老登股估值大溃败，一代资产的估值坐标系之死

Original Author: Xiaobing

On September 19, 2014, Alibaba was listed on the New York Stock Exchange, closing at $93.89 on its first day. That day, Alibaba's market cap was $231 billion, surpassing the combined value of Oracle and Intel.

On June 25, 2026, Alibaba closed at $95.07.

Between these two numbers lies a full twelve years.

At the same time, Meituan closed at HK$65.45, falling below its 2018 IPO price of HK$69.

Pinduoduo hovered around $79, returning to its June 2020 level.

Tencent's P/E ratio compressed to 12 times, nearly halved from its 10-year historical average of 25.7 times.

As for younger Chinese internet companies, Bilibili fell from a high of $156 to $18, a drop of 89%; Kuaishou plummeted from its IPO first-day high of HK$417 to HK$44, erasing nearly 90% of its market value; iQiyi, Zhihu, Douyu, Huya – each suffered a decline between 85% and 98%.

An entire generation of Chinese internet assets has undergone a collective valuation reset. What framework is the market using to price these companies? Or has the framework itself died?

The Establishment and Removal of Anchors

Looking back, the valuation logic of Chinese internet has gone through an exceptionally clear process of "anchor building and anchor removal."

From 2014 to 2017, the core narrative of global capital markets for Chinese internet was "a discount compared to US counterparts."

Alibaba was China's Amazon, Tencent was China's Facebook plus China's Visa, Baidu was China's Google.

This methodology was concise and powerful: first find the valuation multiple of the US counterpart, then apply a growth premium for the Chinese market and a governance discount to arrive at a reasonable price. Within this framework, Chinese internet companies generally enjoyed P/E ratios of 20 to 40 times.

Foreign capital poured in, and Chinese concept stocks were a must-have asset. This was the first anchor.

In 2018, the US-China trade war began. Global capital was forced for the first time to confront a question it had deliberately avoided: if US-China relations shifted from cooperation to competition, were the legal structures of companies operating in China but listed in the US still reliable? The VIE structure had never been explicitly recognized by Chinese law, but no one cared during a bull market. The trade war exposed this hidden vulnerability to sunlight for the first time. The valuation anchor loosened, but was not yet pulled out.

In October 2020, Ant Group's IPO was halted. The international capital market's pricing of "China regulatory risk" transformed from a vague discount factor into a dominant, explicit core variable. The 2021 antitrust storm pushed this logic to its extreme. Alibaba was fined 18.2 billion yuan, Didi was investigated the day after its IPO, and the education and training sector was wiped out overnight. Chinese concept stocks shifted from a "growth premium" to a "regulatory discount."

In 2022, delisting fears for Chinese concept stocks reached a peak.

The SEC placed over a hundred Chinese concept stocks, including Alibaba, Baidu, and JD.com, on a "provisional delisting list." Although China and the US eventually reached a compromise on audit working papers, the damage was done. Global index funds began systematically reducing their weightings in Chinese concept stocks, and some institutional investors liquidated their positions outright due to compliance requirements. This structural exit of capital compressed valuations from being sentiment-driven to being capital-flow-driven.

In early 2025, the sudden emergence of DeepSeek briefly ignited a wave of hope. Deutsche Bank called it China's "Sputnik moment," predicting that the valuation discount for Chinese assets would disappear.

The stock prices of Alibaba and Tencent rebounded by over 60% in the first two months of 2025. But this AI-narrative-driven revaluation fizzled out in less than six months. Entering 2026, the Pentagon added Alibaba and Tencent to its "Chinese military companies" list. Anthropic publicly accused Chinese companies of launching a large-scale distillation attack on its Claude model. Nasdaq introduced new listing rules tightening liquidity thresholds for Chinese concept stocks. Every attempt to rebuild the valuation anchor was quickly shattered by new geopolitical shocks.

At this point, the "US counterpart discount" valuation methodology has completely collapsed. The market is no longer pricing these companies based on their business models, growth rates, or profitability.

But it's not that simple.

The "Old Guard Stocks" on Both Sides of the Pacific

If we shift our focus from Chinese concept stocks on the NYSE to the US tech giants traded in the same building, we find: It's not just Chinese internet being abandoned by the market.

Microsoft was the worst-performing stock among the "Magnificent Seven" in 2026, falling over 20% for the year, from a high near $490 at the end of 2025 down to around $360. Its P/E ratio compressed from a five-year median of 34 times to 22 times, its lowest level in three years.

The company's fundamentals are intact: Azure cloud revenue grew 39% year-over-year, AI business annualized revenue exceeded $37 billion, and quarterly net profit of $31.8 billion hit an all-time high.

The market didn't care about these numbers. It was more concerned with another number: $190 billion, Microsoft's total capital expenditure budget for 2026, almost entirely directed at AI infrastructure. Its single-quarter capex exceeded its total annual spending from five years ago. Free cash flow dropped from $20.3 billion to $15.8 billion, the gap between profit and cash widening ever larger.

