GF Securities has released a research report stating that in 2026, the pricing of market prosperity in the A-share market has become extreme, characterized by: high-growth stocks seeing high gains, high-valuation stocks seeing high gains, and low-dividend stocks seeing high gains. Prosperity indicators such as revenue/profit growth rates and the rate of change in ROIC/ROE/gross margin have all been effective. Among these, companies ranking in the top 10% for revenue growth and adjusted growth in Q1 2026 saw average gains exceeding 40%. Conversely, indicators like cash flow, dividend yield, and valuation have been ineffective, even showing a negative correlation. Stocks with strong cash flow, high dividends, and low valuations have, in fact, declined more.
The current dispersion in valuations and performance between AI vs. non-AI, and the tech sector vs. traditional sectors, has reached historical highs. Does this inevitably lead to valuation convergence, or even a market downturn? Objectively, measuring sector valuation dispersion using the standard deviation of historical P/B percentiles or the difference between maximum and minimum values shows levels approaching historical peaks. Most importantly, extreme valuation dispersion is not a necessary condition for a market downturn. The main points from GF Securities are as follows:
Key Points of Contention
Last October, as innovation and application in the AI industry deepened, stock prices for AI industry chain companies in A-shares and Hong Kong also experienced significant surges, particularly in August last year when the market capitalization of some key companies roughly doubled. Consequently, by September 2025, dissenting voices grew, starting with skepticism over the valuations and market caps of AI industry chain companies, then debates over AI technical details, and eventually escalating to disagreements over the fundamental right or wrong of investment paradigms and philosophies.
Some voices have pointed to the rapid, forward-looking stock price increases during technological progress and industrial revolutions as the speculative original sin under a prosperity investment model, positioning prosperity investing as the antithesis of long-term value investing. Today, following further substantial gains in tech stocks in the first half of 2026, this divergence, particularly regarding the valuation of tech stocks, seems to have become an irreconcilable contradiction in the market.
Valuing Tech Assets
How to value tech assets has always been a major challenge for investors during every tech bull market. During prosperity upcycles, P/B valuations for tech stocks, especially some manufacturing companies, can easily reach 10x, 30x, or even higher. This phenomenon is not unique to A-shares; it occurs in the U.S. market as well. For instance, the overall P/B of the U.S. tech sector currently exceeds 13x (corresponding to an ROE of 30%+), while its historical P/B average is only around 4x (corresponding to an ROE of 20%+).
So, how should we understand tech stock valuations, calmly discern market disputes, and how should we respond to the current extreme valuation dispersion?
Irreconcilable Differences in Investment Philosophy
Before discussing valuation, it is essential to clarify: what type of investment methodology do we prefer (prosperity investing, value investing), and what is our preferred holding period (1-2 years, 3-5 years)? This preference is largely determined by our liability structure. In other words, investors with different liability structures, different investment methodologies, and different holding periods have inherently irreconcilable differences in their understanding of valuation.
Therefore, "valuation" is not an absolute truth nor does it have a standard answer. Stock prices often follow their own operational logic and are not swayed by such valuation disagreements. Valuation discusses the "discounting of the future," and how far that "future" extends depends on the holding period we, as investors, establish.
If one is a value investor, the investment horizon and holding period are correspondingly longer. The goal is to find undervalued quality assets at "cheap" valuations and achieve "mean reversion of value" over a long holding period. Thus, valuation should be seen as a strict disciplinary metric. The key to medium- to long-term value investing lies in buying excellent companies when they are mistakenly sold off and having the courage to hold them long-term when they are out of favor.
If one is a prosperity investor, the short cycle involves buying into the explosive prosperity of a particular industry/stock. The "second growth curve" of many companies cannot be assessed using static valuations or historical financial data. The expected holding period also changes with industry developments, and the consideration of valuation should be different. Compared to valuation, the marginal change in "prosperity" is far more important. The key to short- to medium-term prosperity investing is to avoid excessive attachment; once a marginal weakening in prosperity is observed, one should be willing to realize gains amidst the market fervor.
Furthermore, data will be used later to demonstrate that stock price performance in different industries (value-oriented vs. growth-oriented) inherently has different sensitivities to valuation—some industries naturally price "valuation" more, while others price "growth" more.
Returning to the current wave of AI development, whether in the U.S. or A-shares, investors are debating and discussing the valuation of tech companies. We understand that this is due to the market being composed of different types of investors, which is natural but also an irreconcilable difference. More important than the debate itself is clarifying the industry cycle we focus on, our expected holding period, and assessing whether "valuation" and "source of returns" are aligned.
