The Silicon-Based Era Has Arrived — But Who Gets to Ride It?

As China's chase the silicon-based boom, Chinese economist Li Xunlei warns of a K-shaped economy where a few giants reap the profits and the rest struggle to stay afloat

In recent weeks, the AI-driven boom in the semiconductor industry has lifted the optical communications sector on China’s A-share market, giving rise to a witty saying among Chinese investors: “Stand in the light — don’t just stand around.”(站在光里，不要光站着) urges the public to buy into optical communications stocks. Behind this tongue-in-cheek remark lies a deeper implication: at least among Chinese investors, the arrival of the silicon-based era now feels unstoppable.

Chinese economist Li Xunlei also shared his view about the boom recently. Li Xunlei is the Chief Economist of Zhongtai Securities. He has been engaged in macroeconomic, financial, and capital market research for over thirty years and is one of China's earliest experts in securities market research. Last April, he joined Premier Li Qiang's symposium on economic conditions.

In Li’s view, the prosperity of the silicon-based era is far from evenly distributed. Silicon-based firms and traditional “carbon-based” enterprises are diverging at an accelerating pace in terms of wealth creation, employment absorption, and valuation — giving rise to what he calls a “K-shaped economy.” The so-called Magnificent Seven account for 27% of total profits and more than 33% of total market capitalization in the U.S. stock market, yet together they provide only about 2.5 million jobs. NVIDIA, now the world’s most valuable company by market cap, employs a mere 36,000 people. A handful of firms “stand in the light,” reaping enormous profits, while the vast majority of traditional enterprises struggle to survive along the downward arm of the K. Meanwhile, the wealthiest 1% of Americans hold 50% of all stock market wealth, while 124 million people cannot come up with $400 for an emergency. He warns that AI is likely to widen the wealth gap further and intensify structural pressures on employment.

Li also cautions against the risk of an AI bubble fueled by the ongoing race in capital expenditure. Current projections suggest that the U.S. Magnificent Seven will spend between $700 billion and $750 billion in capex in 2026, a year-on-year increase of 70–80%. As he puts it: “Not expanding capex means certain death, but expanding capex doesn’t guarantee survival either.”不扩大资本开支一定是死，但扩大资本开支也未必能活

In the silicon-based era, Li argues, evaluating a company can no longer rest on commercial value alone; its social value must also be taken into account. The public must think ahead and consider how to respond to the social polarization and employment shocks that the silicon-based era will bring. He believes that only by vigorously developing the service sector and creating new employment opportunities can more people be brought aboard the train of economic progress, rather than left behind on the platform.

Below is the English ver of the article. His work was published on his WeChat account:

---

The Silicon-Based Era Is Here — Are We Ready?

In the first quarter of 2026, the Taiwan region posted real GDP growth of 13.69%, with nominal GDP growth reaching a remarkable 16.88%. As is widely known, this is largely thanks to the Taiwan region’s electronics industry — and, above all, to TSMC. In Q1 alone, TSMC posted a net profit attributable to shareholders of $18.1 billion, up roughly 60% year-on-year. NVIDIA did even better, with a Q1 net profit of $18.78 billion and a market capitalization approaching $5 trillion. For comparison, the entire U.S. GDP in 2025 was just $30.76 trillion.

I still remember visiting financial institutions in Taipei ten years ago, when everyone was lamenting hard times. In Q1 2016, the Taiwan region’s GDP had fallen 0.89% year-on-year — the third consecutive quarterly decline — and salaries had been stagnant for years.

Has the Taiwan region’s industrial structure undergone a radical transformation in the past decade? Clearly not. The electronics industry in the Taiwan region has long been highly developed. The reason for this boom is simple: the silicon-based era has arrived.

In fact, the electronics industry had already entered an upward cycle ten years ago, when everyone was talking about the Apple supply chain. These past few days, Berkshire Hathaway has been holding its annual shareholders’ meeting. At the meeting, Warren Buffett noted that ten years ago, he spent $35 billion on Apple stock, and over the past decade — interest included — it has generated $150 billion in profits for Berkshire, all while he did nothing.

To be fair, expecting a 95-year-old like Buffett to have a forward-looking, penetrating grasp of the silicon-based era is perhaps asking too much. In all likelihood, when he bought Apple ten years ago, he saw it mainly as a high-growth consumer stock.

