Dot-com and Artificial Intelligence (Bubble)

Economic Bubble

An economic bubble is when investors speculate forward, buying an asset based on the expectation that it can be sold at an even higher price in the future rather than its present value, causing the asset’s price to rise far above its intrinsic value to levels that cannot be explained by logic and economics. (Item 1)

Bubbles generally grow with the belief that the growth in a sector can continue indefinitely. For this reason, momentary statistics such as intrinsic value and P/E ratio lose their importance for investors and reach absurd values (Item 2) (it is important to keep this part in mind.)

So how do bubbles burst? There are 2 main reasons for bubbles to burst. The first is that companies fail to meet expectations and cannot generate the expected earnings on their balance sheets, and the second is the increase in interest rates or the deterioration of the general economy. As a result of bubbles bursting, major pullbacks occur in the prices of assets in the affected sector, and prices reach “reasonable” levels again [18]. (Item 3)

Dot-com Bubble

I want to start my article by first explaining the Dot-com bubble. The internet sector, which began to popularize in the 1990s, grows significantly because investors and the media think it will develop much more in the future, and it receives excessive investment from the outside with high income expectations. In addition, with the contribution of economic growths such as low inflation, rapid GDP growth, and low interest rates in the United States, the number of “.com” ventures increases [17]. (Item 1)

Since it is thought that all ventures and companies should have a website, stock prices soar because investors think that companies with websites will be much bigger in the future, whether they make a profit or not. Furthermore, starting from 1998, initial public offerings of internet-based companies increase, and between 1995 and 2000, the Nasdaq index rises from 1,000 points to 5,000 points. (Item 2)

Then in 2000, with capital slowly beginning to disappear, expected high revenues not materializing, and the financial structures of companies slowly beginning to deteriorate, a major correction begins in the markets. In the years before the bubble formed, record-low interest rates, the spread of the internet, and the increased interest in technology companies allowed capital to flow abundantly, especially into newly established companies with no track record of success. Valuations rose rapidly, and eventually, the cash flow stopped. Between March 10, 2000, and October 4, 2002, the Nasdaq dropped from 5,048 points to 1,139.90 points, a decrease of 76.81%. The stock prices of internet giants like Amazon, Cisco, Intel, and Oracle faced a drop of more than 80% [19]. (Item 3)

But let’s talk about the present by looking at this situation from a different perspective. Amazon’s stock price dropped from $3 to $0.35 with the burst of the bubble; today it is $245.05. Oracle’s stock price dropped from $46 to $9; today it is $245. In short, on the stock market, companies that are truly innovative and fill a gap that needs to exist in the future will reach the place they deserve in the long run. Here, I would like to remind you of Cullen Roche’s quote: The stock market is the only market where all the customers run out of the store when everything goes on sale.

Artificial Intelligence (Bubble)

Let’s come to the present. It has been 3 years since the emergence of ChatGPT. If you are reading this article, you probably also know that there is a world of difference between ChatGPT’s initial capacity and its current capacity. At its simplest, the Artificial Intelligence that initially could only share things that happened up to a certain time solely through text, can today exponentially generate videos from just a sentence as if they were shot with a camera. Well, this rapid rise of Artificial Intelligence didn’t just stay within itself; it also leaped to the stock market. In this article, I will address both the similarities of Artificial Intelligence with the dot-com bubble and the fundamental differences that separate it from it.

So why did this “AI Race” flare up? There are basically two reasons why companies adopt AI: The first is that there is a market for AI, and when used correctly, it increases company efficiency. The second is FOMO, that is, Fear Of Missing Out. I will make this point more concrete with two examples. In every era, there are themes that companies must adopt. For example, Google, which learned a great lesson from the decline in Yahoo’s market value in the past, currently possesses the best AI model on the fly. As a second example, in an interview regarding the topic of AI, Mark Zuckerberg says: “AI might be a bubble, but ‘misspending a couple hundred billion dollars’ is worth it to reach superintelligence. It’s highly probable that AI is a bubble, and if we end up misspending a couple hundred billion dollars, it will of course be very unfortunate. But I can say this, I think the real risk is on the other side. At least for a company like Meta, the risk lies in not being aggressive enough rather than being a little too aggressive. [0]”. There is no doubt that AI will be with us in the future as well, but when it comes to the stock market, according to research, 54% of investors state that we are in an AI bubble, and 60% state that stocks are inflated [14].

