Originally published as an article on November 9, 2025
BLUF
Some say we are in an AI bubble because the hype valuations outpace actual revenue and because the AI market is highly concentrated. However, government investment and institutional support, and existing applications of AI in the real economy may argue differently. Whether or not AI is a boom or bust will depend on the speed of AI innovation, adoption, and market corrections.
On November 7th, 2025, the Financial Times reported that tech stocks suffered their worst week in sell-offs since market reactions to Trump’s “Liberation Day” tariffs in April 2025. How bad? The value of the eight most highly valued stocks in AI plummeted by $800 billion and led to a 3% drop in the NASDAQ. This massive sell-off came on the back of reports that we could be in an AI bubble. OpenAI’s CEO, Sam Altman, told reporters back in August of this year, “Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes. Is AI the most important thing to happen in a very long time? My opinion is also yes.” Even Michael Burry, the guy who predicted the 2008 financial crisis, bet $1 billion on the collapse of Nvidia and Palantir. So what’s all the hubbub about AI, and is this “bubble” set to burst?
What’s all this talk about a bubble?
In economics, a bubble simply means a situation in which speculation, hype, and herd mentality cause an asset’s price to outpace its intrinsic value. Stock prices are not driven by how much an asset is actually worth, but rather by how much people think it’s worth. The fervor feels great at first. Yes! I got in before this stock took off! Oh my gosh, I’ve got to get in before I miss the boat! Yet, in the case of industry bubbles, what goes up must come down. There is a juncture in which investors realize that the fundamental value of a conglomerate of companies is actually not worth as much as the frenzy would otherwise dictate. It starts small; a few investors quietly sell off their shares, but it can quickly lead to a panicked mass sell-off that can tank the industry.
A classic case of this lifecycle is the dot-com bubble. In the late 1990s, the explosion of internet-based companies spurred huge and nearsighted optimism about the future of the internet’s application in the economy. Yet, a mixture of investment capital stalling and the inability of the “dot-com” companies to return profits to meet the speculated valuations led to a massive sell-off and the NASDAQ spiraling 77% down from its peak. Other cases, including the 2008 financial crisis, follow this same model:
Displacement moment: a new industry, company(ies), or technology disrupts the market, promising to revolutionize the economy.
Overvaluation period: the stocks for these companies boom as new investors frantically try to enter the market, artificially pumping up the asset value.
The “uh-oh” moment: some investors begin to see the writing on the wall and sell their assets.
The “don’t be the sucker” moment: panic ensues. People sell off their shares rapidly to cut their losses and avoid being the one who ends up with the stock after the crash.
It’s no secret that AI is a booming industry. But is it an overvalued industry? And if so, what stage in the cycle are we in?
Why we might be in an AI bubble
Huge investments and high valuations outpace earnings
Investment in AI is sky high, and rising. Private investment in AI has shot up from $15 billion in 2015 to its $145 billion peak in 2021. In 2024, private investment was $130 billion. Furthermore, more than 1300 AI start-ups exceed a $100 million valuation, while 498 AI companies are valued at or above $1 billion.
Yet, the outlook on actual profits may tell a different story. Take OpenAI, for example, which is one of the biggest players in the industry. The company has made about $1 trillion in AI investment deals but is only set to return around $13 billion dollars in revenue. This story is just one of many stories of investment and valuation outpacing earnings reports.
Many try, few triumph
As a former Stanford student, I can tell you firsthand…the stereotype is true. Stanford kids LOVE their startups. And in Silicon Valley, it’s hard to walk from one end of campus to the other without someone talking about how their startup idea is to do this thing that is already being done, but “hear me out…add AI.” Whether it be the promise of the elusive unicorn status, the Silicon Valley hype, or an attractive post-grad option in a stalling job market (a topic for another discussion), the market is flooded with AI startups and players from all different rounds of seed funding.
However, if you actually look at the big players in AI—Nvidia, Microsoft, and Apple—you’ll see that these three companies alone account for 41% of the S&P 500 Technology index. This market concentration is eerily similar to that of the dot-com bubble, where the top seven NASDAQ stocks made up a significant portion of the market.
