From TSMC to the "FOMO Rampage" sweeping Taiwan: A 160% surge in leverage has already sounded the alarm for the bursting of the AI bubble.

According to analysts on Wall Street, the current AI investment frenzy in the Taiwan stock market is not just a speculative AI investment bubble, but is built on the strong fundamentals of hard technology manufacturing behind the global AI computing power supercycle. TSMC and its extensive AI computing power supply chain collectively produce around 90% of the world's most advanced chip processes, which are crucial for the accelerated expansion of intelligent smartphones, laptops, Siasun Robot & Automation, and particularly global AI data center construction. Meanwhile, Foxconn and its supply chain represent the most crucial assemblers for AI server clusters for American giants such as NVIDIA, Amazon AWS, and Microsoft Azure. TSMC and Foxconn are viewed by investors as the "bellwethers" of the AI computing power industry chain. In other words, Taiwan has shifted from a traditional export-oriented market to a high-beta investment return center for the global AI computing power infrastructure supercycle. Over the past year, the benchmark stock market in Taiwan has surged by over 100%, surpassing the UK, Canada, and India in a matter of weeks to become the fifth largest stock market in the world. This surge reflects the fact that global funds are recognizing the real value of the AI computing power industry chain in the "AI arms race," from the investments by U.S. tech giants, to the AI computing power chain of NVIDIA, Blackwell/Rubin, and the increasing capital expenditures in cloud computing by tech giants, further extending to Asia's most advanced chip processes, advanced packaging systems, and semiconductor/AI server manufacturing clusters. For top analysts on Wall Street, the Taiwanese market is becoming the forefront indicator of whether the capital expenditures of global tech giants in AI data centers will be realized. Should the concerns held by Wall Street about the slowdown in capital expenditures by tech giants, cooling demand for chips/AI servers, or a decrease in demand for advanced processes/advanced packaging arise, the retreat in the Taiwan stock market would be magnified by the concentration of AI weightings, and could even be the first domino to lead to the burst of the global AI investment bubble. Goldman Sachs' latest calculations show that the global AI capital expenditures benchmark model is expected to grow from $765 billion annually in 2026 to $1.6 trillion annually by 2031, with an estimated total capital expenditure of around $7.6 trillion from 2026 to 2031. The power demand of U.S. data centers is expected to increase from 31GW in 2025 to 66GW in 2027. Morgan Stanley forecasts that by 2028, nearly $3 trillion in AI-related infrastructure investments will flow through the global economy, with over 80% of expenditures still ahead. TSMC's faith has fostered the most extreme risk appetite According to the latest statistics compiled by Bloomberg Intelligence, companies in the MSCI Emerging Markets Index reported an "average annual profit" higher than analysts' consensus expectations set a year ago for the first time since April 2022. Asia's tech leaders, including TSMC, Foxconn, SK Hynix, and Samsung Electronics are driving this result. Asia is home to numerous AI computing power infrastructure manufacturing companies, including core chip manufacturers like TSMC, Foxconn, SK Hynix, Samsung, MLCC giant Murata, Taiyo Yuden, Victek, Victory Giant Technology, and others, making Asia's AI computing power infrastructure industry chain the biggest winner in the "AI disrupts everything" trend according to Wall Street analysts. The strongest theme in the current AI investment trend is undoubtedly the AI computing power infrastructure manufacturing/outsourcing stage with "restricted supply + high technical barriers," covering advanced process outsourcing, advanced packaging, HBM/high-end server storage, key AI server components, data center power, liquid cooling, and heat dissipation, as they transfer the unit economics of AI training/inference systems to "computing power and energy consumption per token," with these stages primarily concentrated in Asia. From the perspective of a supercomputing project engineer at an "AI factory," TSMC occupies the most core "physical bottleneck layer" of the AI computing power infrastructure: NVIDIA/AMD AI GPUs, Broadcom/Marvell ASICs and network chips, AMD AI accelerators, and cloud vendors' proprietary ASICs all revolve around advanced manufacturing processes, advanced packaging, and high yield mass production. Information disclosed at a recent shareholder meeting by TSMC's management indicates that demand for advanced processes and advanced packaging will continue to outstrip supply in the coming years - meaning that AI data center construction is not a one-quarter phenomenon, but rather a multi-year capital expenditure cycle from training GPUs to inference, Agentic AI, Siasun Robot & Automation, autonomous driving, and sovereign AI. TSMC itself emphasized in shareholder meeting materials that it will continue to invest in leading processes and packaging in Taiwan, and advance multi-stage capacity construction of 2nm-level advanced processes to meet the robust computing power demand driven by AI. However, the fundamental advantage of AI computing hardware manufacturing dominated by TSMC also brings highly concentrated risks. Technology companies account for about 20% of Taiwan's economic output, but have an approximately 80% weight in the benchmark stock indices, meaning that Taiwan's stock market has essentially become a highly concentrated "AI semiconductor factor asset." As long as global cloud vendors continue to expand AI data centers and demand for advanced GPUs and AI ASIC accelerators remains high, Taiwan's stock assets will enjoy dual support from profit expectations and valuation premiums; however, once the pace of AI capital expenditures slows down, the excessively leveraged Taiwan stock market retreat will be significantly amplified due to the concentration of AI weightings, and may even become the first domino to cause the burst of the global AI investment bubble. FOMO sentiment runs rampant: Investors lever up like crazy, FOMO fire spreads from retail accounts to broker balance sheets Some analysts have started portraying the Taiwan stock market as a typical example of global AI bubble risk: the stock market surges by over 100% in a year, technology stocks account for about 80% of the index, margin financing for stock purchases surges by 160% in the past 12 months, and both retail and institutional investors are leveraging up. The real danger lies not in whether Taiwan has a genuine AI computing power industry moat, but in whether