In the first quarter of 2026, Taiwan's GDP growth rate was 13.69%, with a nominal GDP growth rate as high as 16.88%. As everyone knows, this is mainly due to the contribution of Taiwan's electronics industry, with TSMC making the largest contribution. TSMC's net profit attributable to shareholders in the first quarter reached US$18.1 billion, a year-on-year increase of about 60%. Nvidia's net profit in the first quarter was even higher, reaching US$18.78 billion, with a market capitalization approaching US$5 trillion. In comparison, the US GDP in 2025 will only be US$30.76 trillion. I remember when I visited financial institutions in Taipei 10 years ago, they were all lamenting: Taiwan's GDP fell by 0.89% year-on-year in the first quarter of 2016, marking the third consecutive quarter of decline, while their salaries hadn't increased for many years. Has Taiwan's industrial structure undergone a dramatic transformation in the past 10 years? Clearly not. Taiwan's electronics industry has always been highly developed, and the reason for its prosperity is simple: the silicon-based era has arrived. In fact, the electronics industry had already entered an upward cycle 10 years ago. At that time, everyone was discussing the Apple supply chain. Berkshire Hathaway is currently holding its shareholders' meeting, where Buffett said that he spent $35 billion on Apple stock 10 years ago, and including interest, it has brought Berkshire $150 billion in profit over the past 10 years, while he did nothing. Objectively speaking, expecting a 95-year-old like Buffett to have such a forward-thinking and profound understanding of the silicon-based era is truly asking too much. Actually, when he bought Apple 10 years ago, he probably saw it as a high-growth consumer stock. I remember 10 years ago, the prices of upstream raw materials for the global semiconductor industry saw their first increase in five years, such as silicon wafers, which may have signaled the arrival of the silicon-based era. I even had a conversation with the company's chief electronics executive at the time, who believed that new downstream demands, represented by HPC, IoT, and automotive electronics, were driving the official arrival of the fourth wave of silicon content upgrades. Upstream silicon wafers and tight capacity, coupled with emerging downstream demands for higher silicon content, formed a closed loop, with memory chips, as the core product of this loop, benefiting the most. Today, AI applications are becoming increasingly widespread; for example, the release of ChatGPT marks the entry of large language models into a new era of widespread application. The emergence of multimodal large models has the ability to process and generate multiple modalities, making up for the limitations of traditional large language models in visual, auditory, and other modalities; the form is evolving from chatbots to intelligent agents capable of independent thinking, calling tools, and performing tasks. Major powers have entered an era of computing power competition, from CPU to GPU shortages, and back to CPU shortages; from optical chips to optical modules, and then to CPO optical interconnects—a dazzling and bewildering array. The advantage of working in the securities industry is that, regardless of your learning ability, you have to grit your teeth and be pushed forward by the silicon-based era. If you don't care about the explosion of the A-share optical communication sector, some people will shout, "Stand in the light, don't just stand there." In contrast, Warren Buffett's Berkshire Hathaway has remained steadfast, reducing its holdings of US stocks for ten consecutive quarters and now holding $397 billion in cash. Buffett believes that the current US stock market is like a "church with a casino attached," with valuations that are too high, and that the right time to invest is when "no one else is willing to answer the phone." Global Major Stock Index Valuations and Dividend Yields Source: WIND, Zhongtai International From a valuation perspective, the S&P 500's average price-to-earnings ratio (P/E ratio) is close to 30, its dividend yield is only 1%, and its price-to-book ratio (P/B ratio) is as high as 5.6. Compared to the CSI 300's average P/E ratio of 14.4, P/B ratio of 1.47, and dividend yield of 2.62%, it is indeed expensive, especially considering that the US inflation rate in March was 3.3%, while the 10-year Treasury yield was as high as 4.4%. Of course, this valuation method based on the carbon-based era may be outdated. In the silicon-based era, only a very few companies will create massive value, while most companies are destined to remain stagnant. Statistics show that the total profit of the seven largest US companies in 2025 was $567.25 billion, while the total profit of the S&P 500 was approximately $2.09 trillion, meaning the seven largest companies accounted for 27.1% of the profits; their market capitalization share was even higher, at approximately 33-35%. Therefore, we can argue that we are living in an increasingly polarized era, where a small number of companies earn a large portion of society's profits with astonishingly high profit margins—these companies belong to the silicon-based development model; while the majority of carbon-based companies struggle. This is the so-called K-shaped economy. The problem is that the companies and individuals represented by the upward stroke of the letter K are a minority, while those represented by the downward stroke are the majority. In other words, although we have entered the silicon-based era, our lifestyle will remain predominantly carbon-based for a long time. For example, the recent pilgrimage of over 50,000 people to Omaha to "pay homage" to Warren Buffett, including flying, staying in hotels, visiting Berkshire Hathaway, eating steak, and participating in jogging activities, all fall under the category of carbon-based consumption. Even silicon-based consumption requires significant energy consumption from fossil fuels and releases large amounts of carbon dioxide. Let's examine some characteristics of this era of differentiation to see if they are conducive to healthy economic development. First, consider the asset holding structure in the US stock market. According to data from the Federal Reserve, the top 1% of Americans hold approximately 50% of the total stock market value, while the bottom 50% hold only 1%. Furthermore, the Federal Reserve stated that 124 million Americans cannot afford a $400 emergency fund. Secondly, although the seven major US tech giants account for over 33% of the total market capitalization, the number of jobs they create is quite limited. In 2025, their