Famous investor Michael Burry, who noticed the situation before the 2008 American housing price bubble burst and invested against the housing market, which was thought would never burst at the time, announced in recent months that he also took positions against Palantir and Nvidia, companies that experienced massive increases in their stocks by utilizing AI [15] [20].

Furthermore, Sam Altman, one of the founders of OpenAI, the owner of ChatGPT, said in an interview on August 18, 2025, that the AI industry is heading towards a bubble due to expenses.

The Shiller P/E Ratio, which tracks the Price/Earnings ratio of the S&P500 index containing the top 500 companies in the United States, is at the 40.46 level. (The Shiller P/E ratio, compared to normal P/E ratios, compares stock prices to inflation-adjusted earnings over the past 10 years). You can see the historical background of the Shiller P/E Ratio below:

Shiller P/E Ratio
Shiller P/E Ratio [16]
So what is this Price/Earnings ratio? This ratio is a fundamental financial metric that proportions the stock price of the tracked value to its earnings per share (EPS) and indicates whether the current market price of this stock is high or low relative to its earnings. For example, if one lot of Company A’s stock trades at 58₺ and its earnings per share over the last 12 months is 3₺, the P/E ratio of this stock is calculated as 58/3=19.3, and the stock is priced at 19.3 years of the company’s profit. This means that for the current price of the stock to make sense, it must maintain its annual profit for another 19.3 years. [1]

When looking at the Shiller P/E Ratio, you can see that the previous peak formed during the Dot-com bubble and was at levels of 44. So let’s come to the AI giants and their P/E ratios:

The P/E ratio of the company Nvidia—which in the past generally produced graphics cards for gamers and caused its stock prices to increase due to the increasing importance of graphics cards with the birth of AI—is currently 45.69. That is, it needs to maintain its current profit for 45.69 years.

The P/E ratio of the company Palantir—which develops software platforms in the fields of big data analytics and AI; specifically offering systems that help governments (defense, intelligence) and large companies make sense of complex datasets and make decisions—is currently an astonishing 743.86! Investors who look at the current price of Palantir shares and buy them think that the company can maintain its annual profit level for 743 years (+11 months)…

Finally, I would like to give a general overview of the sector. The P/E ratio of the “Global X Robotics & Artificial Intelligence” fund, which invests in AI companies in general, is 37.12, and the P/E ratio of the S&P 500 Information Technology Index is 40.22.

It is obvious how absurd these ratios are without the need for deeper research. But when this is so clear, how can the stock prices of these companies continue to rise? Because companies are still positively exceeding profit expectations. For example, according to the balance sheet of the company Nvidia announced last month, the company’s annual revenue was announced as $57 billion, exceeding expectations of $54.9 billion. But are these returns really sustainable? This is where things start to heat up.

First of all, the energy need of AI is increasing day by day. Therefore, the only logical solution seems to be nuclear energy. So much so that there are even nuclear energy companies where OpenAI founder Sam Altman is a partner. Because of this need, along with AI, the prices of nuclear energy companies, especially Small Modular Reactors, have increased considerably. The amount of energy needed for OpenAI alone is equivalent to 26 nuclear power plants. Looking at AI in general, by 2030, $7 trillion is expected to be spent just on energy needs [2]. In comparison, the US capital expenditure in 2024 was $8 trillion [3].

But a more critical point than the energy need is circular investments. In the chart below, you can see the investments AI companies make in each other [4]:

The most important point here is that after every deal is announced, the stock prices of the two companies making the deal also increase. Even after the chart above was shared, deals continued to be made, and stock prices continued to rise as if nothing happened. And the company right in the middle of this circular investment, making the most deals, is Nvidia, which I mentioned above. When the chart is examined, it is seen that Nvidia invests in every company, large or small, and in return receives investments, albeit indirectly. I want to make this situation clearer with a small example. Imagine a pickaxe seller. Beyond selling pickaxes as should normally be the case, this seller pays gold miners to buy pickaxes from their shop. The exact same situation, as can be seen above, exists among AI companies.

