AI is starting to pay the bills: why the dot-com bubble story won't repeat itself

Microsoft, Google, and Meta have cheerfully reported revenue and profit growth, not failing to emphasize the contribution of AI technologies to their successes. Individual AI startups also continue to energetically attract investment.  So what is the AI race: a financial bubble or a new economy? It's an unprecedented question that worries both private investors and CEOs of major tech companies that have invested billions in AI. Let's try to figure out what's going on.

The classic bubble is the dot-com crash that happened in the early 2000s, at which time investors lost trillions of dollars. I wrote about it in detail last month;

And it's not even that the Internet was a failed technology, quite the contrary - in many aspects it has really changed our lives and allowed the growth of such huge companies as Microsoft, Alphabet (Google's parent company), Amazon, Meta (formerly Facebook) - together they now account for 18.56% of the capitalization of the S&P 500 index.

It is just that most of the Internet companies that went public before March 2000 did not have a sustainable business model and were grossly overvalued on the wave of excitement, which eventually led to panic and collapse. The fears are the same now: the share price is rising so rapidly that one wonders how much this corresponds  to the fundamental value of the company (profits, assets, etc.). Obviously, the current AI boom is nothing short of alarming.

"Generative AI is great at spending money but incapable of generating revenue," excited John Naughton, a professor of public understanding of technology at the Open University, UK, last year.

He answers the question "are we in an AI bubble" in his column with the same question: "Is the Pope a Catholic?". So, in other words, yes, of course we are.

But is that really the case?

The current reporting season seems to refute these fears. Meta, for example, reported record revenue of $47.5 billion (up 22% from the same period last year) and net income of $18.3 billion (up 36%) for the second quarter of 2025. "This quarter's strong results were largely due to AI, which has significantly improved the efficiency and profitability of our advertising system," noted CEO Mark Zuckerberg.

Alphabet also reported revenue growth of 14% to $96.4 billion in the second quarter of 2025. "We are leading the way at the forefront of AI and growing incredibly fast. AI is positively impacting all aspects of the business, providing powerful momentum. Search delivered double-digit revenue growth, and our new AI features - such as AI Overviews and AI Mode - are performing well."- commented on the numbers CEO Sundar Pichai.

How is Microsoft doing? "So far, shareholders have praised [Microsoft CEO Satya] Nadella's performance... But if he fails, [implementing artificial intelligence - Oninvest note] his job could be one of the first to be jeopardized by AI,"  wrote Bloomberg in May of this year. However, he doesn't seem to be in danger of losing his job just yet; the company is doing good: revenue in the fourth quarter of fiscal 2025, ending June 30, rose 18% to $76.4 billion. The next day Microsoft joined Nvidia in the elite club of companies with a capitalization of more than $4 trillion. 

Nadella didn't fail to emphasize at the conference call that Copilot - the AI chatbot the company offers to private users, businesses and developers - has more than 100 million monthly active users. About 800 million customers interact with AI features built into Microsoft products, he said, writes Bloomberg.

Of course, there is a bit of trickery here - the companies do not give a detailed breakdown of exactly how much they spent on AI creation and implementation and how much they earned thanks to it. However, these are all large technology corporations with established revenue-generating business models, which gives them both a solid foundation for investing in AI from their own funds and various options for monetizing the new technology;

For example, when the "DeepSeek moment" happened in February,   temporarily dropping $1 trillion in U.S. tech companies, what did Microsoft CEO Satya Nadella do while everyone was panicking? He decided to host this Chinese AI (good thing it's open source) in his cloud service and sell access to it along with the other 1,900 AI models already there. Conventionally speaking, you can spend years trying to grow the most beautiful flower, or you can open a flower shop and make money off all the flower growers - which is exactly what Nadella did, Bloomberg notes. "Whether using the model costs a customer $10 through OpenAI or 90 cents through DeepSeek, Microsoft gets paid for cloud computing, cyberthreat protection, data storage and other value-added services," the agency writes.

