Over the past month, I've been asked at least four times, "Is there a bubble in the current AI bull market?" I think, of course there is—every major bull market has a bubble, and the three-year AI bull market that began in December 2022 (note: I'm mainly referring to the Nasdaq here; I don't understand the STAR Market and don't want to comment on it) couldn't possibly be without one. However, we must avoid a misguided notion: comparing this localized AI bubble to the dot-com bubble of early 2000.

As is well known, the US stock market experienced an unprecedented super bull market in the late 1990s, until it crashed in March 2000. After a few brief rebounds, it became completely incurable in the latter half of that year, entering a dormant period of more than two years. During this period, many of the first-generation internet companies that created capital myths collapsed, leaving only a handful, such as Amazon, to become new-era tech giants. As for computer software and communication equipment companies, even those that survived often took a long time to recover their original valuations. The most typical example is Microsoft: it was the first company in human history to break the $500 billion market capitalization mark, but after the bubble burst, it didn't recover to the $500 billion mark until 2017.

Various people, driven by various motivations, repeatedly emphasize to us that the artificial intelligence industry is about to repeat the tragedy of the last bubble burst! At that time, the currently hot computing power industry chain, AI large-scale model and application industry chain companies will face a major downturn, just like Yahoo, eBay, and Nortel Networks did. However, I think the above statement is completely lacking in common sense because it ignores two key characteristics of the 2000 Nasdaq dot-com bubble:

1. In the 1990s, the Federal Reserve implemented excessive monetary easing, which catalyzed a capital bubble;

2. At that time, the internet industry had not found a viable business model.

Let's start with the first point. The statement that "Greenspan was the most competent Federal Reserve Chairman in history" is one of the biggest lies in history. During the formation of the first dot-com bubble, Greenspan claimed it was "irrational exuberance" while maintaining low interest rates for an extended period, only belatedly tightening monetary policy when the bubble reached unprecedented levels of severity. To this day, economists generally believe that the Federal Reserve under Greenspan bears undeniable responsibility for that "irrational exuberance."

This time, however, the Federal Reserve under Powell is completely different. Since 2021, Powell has been known for his unconventional approach, implementing aggressive monetary tightening (although he previously made the mistake of raising interest rates too late) . This AI bull market started before the Fed's rate hike cycle had even ended, which is historically rare. The Fed didn't confirm entering a rate-cutting cycle until 2025, and the pace of rate cuts was slow, much to the disappointment of Wall Street. In any case, monetary tightening did not prevent the three major US stock indices from reaching new highs. We can say with certainty that the origin of this AI bubble is definitely not excessively loose monetary policy.

In the late 1990s, dot-com stocks enjoyed a period of prosperity in a low-interest-rate environment, which ended when the Federal Reserve began aggressively raising rates. The situation is reversed now; artificial intelligence companies are enjoying a boom in a high-interest-rate environment, and the Fed's rate cuts are clearly beneficial to them. Even if a bubble bursts, monetary policy can play a mild buffering role, rather than adding fuel to the fire as it did in 2000.

Now, let's discuss the second point. Around 2000, the entire Silicon Valley, and indeed the world, had no idea what the core business model of the internet would be. At that time, the concept of "consumer internet" remained merely on paper. The "consumer internet platforms" that are now so influential either hadn't yet been established or had been established but hadn't yet gone public. Back then, the capital markets were most enthusiastic about several concepts: first, "clicks and mortar" ( referring to the basic internet model); second, portal websites; and third, ISPs (Internet Service Providers) . In hindsight, these were clearly not the right path for the development of internet business models.

（Netscape）

Back then, tech giants generally approached internet business with the mindset of the PC era. A prime example is Microsoft's "first browser war": it genuinely believed that web browsers were the gateway to the internet, just as operating systems were the gateway to the PC era, and therefore was willing to do anything to defeat Netscape , even at the cost of antitrust investigations. Other tech giants thought the same way; for instance, AOL spent heavily to acquire Netscape, attempting to combine browsers and ISP services to create a new generation of "super internet gateways." Incidentally, AOL later merged with Time Warner, attempting to create a dual super gateway of "internet + traditional media," ultimately resulting in the largest corporate merger disaster in history.

The situation is entirely different in this era. Artificial intelligence has already found a viable business model. Let's set aside the computing power industry chain for now and focus solely on applications: hundreds of millions of users have developed the habit of subscribing to certain large-scale model chat services, generating hundreds of billions of dollars in annual revenue for large-scale model vendors (including not only OpenAI but also its competitors) . Enterprise clients, by calling large-scale model APIs, have also contributed similar or even larger amounts of revenue. By 2025, the revenue scale of large-scale model and application vendors will be unimaginable for the first-generation internet companies of 2000!

Please note: When we say that "the business models of companies like OpenAI are not yet fully viable," we don't mean "they don't generate significant revenue," but rather that "their revenue is not yet commensurate with their overall costs (including capital expenditures) ." OpenAI, Anthropic, and Google's Gemini can all generate substantial operating revenue. The capital market's concern is that their revenue growth rate cannot keep up with the growth rate of capital expenditures—and this concern is valid. In the last dot-com bubble, almost no internet unicorns could generate substantial revenue, even the internet businesses of tech giants; for example, Microsoft's browser business was commercially insignificant.

（platforms）

Therefore, after the bursting of the dot-com bubble in 2000, the global technology industry underwent approximately a decade of arduous exploration before finding a business model truly suitable for the internet: on the consumer side, this manifested as so-called " platform companies" that combined massive amounts of user-generated content (UGC) with merchants; on the enterprise side, it was represented by cloud computing and SaaS (Software as a Service) , enterprise information services integrated with the internet. Both ends saw the emergence of a number of companies with trillion-dollar market capitalizations. Unfortunately, this exploration process was far too long. Strictly speaking, the 2000 bubble was largely deflated in about two years; the subsequent long period of dormancy was primarily due to the lack of long-term growth drivers, rather than excessively high valuations.

This time, the situation may be much better. After the inevitable bursting of localized bubbles (perhaps now, or perhaps several quarters later) , the "post-bubble era" will be relatively short, as AI companies have generally found viable business models. Once the capital market realizes that large-scale application companies can consistently deliver solid revenue growth, their concerns about the computing power industry chain and cloud service companies will also dissipate. At that time, we may witness another genuine revaluation of technology companies lasting several years—this time not a bubble, but a long and moderate period of growth like that of 2011-2019.

That will be the long-awaited "deep snowfall" that value investors throughout history have been eagerly anticipating: the transformation of human society by artificial intelligence is tangible and profound, and its impact will inevitably be very long-lasting. If the impact of the internet on human society is 100 points, then artificial intelligence is at least 300 points, perhaps 1000. The integration of artificial intelligence with technologies such as new energy, space exploration, and virtual reality will create an impact as high as 2000 points or even 10000 points. All of this will happen in our lifetime, which is something to look forward to and something to be very excited about!

Of course, this is predicated on the premise that we must acknowledge the existence of localized bubbles. Anyone who denies this is being overly optimistic in the short term. Long-term optimism and short-term optimism often cannot coexist, and this is precisely the case at this moment.