residents and financial institutions are chasing the same grand AI narrative with excessive leverage. According to a series of leveraged investment dynamics compiled by the media, 26-year-old Andy Cheng, who is facing unemployment, borrowed money in a leveraged manner to hold about $60,000 worth of Taiwan technology stocks, claiming that "buying any AI-related stocks in the Taiwan stock market will make money"; 39-year-old financial social media influencer Ada Hung, who had refused to invest with debt for many years, borrowed 5 million New Taiwan dollars, approximately $158,302, after seeing friends earn far more than herself in May. These cases actively illustrate that the market mood has shifted from "believing in the long-term fundamental competitiveness of TSMC" to "FOMO-driven extreme fear of missing out on wealth acceleration opportunities." For some investors wary of the "AI bubble" narrative, the data on leverage is undeniably more alarming. Over the past 12 months, the scale of investing in stocks through local brokerage firms has surged by 160%, nearing the historical high before the global internet bubble burst in 2000; not only exceeding the 50% increase in the last 12 months of the internet bubble, but also surpassing the roughly 94% increase in leverage seen in another strong AI market, South Korea. The surge in financing balances itself is not a signal of collapse, but when it resonates with the narrative of retail investors believing that "any stock will rise," it indicates that risk appetite has shifted from reevaluating fundamentals to leveraging up for irrational prosperity. The frenzy of leveraging is also starting to squeeze liquidity in the financial system. Some brokerage firms have hit their internal loan limits, forcing them to request more collateral and raise rates; some investors rejected by brokerage firms are turning to bank loans, or canceling financial products to release cash. More symbolically, the frenzy of borrowing has spread so intensively throughout the entire financial system, even disrupting recent central bank debt auctions. On June 3, there weren't enough buyers to purchase all the debt offered for sale, the first time this has happened. The debt sale by the Taiwan central bank on June 3 was impacted by insufficient buying, marking the first time in history that the entire debt was not sold, indicating that the stock financing frenzy is no longer just an internal phenomenon in the stock market, but is starting to affect the fixed income market and capital allocation in the financial system. Regulators still say it's manageable, but if AI capital expenditure momentum suddenly wanes, the spillover impact won't just stop at the stock market Brokerages themselves are being drawn into this wave of leverage expansion. To supplement operating capital, Taiwan's brokerages have issued nearly $1.2 billion in bonds this year, more than seven times the total financing amount in 2025, with some institutions turning to the syndicated loan market. At the same time, stock trading defaults in June have doubled to over an astonishing $2 billion New Taiwan dollars, reaching the highest monthly level since there is data available from 2019. As the stock market continues to rise, these numbers may be easily overlooked; however, once the market experiences a deep retracement, defaults, margin calls, and forced selling could become typical amplifiers of the cycle. According to Taiwan's financial regulatory authorities, the overall leverage in the industry is still considered manageable. The Securities and Futures Bureau stated that as of May, the 34 brokerage firms engaged in various leverage financing businesses had not surpassed regulatory restrictions, and stock market default amounts were still below 0.002% of total transactions. However, brokerages have begun to take proactive measures to reduce risk: some institutions have cut leverage ratios, suspended online loan applications, adjusted interest rates and margin requirements; industry insiders say that financing rates have been raised by 0.2 percentage points or more, and the interest rate adjustment for non-restrictive loans secured by stocks and ETFs can be as high as 1 percentage point, which, against a backdrop of Taiwan's central bank's benchmark interest rate of just 2%, denotes a clear repricing of risk. More crucially, stress testing has moved to the operational level of Taiwan's brokerage firms. Some institutions have begun to simulate how their loan portfolios would be impacted in the event of a sudden 20% or 30% market downturn; brokers such as KGI, Fubon, SinoPac, Cathay, have all acknowledged taking measures to lower leverage, adjust interest rates, modify margin requirements, or restrict non-restrictive loans for high-risk stocks. While the financial system has not yet spiraled out of control, risk management has switched from "post-event observation" to "preemptive firefighting." At the macro level, the tail risk lies in the potential spillover effects if the AI momentum suddenly fades, with impacts not just stopping at stock accounts. Alicia Garcia Herrero, the Chief Economist for Natixis Asia-Pacific, pointed out that potential spillover effects may include broker pressure, a decline in consumer spending, and a hit to export growth. For an economy like Taiwan with an extremely high weighting of tech stocks, a wealth effect intertwined with external demand cycles, a market adjustment could potentially transmit continuously through asset prices, consumer willingness, financial intermediaries, and export orders into the real economy. Taiwan is undoubtedly becoming the most typical dual-faced sample of the global AI super bull market: on one side lies a strong AI computing power supply chain, leading-edge processes, advanced packaging systems, and the increasing capital expenditures by tech giants like TSMC, Foxconn, and Wistron; on the other side lies the retail FOMO sentiment exploding, soaring leverage balances, brokerages borrowing to replenish capital, and rising default amounts, exhibiting the fragility of financial market leverage. While institutions like Goldman Sachs still maintain a bullish outlook, most analysts also expect the Taiwan stock market to continue rising in the coming months, indicating that the trend is not over; however, the real investment insight lies in the fact that the stronger the fundamentals of AI-related assets, the more likely they are to be overdrawn prematurely in a low-rate, high-belief, and high-leverage environment. Whether the AI super bull market in Taiwan can continue does not only depend on the orders related to AI chips dominated by TSMC and AI server clusters led by Foxconn, but also on whether the enormous leveraged exposure can be digested by the market without triggering forced de-risking.