In addition, Microsoft has another strong competitive advantage: virtually all of corporate America runs on its office programs such as Word, Excel, Teams and others. Nadella decided to develop an AI assistant, Copilot, and began actively marketing it through established corporate sales channels. Unlike the startup OpenAI, which could afford to release ChatGPT with hallucinations and other flaws into the world, Microsoft understands that corporate consumers need reliability, legal clarity, and security. In the end, it became a competitive advantage. "We were very pleased to reach out to customers and say, 'Do you like ChatGPT?' This [Copilot] is the same but better: more powerful user interface, better security,'" Bloomberg quoted Jeff Teper, head of Microsoft Teams, as saying.  In January 2025. Nadella said that annual revenue from AI products has already reached $13 billion (acknowledging that some of that was due to OpenAI). It's true that the company will spend $30 billion next quarter to create new data centers (50% more than a year earlier), but you have to take into account that not only AI models work there, but also other cloud services, which for the year ended July 30 brought Microsoft a total of  $75 billion in revenue. Investors greeted the report positively, with the company's shares rising 8.2% on July 31 and its capitalization exceeding $4 trillion for the first time (albeit briefly), Bloomberg writes.

Techno-giants, as we can see, are generally doing quite well so far. And what about AI startups? This is also an interesting situation, not unlike the dot-com bubble of 2000.

Take for example  the pioneer of the AI revolution, the company OpenAI. Leaving aside such indicators as the growth of the number of users (these metrics  strongly failed investors during the dot-com boom) let's look at the money.  A Bloomberg source reported that OpenAI's revenue from paid subscriptions to its tools in 2025 may exceed $12.7 billion - more than three times more than $4 billion a year earlier. And in 2026, according to the agency's interlocutor, it will double and reach $29.4 billion. And this is quite impressive for a company that nobody had even heard of three years ago.                                                                                                                                                                                                                                                                                                                                            It is still a long way from profitability, though. OpenAI is expected to "burn" about $8 billion in investor money in 2025, which will be used to lease computing power, train increasingly powerful AI models, and other expenses. According to Bloomberg, the company expects to become profitable by 2029, when its revenue will exceed $125 billion. Ambitious, but people believe - so far the company has no shortage of investor support. In early August, it was reported that OpenAI had raised $8.3 billion in an effort to raise another giant $40 billion round, with no shortage of takers, writes CNBC.

Or there's the startup Anthropic, founded in 2021 by OpenAI alumni Dario Amodei and his sister Daniela Amodei. According to The Information's source, its revenue is approaching $5 billion a year. The story is similar - so far the company is without a profit and "burning through" about $3 billion a year, but investors are already lining up to put $5 billion into it at a total valuation of $170 billion.

Or, finally, the story of Thinking Machines Lab, founded by former OpenAI CTO Mira Murati - I wrote about it recently. The company is less than a year old, has no product on the market yet, but investors have already given it $2 billion at a total valuation of $12 billion. Isn't that a bubble?

In my opinion, the current situation has an important difference from the dot-com bubble. Back then, mostly retail investors invested in Internet startups through IPOs against the backdrop of loose monetary policy (MP). New offerings occurred almost daily or even more frequently - 457 IPOs in 1999, and then another 91 in the first quarter of 2000 alone.

Things are very different now. The tech IPO market, which froze in 2022 amid the tightening of the DCA, is barely alive: since January 2025, there have been an average of two per month, and by early July the number had jumped to five, wrote CNBC.

Either techno-giants invest in new AI technologies from their own funds or professional venture capital investment funds. It is not that top managers of large companies and qualified venture capitalists cannot make mistakes - they are also human beings and are subject to the so-called FOMO effect (fear of missing something important).

However, as we can see, AI technologies have already demonstrated, if not profitability, then the ability to generate substantial cash flows, and they can grow them very quickly, in fact, faster than anyone has been able to before. According to a study by fintech group Stripe, AI startups with annual revenues of more than $30 million have reached this milestone in an average of 20 months - five times faster than, for example, SaaS companies (which have also been a recent hit), writes the Financial Times.

"When the dot-com bubble burst in 2000, the correction was swift and brutal. Companies valued at billions of dollars evaporated overnight, leading many to conclude that Internet technology was just a speculative fad," writes portal Geeky Gadgets. The pain was long-lasting. It took the NASDAQ index 15 years to regain its early 2000 highs;

"Today's AI boom exhibits remarkably similar characteristics, but here's the key difference. It is based on technologies that are already embedded in everyday life and scaling rapidly across industries. While investment enthusiasm has inflated some valuations, key capabilities such as natural language processing, generative tools and autonomous agents are not speculative ideas. Even if the market cools, AI will not disappear or retreat, as has happened to many dotcom startups. Its driving force is structural, not HYPE," summarizes Geeky Gadgets.

This article was AI-translated and verified by a human editor