This is the biggest argument of those who think AI is a bubble. During the Dot-com bubble, such circular investments were particularly famous among companies like Nortel. (Nortel stocks had reached the $87 level with the Dot-com bubble, and practically dropped to 0 with the burst of the bubble [5].) So what will be the end of so much investment? Will this bubble grow forever? Of course not. The investments made bring along revenue targets as well, and as I said before, the end of bubbles is brought about by failing to meet expectations. Studies predict that for AI companies to meet expectations, their annual revenues need to be around $2 trillion by 2030 [6]. For comparison, this amount is MORE than the TOTAL 2024 revenues of Microsoft, Meta, Google, Amazon, Apple, and Nvidia. In contrast, according to research by Citi Group, AI revenue in 2025 is projected to be only $43 billion [7].

While major technology companies are aggressively making CAPEX expenditures (CAPEX (Capital Expenditure) is major investments a company makes to purchase, upgrade, or maintain physical assets to increase long-term growth and operational capacity), the information that Oracle’s gross profit margin in new AI data centers is only around 14% shows how difficult it might be to get a return on these investments [23]. According to calculations, for AI companies to recover the capital they invested (in the base scenario), they are expected to generate approximately $250 billion in AI revenue annually. [23]

Let’s look at possible scenarios. First of all, building the energy infrastructure required for AI to achieve these returns is a long process in itself. Economics is the social science that studies how people will meet their unlimited needs due to the scarcity (insufficiency) of existing resources; scarcity is the insufficiency of factors of production in the face of unlimited human wants. In my personal opinion, in the AI sector, there is a dilemma between energy scarcity and the time required to resolve this scarcity. Although nuclear power plants might seem like a quick solution, for instance, in the US, the acceptance of Nuclear Reactor applications alone can take 5 years [8].

Another important point is OpenAI. Although there are alternatives to Nvidia, the largest circle in the chart above, the fact that every company makes deals with OpenAI, which is relatively not that big, shows that the AI ecosystem is dependent on a small company (OpenAI is a private company, and according to recent investments, the company’s value is $500 billion; in comparison, Nvidia’s value is $4.5 trillion). If OpenAI, which has made deals with almost all AI companies, cannot sustain the deals, other companies in the AI ecosystem may have to revise their plans. OpenAI, in this position, at least looking from the outside, seems to rely on the support that could come from the US government. Of course, the company that would profit in such a possible scenario is Microsoft. Holding approximately 25% of OpenAI shares, Microsoft might try to acquire OpenAI completely in a potential situation, but this would only be positive for Microsoft.

As you may have noticed, I haven’t mentioned a country other than the US throughout my article, but our third scenario is about China, the rival of the US. Actually, thinking about it now, my not mentioning China throughout my article shows that the probability of this scenario occurring should not be ignored. Almost all publicly traded AI companies are in the United States, and no one knows what China is doing regarding AI. China is like a black box, but despite this, the US is trying to restrict chip shipments to China as much as possible to win the AI war. The last thing that came out of China and turned US stock markets upside down in a day was DeepSeek. Until that day, because there appeared to be no competition between countries, not much attention was paid to optimization on the US side, and therefore on the day DeepSeek was announced, with the entry of a model from a rival country into the market, Nvidia dropped by 17% and the S&P 500 Technology Index dropped by 5.6% on the same day [11]. In addition, chips such as the Ascend chips produced by Huawei are being produced in China to reduce dependence on the US [12]. China’s other trump card is open source. Compared to US-origin AIs like ChatGPT and Gemini, China is aware that it needs to follow a different route to capture the market, and therefore, for example, DeepSeek was released as open-source. People who wanted could “download” DeepSeek to their computers and use it offline [13]. [21]

Looking at the problems of AI independent of all external factors, the biggest problem is definitely the data limit. Leading figures in the AI world state that the limits of scaling laws, which have been the strategy of “larger model, more data, more processing power” ongoing for the last 5 years, are being approached. OpenAI co-founder Ilya Sutskever said “data is the fossil fuel of AI” and stated that peak data on the internet has been reached, and simply scaling up will no longer be enough [24].

