What Is an AI Bubble? Why Investors Are Worried

The AI bubble shares similar characteristics with almost all economic bubbles in history: asset prices climb rapidly amid market euphoria, diverging significantly from their intrinsic value and fundamental support, and ultimately leading to a rapid price contraction and a burst bubble. The global AI craze currently sweeping the world is increasingly exhibiting these classic bubble characteristics.

For a simple example, during the dot-com bubble more than 20 years ago, any company with ".com" in its name could see its stock price double instantly, even if it had only a rudimentary website and virtually zero earnings. A massive influx of investors, driven by the belief that "the internet will change everything," rushed into the market. No one cared about revenue or profits; they only feared missing out on the next big wave, bidding many shell companies up to astronomical valuations.

Although the internet ultimately reshaped the world, it still went through a severe bubble burst: over a span of two and a half years, the Nasdaq Composite plunged from 5,048 to around 1,100, a cumulative decline of approximately 78% from its peak.

While the long-term value of AI is likewise unquestionable, we must also remain vigilant against a bubble burst that could occur at any moment.

The key to determining whether a bubble exists is not rising stock prices or crowded capital, but rather whether asset prices have completely decoupled from actual earnings and fundamental support.

CICC's latest research report highlights current structural risks: market capital is overly concentrated in hardware, there is insufficient synergistic development between models and applications, and crowded trades in a single segment could exacerbate bubble risks instead.

According to the firm, the core risk of AI is not a bubble in demand itself, but rather structural concerns surrounding the model-side business model—the Scaling Law continues to drive up R&D investment, while intense, homogenized competition compresses the window for companies to build differentiation barriers. Additionally, under a subscription-dominated business model, the industry faces ongoing headwinds from low-price competition.

Meanwhile, capital expenditures of leading tech firms are expanding rapidly. This year, the combined AI-related capex of the four major tech giants, including Meta (META) and Microsoft (MSFT), is projected to exceed $500 billion, with their combined Q1 capex surging 70% year-over-year. As some companies rely on external financing to supplement their capital, pressures from depreciation, financing costs, and free cash flow are expected to gradually emerge.

Current market perspectives on the bubble are primarily split into optimistic and pessimistic scenarios.

Under the optimistic scenario, the current AI boom is seen as far from a bubble. Unlike the dot-com bubble of 2000, which was characterized by widespread conceptual hype, the core participants in the current AI cycle are highly profitable industry giants with robust cash flows.

Commercialization has already been clearly validated. As the most transparent monetization pathway, cloud computing saw the three major cloud providers post a 39% year-over-year revenue increase in Q1, while backlogs surged 145%. AI has also delivered quantifiable gains in areas like ad placement and content recommendation, driving double-digit growth in advertising revenue for leading firms. Coupled with the continuous iteration of technology such as Agentic AI, there is still immense room for the industry's long-term value proposition.

On the pessimistic side, critics argue that current tech stock valuations have reached historical extremes, with cyclically adjusted price-to-earnings (CAPE) ratios trailing only the peaks of the 2000 dot-com bubble and the pandemic-era earnings trough.

Even more concerning is the "circular revenue" issue within the AI supply chain: a significant portion of AI revenue is generated through intra-industry transactions, while actual external demand from end-users remains low. Coupled with highly leveraged debt in data center infrastructure, any macro interest rate hikes or tightening credit conditions will put highly indebted companies under immediate pressure, potentially triggering a valuation correction.

In conclusion, the AI industry is currently in a phase where real demand and speculative sentiment coexist, and a systemic bubble has not yet formed. However, risks such as localized valuation overheating, crowded trades, and debt accumulation warrant caution. While the long-term value proposition of the technology remains valid, this does not rule out the potential for near-term valuation corrections.

The current wave of AI investment continues to push up technology stock valuations, while successive warnings from several legendary investors are constantly amplifying market anxiety.

Seth Klarman, known as the 'Boston Buffett,' stated that current technology market valuations are 'extremely stretched,' and investors have made overly optimistic assumptions about the long-term prospects of AI. His Baupost Group has completely avoided leading large language model companies such as OpenAI and Anthropic, believing that these firms continue to burn massive amounts of cash and will face existential crises if they fall behind in technological iteration. Instead, he has focused on assets whose businesses are insulated from the impact of AI.

Ray Dalio, founder of Bridgewater Associates, also issued a warning, pointing out that historical technological revolutions have always been accompanied by bubble cycles. Currently, AI companies are generally caught in a dilemma of 'burning cash to grab market share or remaining conservative and losing the market.' He emphasized that if hundreds of billions of dollars in capital expenditures ultimately fail to translate into tangible profits, this sentiment-driven valuation carnival will eventually face an inevitable correction.

At this current juncture, the main risks that could trigger a sharp correction in U.S. equities are: the retreat of highly leveraged capital (deteriorated trading sentiment caused by escalating geopolitical conflicts), corporate earnings falling short of expectations (U.S. stocks will enter the second-quarter earnings season in July), and rising expectations of Fed interest rate hikes (which suppress the future cash flows of tech stocks while adding pressure to borrowing costs). Among these, the strain on leveraged financing is the warning sign most worthy of vigilance right now.

A Morgan Stanley research report regards elevated equity financing costs as the market's 'canary in the coal mine.' Currently, the AXW futures financing spread has risen to a multi-year high, and the cost for hedge funds to borrow money to trade stocks has surged. Coupled with the supply shortage of funding caused by banking regulatory constraints, the room for leveraged investors to increase positions has approached its limit. Currently, the scale of equity leveraged financing has reached a record high, and dependence on stock financing has soared within a year. The market's rally is heavily dependent on borrowed funds, with leverage highly concentrated in the semiconductor and information technology sectors, resulting in extreme market divergence.

The institution warned that once leveraged capital is unable to continue entering the market, passive unwinding of positions will begin, and deleveraging will continuously amplify selling pressure, triggering a deep correction in major indices. Meanwhile, geopolitical conflicts have driven Treasury yields higher and strengthened the dollar, quietly tightening financial conditions. The stock market's prior gains have masked the risks of contracting liquidity.

Should the market decline, investors will reprice monetary policy expectations, further exacerbating the correction pressure on U.S. equities.

This content was translated using AI and reviewed for clarity. It is for informational purposes only.