Ten years ago, I recall, upstream raw materials in the global semiconductor industry — silicon wafers and the like — saw their first price increases in five years, which arguably signaled the dawn of the silicon-based era. At the time, I had a conversation with our firm’s chief electronics analyst. He argued that new downstream demand, led by HPC, IoT, and automotive electronics, was driving the official arrival of the fourth wave of rising silicon content. Upstream tightness in silicon wafers and capacity, combined with the surge in downstream silicon-intensive applications, was forming a closed loop — and memory chips, as the core category within that loop, would benefit the most.

Today, AI applications have become ever more pervasive. The launch of ChatGPT marked the dawn of a new era in which large language models are being widely deployed. Multimodal large models can now process and generate content across multiple modalities, overcoming the limitations of traditional LLMs in vision, hearing, and other domains. The dominant form is evolving from chatbots into agents capable of independent reasoning, tool use, and task execution. Major powers have entered an era of competition over computing power: from CPU shortages to GPU shortages and back to CPU shortages again; from optical chips to optical modules to CPO optical interconnects — a dazzling, kaleidoscopic progression.

One benefit of working in the securities industry is that, regardless of how quick a learner you are, you have no choice but to grit your teeth and let the silicon-based era push you along. If you fail to pay attention to the surge in A-share optical communications stocks, someone will shout at you: “Stand in the light — don’t just stand around.”

By contrast, Buffett’s Berkshire has shown considerable discipline, reducing its U.S. equity holdings for ten consecutive quarters and now sitting on $397 billion in cash. Buffett has likened today’s U.S. stock market to “a church with a casino attached” — valuations are simply too high. The right time to invest, he says, is when “no one else is willing to pick up the phone.”

From a valuation standpoint, the S&P 500 trades at an average P/E of nearly 30, with a dividend yield of only 1% and a P/B ratio of 5.6. By comparison, the CSI 300 has an average P/E of 14.4, a P/B of 1.47, and a dividend yield of 2.62%. The U.S. market does look expensive — especially given that U.S. inflation in March was still 3.3% and the 10-year Treasury yield stood at 4.4%.

Of course, valuation methods rooted in the carbon-based era may already be outdated. In the silicon-based era, only a handful of companies create enormous value, while most are destined to languish. According to available data, the Magnificent Seven collectively earned $567.25 billion in profits in 2025, against the S&P 500’s total of roughly $2.09 trillion — a 27.1% share of total profits. Their share of total market cap is even higher, at roughly 33–35%.

This raises the question: are we not living in an era of ever-deepening divergence, in which a handful of companies capture a vast share of society’s profits with staggering margins — the silicon-based growth model — while the majority of carbon-based firms struggle to stay afloat? This is the so-called K-shaped economy: the problem is that only a few companies and individuals ride the upward stroke of the “K,” while the majority slide down the other.

In other words, although we have entered the silicon-based era, our way of life will remain carbon-based for a long time to come. For example, these past few days, more than 50,000 people made the pilgrimage to Omaha to see Buffett — flying on planes, staying in hotels, touring Berkshire, eating steak, joining morning runs. All of this is carbon-based consumption. Even so-called silicon-based consumption requires enormous amounts of fossil fuel energy and releases vast quantities of carbon dioxide.

Let us now look at some features of this age of divergence and ask whether they bode well for healthy economic development. First, consider the ownership structure of assets in the U.S. stock market. According to Federal Reserve data, the wealthiest 1% of Americans hold about 50% of total stock market value, while the bottom 50% hold just 1%. The Fed has also reported that 124 million Americans cannot produce $400 in an emergency.

Second, although the seven U.S. tech giants account for more than 33% of total market cap, the number of jobs they create is strikingly limited. In 2025, their combined employment was around 2.5 million, of which Amazon alone accounted for 1.56 million. Nvidia — the world’s most valuable company by market cap — employs a mere 36,000. Moreover, in order to expand their capital expenditures, these giants have been laying off workers in droves: reportedly more than 100,000 jobs were cut in the first two months of this year alone.

In short, in the silicon-based era, AI firms can create far more wealth than their carbon-based counterparts, driving GDP growth — but at the same time widening the wealth gap and generating new employment pressures across society.

So the silicon-based era has arrived. How will it evolve, and what problems will it bring? Everyone is thinking about these questions, but clear answers remain elusive.