Historical Perspective: The 90s Dot-com Era
The peak P/B does not correspond with the peak stock price: using the 90s dot-com era as a case study, how should we understand the high P/B ratios of hardware "pick-and-shovel" companies?
During the 90s dot-com era, leading hardware manufacturing companies also reached P/B valuations of 30-50x. The real danger is not the high P/B, but when the high ROE becomes "unsustainable." During the explosive phase of the 90s dot-com industry cycle, some hardware/manufacturing tech leaders saw their P/B ratios rapidly rise to high levels of 30-50x. For example, Dell's P/B was around 48x in 1997, Cisco's peaked at about 33x in 1999, and Qualcomm's peaked at about 40x in 1999.
Regarding the market discussion of "whether a P/B over 40x for a manufacturing company signifies a bubble," we derive the following insights from the dot-com era:
1. The core driver of high P/B for tech leaders was also "extremely high ROE." For instance, Dell's ROE remained around 80% in 1997-1998 and around 70% in the first half of 1999, pushing its P/B close to 50x. (Of course, ROE began to decline in the second half of 1999, further collapsing after the Y2K issue was debunked in 2000, highlighting the importance of fundamental tracking.)
2. For tech manufacturing companies, especially "pick-and-shovel" beneficiaries of explosive industry trends, their core assets cannot be quickly recorded on the balance sheet, making short-term book value of limited significance. Factors like R&D capability, technology patents, and customer barriers—summarized as "intangible assets"—lead to understated assets and higher P/B ratios. Compared to other asset-heavy (e.g., automotive, resources) companies, the asset/revenue ratio for tech manufacturing is much lower, and book value reflects more than just factory, production line, or machinery value. Greater R&D investment may also be expensed. Dell's asset/revenue ratio consistently ranged between 30-50%; whereas Alcoa, with a similar revenue scale, had an asset/revenue ratio above 100%.
3. If profit growth can be sustained, net profit will be converted into shareholder equity (net assets) year by year. The increase in net assets dilutes and amortizes the high P/B, which we summarize as—"digesting valuation through the denominator of P/B." Taking Dell as an example, its P/B peak occurred in 1997. During the craziest years of the dot-com boom, 1998-1999, although market cap rose, net asset growth was even faster, causing the P/B to fall back to around 30x. In the 90s, tech companies increased net assets mainly through sustained profitability (retained earnings), and acquisitions bringing goodwill.
4. However, when the high-growth narrative cannot be sustained and EPS declines, the market enters a phase of—"digesting valuation through the numerator of P/B," characterized by rapid market cap decline, which is difficult for investors to bear. A detailed review of the dot-com bust in the 90s shows the fundamental change was the confirmation on January 1, 2000, that the "Y2K bug" did not materialize, debunking expectations for high replacement demand. Faced with extremely high global hardware inventory levels, whether the industry chain had high P/B or high dynamic P/E, it meant the myth of high growth was shattered, and valuations became unsustainable.
Therefore, for the current global fervor in AI investing, the history of the 90s is worth careful consideration. However, history shows that what bursts a bubble is never high valuation itself, but the trajectory of industry development. The strict discipline of "prosperity tracking" is further elaborated in the data discussion below.
The Starting Point: Investment Horizon
In fact, the starting point for valuation discussion is the investment horizon. If the starting points differ, the debate is meaningless. The starting point for valuation discussion lies in setting the investment horizon—long-term value investing or medium- to short-term prosperity investing. For long-term value investing, valuation tends to revert to a mean; for medium- to short-term prosperity investing, one should not be overly fixated on valuation.
Historical experience in A-shares shows: on a 1-year horizon, the correlation between P/B/P/B percentile levels and price changes is not significant. However, extending the investment horizon to 3 or 5 years significantly strengthens the effectiveness of low P/B. In other words, for short-term investments on a quarterly or annual basis, looking at valuation is of little significance. But for long-term investments of 3 years or more, many assets revert to a valuation center priced by profitability, so the level of valuation at purchase largely determines investment returns.
Sector Sensitivity to Valuation
The valuation sensitivity of a sector depends on the volatility of its profitability: the effective ranges for value investing and prosperity investing are significantly different.
From a sector perspective, some sectors are insensitive to valuation, while others are highly sensitive, depending on the volatility of profitability. Contrary to intuition, sectors with high profit volatility (like tech) often have low valuation sensitivity, while sectors with low profit volatility have high valuation sensitivity. This is because high-volatility sectors focus more on future profit expectations and market potential, lacking a clear valuation anchor. For sectors with stable profitability, the market can easily assign valuation pricing, with the long-term valuation center mainly influenced by macro growth and interest rates.