total workforce is estimated at around 2.5 million, with Amazon alone accounting for 1.56 million, while Nvidia, the world's largest tech company by market capitalization, has only 36,000 employees. Moreover, these giants have been laying off large numbers of employees to expand capital expenditures; reportedly, they laid off over 100,000 people in the first two months of this year alone. In short, in the silicon-based era, AI companies can create far more wealth than carbon-based companies, driving GDP growth, but simultaneously widening the wealth gap and creating new employment pressures across society. So, the silicon-based era has arrived. How will it develop in the future, and what problems will it bring? Everyone is thinking about this, but there are few clear answers. In the US, apart from Walmart, most companies with a market capitalization exceeding one trillion US dollars are silicon-based. Over the past 30 years, traditional manufacturing, energy, telecommunications, and finance have all fallen out of the top ten companies by market capitalization. Even globally, among companies with a market capitalization exceeding one trillion US dollars, besides the seven US giants plus Broadcom, Berkshire Hathaway, and Walmart, there are TSMC from Taiwan, Samsung from South Korea, and Saudi Aramco; no mainland Chinese companies are on the list. Comparing the stock markets of China and the US, a significant difference emerges: the market capitalization of A-share companies is not large enough, and the degree of differentiation is not as pronounced as in the US stock market. Furthermore, globalization is generally low, with large companies accounting for a small percentage of transactions and relatively cheap valuations, while small companies are relatively expensive and actively traded. In addition, large American companies typically grow stronger through continuous mergers and acquisitions, a model with few successful examples among large domestic companies. Of course, my country still has many companies with market capitalizations exceeding one trillion RMB, but these are mainly in traditional industries and state-owned enterprises (SOEs). These mega-enterprises provide employment for tens or even hundreds of thousands of people. Moreover, the actual number of jobs created by Chinese SOEs is underestimated because, in addition to regular employees, SOEs employ a large number of dispatched workers or service outsourcing workers, whose numbers in some central SOEs are several times greater than regular employees. Among large internet companies, the gap in employment numbers widens due to differences in business models. For example, Tencent's market capitalization is 13 times that of JD.com, and its net profit in 2025 is over 220 billion yuan, but its employment is only over 100,000, only about one-ninth of JD.com's. This means that evaluating a company requires considering not only its commercial value but also its social value, especially given the increasing impact of the silicon-based era on employment. Since the advent of the internet age in 2000, online transactions have become increasingly active, posing a huge challenge to brick-and-mortar stores. Simultaneously, the widespread adoption of express delivery and food delivery services has led to a large amount of "white pollution" (packaging bags, boxes, and tape, etc.). Furthermore, price wars among large companies have resulted in the misallocation and waste of social resources. Currently, my country's flexible employment population is estimated at 287 million (as of November 2025, see the Tianjin Municipal CPPCC website), accounting for nearly 40% of the total employed population of 725 million. The excessive number of flexible employment positions raises concerns about their potential impact on future employment and social security, requiring thorough assessment. After the high-growth phase of the AI industry ends, will these highly profitable, low-employment silicon-based companies be able to maintain their large market capitalization? Currently, major US AI companies are aggressively expanding capital expenditures. The capital expenditures of the seven largest US AI companies are projected to reach $700-750 billion by 2026, a 70-80% increase from 2025, contributing to a shift in US GDP growth towards investment. Chinese AI companies will also see significant increases in capital expenditures. One argument is that not expanding capital expenditures will inevitably lead to failure, but expanding them doesn't guarantee survival. In this fiercely competitive environment where only the strong survive, the bursting of the AI bubble is likely inevitable, just like the bursting of the dot-com bubble. For example, Buffett has been continuously reducing his stock holdings and accumulating more and more cash in preparation for the coming economic downturn. Looking at history in the long run, the bursting of various bubbles is actually a good thing, allowing investors and markets to return to rationality and further improving labor productivity and resource allocation efficiency through natural selection. For instance, after the bursting of the US internet bubble in 2001, the internet became more widespread, leading to the rise of a group of tech giants and ushering in the silicon-based era. In terms of the global economy and society, the AI revolution is driving increased labor productivity and human progress, rewriting traditional disciplines such as economics and sociology. For example, from the perspective of modern development economics, it is generally agreed that population aging will inevitably lead to economic slowdown, especially in the super-aged stage, where economic growth will further decline to below 3%. In the first quarter of this year, Taiwan, which has already entered the super-aging stage, actually achieved double-digit economic growth; South Korea, which is also super-aging, did not lag behind in economic growth. Therefore, the technological development of the silicon-based era will continue to drive the wheels of history forward, although it will raise dust and may even collapse the roadbed. From "Internet+" to "AI+", we need to prepare in advance for how to cope with the increasing differentiation among society, industries, enterprises, and households, and how to mitigate the resulting shocks; how to create new job opportunities by vigorously developing various service industries to cope with the impact of the silicon-based era on the significant reduction in employment demand; and how to plan ahead to reduce the risks to various industries from the potential bursting of the AI bubble.
Brian
Jixu
YouQuan
Joy
Hui Xin