What the Dot-Com Bubble Can Teach Us About AI 

If you spend any time online, it’s hard to miss the chorus of pundits declaring that artificial intelligence is the new dot-com bubble. The comparison is tempting. A small group of companies is worth staggering sums on the stock market, venture money is chasing any startup with “AI” in its pitch deck, and skeptics warn that many of these firms will never earn back what investors are paying today. 

But if you only compare the market hype, you miss one important element: Washington’s role in this boom looks very different from its role in the last one. During the dot-com era, the federal government helped create the conditions for the internet to take off, then mostly stood aside while private capital inflated the bubble. In today’s AI boom, Washington is pouring in more money, asserting more control, and trying to manage more risks all at once. 

That shift matters for anyone who cares about innovation, stable rules, and the long-term health of the U.S. economy. 

In the late 1990s, most of the big public investments that made the internet possible were already in the past. Beginning in the 1960s and 1970s, the Defense Department and the National Science Foundation funded the early networks that became today’s internet. By the time Pets.com and Webvan were buying Super Bowl ads, the government’s annual networking and information technology research budget, across all agencies, was only about $1.3 to $1.9 billion a year. Adjusted for inflation, that is roughly $2.5 to $3.5 billion in today’s money. 

What Washington did do during the dot-com years was set the rules of the game. In 1996, Congress passed the Telecommunications Act, which explicitly aimed to let “any communications business compete in any market against any other.” That deregulation opened the door to massive private investment in telecom and broadband infrastructure. In the same year, Congress and the courts interpreted Section 230 of the Communications Decency Act to give online platforms broad immunity for most user content, which reduced legal risk for new online services. 

Congress also cut capital gains taxes in 1997 from 28 percent to 20 percent for most long-term investments. That decision made speculative stock market gains more attractive just as Wall Street was rushing young internet companies to an IPO. The Federal Reserve, for its part, did not seriously try to lean against the bubble with tighter money. When the crash finally came in 2000 and 2001, Washington responded after the fact with interest-rate cuts and post-mortems. 

In other words, federal policy during the dot-com bubble combined modest new R&D spending with a broadly permissive environment for risk-taking. The government had planted the seeds decades earlier, then mostly let the market pour on the fertilizer. 

Today’s AI boom looks different in at least two ways. 

First, the federal government is spending far more, in real terms, on the technologies underneath the current boom than it spent on networking and computing during the dot-com era. According to the government’s own cross-agency figures, total federal networking and information technology R&D for fiscal year 2025 is a little over $11.1 billion. That is about three to four times the inflation-adjusted annual level around the turn of the century. 

Within that total, AI research is now a line item of its own. President Trump’s budget requested about $3.3 billion in non-defense AI R&D for 2025, including both “core” AI research and AI projects captured in other program areas. That AI slice alone is roughly the same size, in today’s dollars, as the entire federal networking and IT budget at the height of the dot-com boom. 

On top of those research budgets, Congress has moved into explicit industrial policy. The CHIPS and Science Act, passed in 2022 with bipartisan support, set aside roughly $52.7 billion over five years for semiconductor manufacturing and research. That money is now flowing into new chip fabrication plants and research centers that will power AI models for decades. Agencies like the National Science Foundation are also building a network of National AI Research Institutes, with new awards of roughly $100 million announced this year, to support long-term, open AI research and workforce development. 

Second, Washington is trying to manage the risks of AI much earlier in the technology cycle than it did with the internet. The 2023 White House executive order on AI directs agencies to set safety standards, evaluate high-risk systems, and think through ethical and national security implications. The Commerce Department is writing rules to guardrail the export of advanced chips and AI hardware to strategic rivals like China. Lawmakers in both parties are debating how to update liability rules, privacy protections, and competition law for an AI-driven economy. 

Some of this caution is understandable. AI systems are already in use in sensitive domains like healthcare, hiring, and critical infrastructure. The stakes for national security are also higher, since advanced AI and the chips that power it are central to military capabilities and cyber operations. 

The most important thing is that Congress does not sit idly by while this technology transforms the world. Both parties should resist the temptation to turn AI into another culture-war football that is used for partisan battles. The history of the internet shows that politics can lag technology by a decade or more. In the AI era, Washington is engaged much earlier. That is an opportunity, but only if leaders treat AI as a serious national project, not as a prop for the next press conference. 

The dot-com bubble left behind real damage, but it also left behind the infrastructure of today’s online economy. If Washington can learn the right lessons, today’s AI boom can leave behind something even more valuable: a stronger innovation base, a safer and more open digital environment, and a model for how to handle the next wave of transformative technology with a steadier hand.