Why we might NOT be in an AI bubble
Governmental institutions are poised to support AI development
Recent government initiatives, such as President Trump’s AI Action Plan, which emphasized accelerating AI innovation and adoption through streamlined regulation processes, supporting AI infrastructure projects, and encouraging private-public AI partnerships, have boomed. Furthermore, the Chief Digital and Artificial Intelligence Office (CDAO) recently announced a $200 million contract each with Anthropic, Google, OpenAI, and xAI to develop agentic workflows that will assist with mission-critical areas within the Pentagon.
This broad institutional backing signals that there will at least be one major stable partner in the AI market, the U.S. government itself. Like the airline industry post 9/11, banking industry in 2008, and automotive industry in 2009, the AI market could be classified as “too big to fail,” in which the U.S. government would bail out or intervene in certain industries that are so important to the U.S. economy that their failure would cause broad economic harm.
Evidence of long-term value
The dot-com bubble burst because internet-based companies were unable to find sufficiently broad or profitable applications for their services. There is reason to believe that AI is already different from this. AI is already demonstrating broad applications in cloud computing, semiconductors, enterprise software, and other key sectors. Because of these broad-based offerings, AI promises to add concrete value to a broad range of industries, which can help it build resilience to failures in individual sectors.
Boom or bust? What’s the verdict?
In short, whether or not the “AI bubble” will burst depends on a couple of factors:
Evolution or involution: AI advocates say that models will only get better at an increasing rate. A recent example of this is the release of AlphaEvolve earlier this year, which uses a recursive self-improvement (RSI) process to iteratively learn and improve by itself. On the flip side, the AI cynics cite the release of ChatGPT 5 as a prime example of the AI plateau, where the GPT 5 model did not demonstrate the exponential improvement that some thought, and even fell short in other areas, such as accuracy and speed. In this scenario, the AI market would more so be defined by intense competition among an increasing number of players trying to remake the same wheel to fit a finite range of narrow applications.
Adoption, adoption, adoption: AI will only be as good as how it is applied to the real economy. What sectors can AI be used to optimize data analysis? What labor-intensive industries can embodied AI be substituted for humans to increase production efficiency and mitigate high-risk jobs? How can recursive self-improvement and agentic AI be used to discover new drugs, test their safety, and optimize the roll-out of those drugs to communities that are giving pharmaceutical companies demand signals for cures to certain diseases? And notwithstanding how much of this will actually happen, will we even let it? What are the socioeconomic barriers to AI adoption? Amazon is already planning to replace half a million jobs with robots from Kiva Systems. Where will these workers go, and if this happens on a large scale, will we really be prepared to address the major distortions in the labor market?
Bear walk or bear run: Bubble bursts tend to happen because of phase 4, the “don’t be the sucker” moment, or in other words, the panic that ensues when people intractably sell their shares, sometimes known as a bear run on the stock. This would be bad. However, if the AI market can self-regulate by distinguishing the “AI hype” companies from the here-to-stay companies, which have diversified revenue streams beyond AI and have the earnings reports to prove value beyond seed funding, there might not be the sudden crash we saw in 2000 or 2008.
So, the bottom line: Some say we are in an AI bubble because investments, valuation, and “hype” are outpacing fundamental markers of company value. Furthermore, the high volume of competitors with a significant market share concentrated among three companies means that the entire market hinges almost entirely on the performance of the big players. However, government investment and institutional support signal that the current administration will not allow this bubble to burst quite yet. And on top of that, AI is already proving to have broad applications in the real economy. Finally, whether or not AI is a boom or bust will depend on whether AI innovation can keep pace with the hype, whether AI can be sufficiently adopted into the real economy, and whether market corrections adjust slowly enough to allay fears of a mass sell-off.
Discussion about this episode
Bottom Line Up Front
Weekly updates in international security in a nutshell.
Weekly updates in international security in a nutshell.