Among U.S. companies with market caps above $1 trillion, nearly all — aside from Walmart — are silicon-based enterprises. Over the past 30 years, traditional manufacturing, energy, telecoms, and financial firms have all dropped out of the top ten by market cap. Looking globally, the trillion-dollar club includes America’s Magnificent Seven along with Broadcom, Berkshire, and Walmart, plus TSMC from the Taiwan region, South Korea’s Samsung, and Saudi Arabia’s Aramco. Not a single firm from the Chinese mainland makes the list.

Comparing the equity markets of the Chinese mainland and the United States reveals one striking difference: the largest companies in the A-share market are simply not large enough, and the degree of divergence is less pronounced than in the U.S. Globalization levels are generally lower, large-cap trading volumes are relatively modest, and valuations for large firms are comparatively cheap — while small-caps trade at higher multiples and enjoy livelier turnover. Moreover, America’s big companies have grown largely through continual mergers and acquisitions, whereas successful M&A stories are far less common among the large companies on the Chinese mainland.

To be sure, the Chinese mainland has quite a few companies with market caps above one trillion RMB, but most are in traditional industries or are state-owned enterprises. These large firms employ tens or even hundreds of thousands of workers. Furthermore, the real employment generated by SOEs on the Chinese mainland is often underestimated, because in addition to formal employees, SOEs also rely heavily on labor dispatch and service outsourcing arrangements — in some central SOEs, these workers outnumber formal employees several times over.

Among the Chinese mainland’s internet giants, employment levels vary dramatically depending on the business model. Tencent, for example, has a market cap thirteen times that of JD.com and posted net profits of more than 220 billion RMB in 2025, yet its workforce numbers only around 100,000 — about one-ninth the size of JD’s. This underscores the point that evaluating a company requires looking not just at its commercial value, but also at its social value — especially as the silicon-based era delivers ever-larger shocks to employment.

After the internet era arrived in 2000, online transactions became increasingly active and dealt heavy blows to brick-and-mortar retailers. The rise of express delivery and food delivery services also brought with it massive “white pollution” — plastic bags, packaging boxes, tape, and the like.

Moreover, price wars among the internet giants have led to misallocation and waste of social resources. Today, the Chinese mainland’s flexibly employed workforce is estimated at 287 million (as of November 2025, according to the Tianjin CPPCC website), accounting for nearly 40% of the 725 million total employed population. The sheer size of this flexible workforce raises serious questions about future employment and social security that warrant thorough assessment.

Once the AI industry’s high-growth phase comes to an end, will these high-profit, low-employment silicon-based companies still be able to sustain such lofty valuations? Right now, America’s major AI firms are engaged in a frenzied expansion of capital expenditure: the Magnificent Seven are projected to spend between $700 billion and $750 billion on capex in 2026, a 70–80% increase over 2025. This is shifting U.S. GDP growth onto an investment-driven footing, and AI firms on the Chinese mainland are also ramping up capex sharply.

As the saying goes: not expanding capex means certain death, but expanding capex doesn’t guarantee survival either. In such a fierce, winner-takes-all environment, the bursting of the AI bubble is likely only a matter of time — just as the internet bubble burst years ago. Buffett has been steadily trimming stocks and accumulating cash, preparing for the coming winter.

Viewed over a longer historical horizon, the bursting of bubbles is actually a good thing. It restores rationality to investors and markets, and — through the elimination of weaker players — further improves labor productivity and the efficiency of resource allocation. After the U.S. internet bubble burst in 2001, for example, the internet became even more widely adopted, giving rise to a generation of tech giants that have since led the world into the silicon-based era.

For the global economy and society, the AI revolution is driving gains in labor productivity and human progress, rewriting traditional disciplines such as economics and sociology. Modern development economics, for instance, holds as a matter of consensus that an aging population must bring slower growth — and that once a society becomes super-aged, growth will fall below 3%.

Yet in Q1 of this year, the already super-aged Taiwan actually posted double-digit growth; South Korea, also super-aged, has performed quite respectably as well. The technological advance of the silicon-based era, then, will continue to push the wheel of history forward — even if it stirs up dust along the way, and even if it may at times crush the very roadbed beneath it.

From “Internet+” to “AI+,” we must prepare ourselves in advance. How do we confront the deepening divergence among societies, industries, companies, and households, and cushion the shocks this divergence will bring? How do we vigorously develop the service sector to create new jobs and counter the sharp drop in labor demand that the silicon-based era threatens to cause? And how do we plan ahead, so as to reduce the risks that a future AI bubble burst would spread across every industry?