Although AI is similar to the Dot-com bubble in many ways, there are also ways in which it is not. For example, unlike the Dot-com bubble, as I just mentioned, there are also financials and earnings supporting most of the valuations AI companies possess. Looking at the overall market, Morgan Stanley stated that the cash flow generated by S&P 500 companies compared to their current valuations is three times higher than the period before the dot-com bubble burst in 2000 [9]. Furthermore, compared to 2000 when 36% of technology companies were making losses, today only 19% of technology companies are making losses [10]. AI giants like Nvidia, Microsoft, and Google are recording massive year-over-year revenue increases between 30% and 80% [22].

During the Dot-com era, many companies like perts.com and Webvan had billion-dollar valuations with almost no revenue. Looking at today, companies like OpenAI and XAI, contrary to this, generate revenues in the billions of dollars [22]. At the same time, AI is closely related to the real economy due to requiring infrastructures such as data centers, energy, etc., and creates value in the economy not only with its products but also through the process of building these products. Although the internet was needed as infrastructure in the Dot-com bubble, the difference between market expectations and application in the real economy was much higher [22]. Companies like Google, Microsoft, and Nvidia also have a debt/equity ratio of around 5% to 30%, and these companies have large cash positions in their coffers.

Finally, against the currently irrational-looking AI market, I would like to remind you of John Maynard Keynes’s quote: “Markets can remain irrational longer than you can remain solvent.”

———-

Sources:

[0] https://sherwood.news/tech/meta-ceo-mark-zuckerberg-ai-might-be-a-bubble-but-misspending-a-couple-of/

[1] Shiller, Robert (2005). Irrational Exuberance (2nd edition). Princeton University Press

[2] https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-cost-of-compute-a-7-trillion-dollar-race-to-scale-data-centers

[3] https://fred.stlouisfed.org/series/BOGZ1FA895050005Q

[4] https://flowingdata.com/2025/10/13/circular-deals-among-ai-companies/

[5] https://x.com/NewLowObserver/status/1361770009303814146

[6] https://www.bain.com/about/media-center/press-releases/20252/$2-trillion-in-new-revenue-needed-to-fund-ais-scaling-trend—bain–companys-6th-annual-global-technology-report/

[7] https://www.citigroup.com/global/insights/ai-the-information-era-apex-technology

[8] https://www.eia.gov/energyexplained/nuclear/us-nuclear-industry.php

[9] https://www.morningbrew.com/issues/on-the-bubble

[10] https://www.janushenderson.com/en-dk/investor/article/ai-versus-the-dotcom-bubble-8-reasons-the-ai-wave-is-different/

[11] https://www.intuition.com/deepseeks-market-shock-what-you-need-to-know/

[12] https://www.tomshardware.com/tech-industry/semiconductors/huaweis-ascend-ai-chip-ecosystem-scales

[13] https://www.getlago.com/blog/deepseek-open-source

[14] https://www.bloomberg.com/news/articles/2025-10-14/most-investors-say-ai-stocks-are-in-a-bubble-bofa-poll-shows

[15] https://www.bbc.com/news/articles/c867vyn2evlo

[16] https://www.multpl.com/shiller-pe

[17] https://www.investopedia.com/terms/d/dotcom-bubble.asp

[18] Dogic, Nina (2015). Theories of Finance and Financial Crisis – Lessons for the Great Recession

[19] https://web.archive.org/web/20100610011943/http://www.socialmediatr.com/blog/dot-com-krizinin-10-yilinda-internet-nereden-nereye-geldi/

[20] https://www.theguardian.com/business/2025/nov/05/global-stock-markets-fall-sharply-over-ai-bubble-fears

[21] https://www.youtube.com/watch?v=j4CiDOSedmQ

[22] https://www.youtube.com/watch?v=m9PBVlLizzQ

[23] https://borsadabibasina.com/yapay-zeka-yatirimlari-balon-mu-altyapi-devrimi-mi

[24] https://www.youtube.com/watch?v=LYoJnCpRswo

https://www.youtube.com/watch?v=2U6bJEFC97c

https://www.youtube.com/watch?v=tAXKxKTGWFQ

https://en.wikipedia.org/wiki/AI_bubble

https://www.youtube.com/watch?v=fIcWfHikAOo

 


Subscribe to get the latest posts sent to your email.