1. Some sectors are insensitive to valuation, typically focusing only on the prosperity cycle. For example, growth sectors like electronics, communications, computers, power equipment, and defense. For the short to medium term (1-year horizon), the order of importance is: first-order profit changes (revenue/profit growth, rate of change in ROIC/ROE/gross margin) > second-order profit changes (rate of change in growth) > profitability level (ROIC/ROE). Valuation indicators (P/E, P/B, P/E percentile, P/B percentile, dividend yield) show only weak negative correlation or no correlation, offering limited reference value.
2. Some sectors have high valuation sensitivity, mainly those with stable profits. Sectors with high valuation sensitivity are primarily those with low ROE volatility, such as utilities, transportation, home appliances, and automobiles. Most of these sectors have high operational cycle stability and visibility, low ROE volatility, and usually a relatively stable valuation center. Therefore, the level of purchase valuation to some extent determines the level of returns.
2026: Extreme Prosperity Pricing, Inverse Valuation Correlation
1. In fact, a characteristic of A-share market pricing is the exclusive advantage of prosperity indicators. The overall pricing characteristic of A-shares is that the market rewards clear prosperity signals. The most effective are indicators representing prosperity, such as revenue/profit growth rates and the rate of change in ROIC/ROE/gross margin. For valuation indicators (P/E, P/B, P/E percentile, P/B percentile, dividend yield, PEG), effectiveness is unstable (sometimes positive, sometimes negative), depending on market style.
2. In 2026, prosperity pricing in A-shares became extreme, with valuation pricing showing an inverse correlation. In 2026, the pricing of market prosperity in the A-share market became extreme, characterized by: high-growth stocks seeing high gains, high-valuation stocks seeing high gains, and low-dividend stocks seeing high gains. Prosperity indicators such as revenue/profit growth rates and the rate of change in ROIC/ROE/gross margin were all effective. Among these, companies ranking in the top 10% for revenue growth and adjusted growth in Q1 2026 saw average gains exceeding 40%. Conversely, indicators like cash flow, dividend yield, and valuation were ineffective, even showing a negative correlation. Stocks with strong cash flow, high dividends, and low valuations, in fact, declined more.
Looking Ahead: The Key Lies in Marginal Changes in Prosperity
From the perspective of the three stages of prosperity investing:
(1) The simple stage of prosperity investing: acceleration in growth or rising ROE phase, profit and valuation double expansion (Stage 1); when an inflection point in prosperity appears, it may enter a phase of profit and valuation double compression (Stage 3).
(2) The complex stage of prosperity investing: growth decelerates but has not yet reached an inflection point, stock prices oscillate at high levels, possibly facing crowded trades, increased speculation, and even preemptive selling before the peak (Stage 2).
The key to valuation judgment lies in the marginal change in prosperity (marginal change in ROE or growth rate). During accelerating growth or rising ROE, valuation generally increases, and the level of immediate valuation is less important. During decelerating growth or falling ROE, valuation generally declines, and one can easily fall into a "value trap."
Extreme Dispersion Does Not Necessitate a Downturn
The final point of contention is: the current dispersion in valuations and performance between AI vs. non-AI, and the tech sector vs. traditional sectors, has reached historical highs. Does this inevitably lead to valuation convergence, or even a market downturn?
First, objectively, measuring sector valuation dispersion using the standard deviation of historical P/B percentiles or the difference between maximum and minimum values shows levels approaching historical peaks.
Historically, there have been four instances where severe valuation dispersion converged in the form of a broad bear market: late 2010-2011, the second half of 2015, 2018, and the second half of 2021-2023. There were also two instances where severe valuation dispersion converged in the form of a broad bull market: the second half of 2006-2007, and 2014 to the first half of 2015.
Most importantly, extreme valuation dispersion is not a necessary condition for a market downturn.
(1) In bull markets supported by industry trends, high levels of valuation dispersion can persist for a long time. For example, after valuation dispersion reached 95% in July 2020, it remained elevated for 20 months.
(2) In bull markets supported by industry trends, the average period from the peak of valuation dispersion to the market peak is 19 months: 2006-2007 (May 2006 - September 2007, 17 months), 2013-2015 (July 2013 - May 2015, 22 months), 2020-2021 (July 2020 - December 2021, 18 months).
Comments