inflated Expectations” when the technology is seen as game-changing, or in the case of AI, universe changing.
AI appears to be moving from “inflated expectations” to “trough of disillusionment.” Many latched onto the findings of a recent MIT study that found that most companies using AI are not getting the results they want. And there will likely be more of these “see, it’s not all its cracked up to be,” claims. What else should we expect when so many pundits and AI leaders told us that AI is more important than the invention of fire?
There’s another aspect of the narrative around AI, though. In earlier years, technology hype was mostly about the massive benefits of a new innovation. Today, the inflated expectations are mosly about how bad things will get — a phenomenon the Information Technology and Innovation Foundation (ITIF) has termed the “tech panic cycle.” As one tech journal wrote: “When a technology is subject to overhype, it is easy for policymakers to assume vast and unexpected impacts, many of which could be, or likely will be, negative.” This is especially true as American culture has shifted from its pro-technology, pro-business orientation to one that is decidedly negative. After all, we are regularly bombarded with (incorrect) messages that AI is biased against minorities, kills jobs, violates privacy, jacks up electric rates, concentrates power into the hands of a few tech bros, floods the internet with garbage, exploits labor, undermines learning, powers misinformation and disinformation, exacerbates economic inequality, deskills professions, powers the surveillance state, steals artist’s content, and is leading to financial/economic collapse.
The reality is that new technology has always presented challenges (e.g., cars kill people, computers put typists out of work, etc.), but the benefits have vastly outweighed the costs. And America got to be number one technologically because, more than any almost any other nation, we believed that.
But what about immediate impacts? What about the benefits? Well, one benefit is the massive investment the industry is making in the American economy. According to McKinsey, the United States is expected to account for 40 percent of global spending on AI in 2025 – $592 billion, which equates to almost 2 percent of America’s GDP. And as of August 2025, $40 billion had been spent on U.S. data center construction. All of that creates jobs.
But what about AI stock prices? Are we headed for a bust or another dot-com-like crash? If I knew the answer to that question, I’d be a rich man. Stock valuation, however, is not the right question. Just as in the early 2000s, although there were some high profile dot-com bankruptcies, the industry grew overall. The same is surely true today. But which exactly are the firms that will succeed is something that only time will tell.
So what should policymakers do to ensure the growth of AI? First and foremost, they should do no harm. Don’t give in to the Luddites who want to take a page out of the failed EU playbook and tax and regulate AI to keep it in its cradle. This goes doubly in response to the AI job doomers who would have us believe that we will be lucky if most of us are on the Universal Basic Income dole in a few years. The reality is that new technology has never led to fewer jobs. But it has always led to more income. What Congress should do is improve the workforce development and worker adjustment system.
What should policymakers do to ensure the growth of AI? First and foremost, they should do no harm.
Second, we should invest in data centers. AI doesn’t work without data centers. And yet many seem to think that building data centers is the end of the world. Since when did America not want to build? Congress should press the Federal Energy Regulatory Commission to require utilities to include new technologies in transmission review, and to use new technology when it is the best solution. And it should pass the House Speed Act to reform NEPA.
Third, AI has been built on open Internet data. But the next stage requires much more data and deeper data sets, in particular applications, such as health, education, robotics, urban systems, and logistics. Much of this data is locked up in data silos. The federal government should launch sector-specific data strategies to facilitate more data collection and sharing, such as by reforming restrictive data privacy laws, developing model data contracts, improving data standards, and encouraging data “donation” where individuals can contribute their personal data for the common good.
Fourth, Congress should be more vocal about standing up to foreign governments that launch attacks on U.S. AI companies just because they are American.
Pogo famously said, “we have met the enemy and he is us.” This is certainly true with AI. If we stop attacking it and instead enable it, there is no reason why the United States can’t maintain its global AI leadership and significantly boost U.S. living standards.
Robert D. Atkinson is President of the Information Technology and Innovation Foundation.