Microsoft's experience is not an isolated case.

In 2026, all of the Magnificent Seven underperformed the S&P 500. The four major hyperscale cloud providers (Amazon, Microsoft, Alphabet, Meta) are set to spend a combined nearly $700 billion on capex this year. The GPU clusters and data centers this money buys will only generate revenue over their 3-to-5-year depreciation cycle – investment comes first, returns follow later, and free cash flow gets crushed in the middle.

The deeper problem is: these companies are using massive amounts of capital to chase a technological paradigm that could potentially disrupt their own business models.

Microsoft's core revenue comes from Office subscriptions and Windows licenses, a SaaS model charging per user, with growth nearing its ceiling. The business logic of the AI era is consumption-based billing – pay per token used.

CEO Satya Nadella has already publicly admitted that every one of Microsoft's per-user billing businesses will transition to a hybrid "user + usage" model. GitHub Copilot already switched entirely to consumption-based pricing in June 2026. But the market's fear is precisely this: the old model had extremely high profit margins; whether the new model can maintain the same level remains unknown.

Looking from a distance, this picture forms a structural mirror image of the predicament faced by Alibaba and Tencent.

Alibaba's core e-commerce business is a highly profitable advertising machine, as stable as Microsoft's Office, yet the market gives it an increasingly lower valuation multiple. Tencent's WeChat ecosystem is still the most formidable moat in the Chinese internet, but slower game revenue growth and advertising business erosion by short-video platforms mirror Microsoft's predicament of being squeezed by Alphabet in search advertising.

The old giants on both sides are desperately pouring money into AI to save themselves. Alibaba is spending $55 billion to build AI infrastructure; Microsoft is spending $190 billion. But markets on both sides are casting a vote of no confidence on whether "this money can be earned back."

Chinese internet professionals habitually blame their companies' declines on regulatory pressure and geopolitics. US tech professionals habitually blame Microsoft's decline on "spending too aggressively." Strip away the surface narratives, and the same underlying phenomenon is occurring: AI-native companies are redefining the entire value chain of the tech industry, and the previous generation of platform giants, regardless of nationality, are transitioning from "companies defining the future" to "companies needing to prove they won't be eliminated by the future."

On the Chinese internet, this type of stock has earned a precise nickname – "Old Guard Stocks."

Nikkei: A Precedent for the Death of a Valuation Framework

This phenomenon of "the valuation coordinate system itself being replaced" is not the first time in global capital market history. The closest parallel is Japan after 1989.

On December 29, 1989, the Nikkei 225 index closed at 38,915 points, its all-time high.

That year, eight of the ten largest global companies by market cap were Japanese. NTT's stock price surged to 3 million yen per share two months after its IPO in 1987, giving it a market cap exceeding the combined value of the eight largest US companies at the time. Land prices in Tokyo were 350 times those in Manhattan. Sony bought Columbia Pictures, and Mitsubishi purchased Rockefeller Center.

Japanese investors of that era, much like Chinese internet professionals in 2020, genuinely believed the system they were part of would dominate the future of the global economy.

The trigger for the bubble's burst was the Bank of Japan's interest rate hike. But the magnitude of the decline was just the shallowest feature of this crisis. The duration and nature of the decline were truly suffocating.

The Nikkei lost half its value in the first half of 1990, halving to 14,000 points by 1992. If it had stopped there, it would have been just an ordinary bubble burst and valuation correction. But the Nikkei didn't stop. It continued a grinding decline for another decade, falling to 7,600 points in 2003, an 80% drawdown from its peak.

The core reason for this decade-long decline was not a collapse in the competitiveness of Japanese companies. Toyota was still the world's best automaker; Sony was still creating epoch-making consumer electronics. The problem lay deeper: Global capital no longer believed in the "Japan premium."

Before 1989, the market's valuation framework for Japanese companies was "the world's most efficient manufacturing civilization + a perpetually growing domestic market + unique corporate governance advantages."

After the bubble burst, these three assumptions were negated one by one. Manufacturing advantages were caught up by South Korea and China, the domestic market fell into deflation and population aging, and corporate governance proved to be a shelter for inefficiency. The old valuation framework died, but a new one was slow to emerge.

In 1989, 32 of the top 50 global companies by market cap were Japanese. By 2018, only Toyota remained.

How long did this period of vacuum last? About 25 years. The Nikkei didn't begin a truly trend-setting recovery until 2012, and it took until February 2024 to reclaim the 38,915-point level. The catalyst for this revaluation was not a comprehensive revival of the Japanese economy.

A specific person used a new language to redefine "why you should buy Japanese assets."

In the summer of 2019, Warren Buffett began buying shares of Japan's five major trading houses. The logic behind this investment was completely different from how the market had viewed Japan over the previous three decades. Buffett didn't talk about GDP growth rates, demographic trends, or technological innovation. His reasons were extremely simple: these five companies had low valuations, high dividends, stable cash flows, and were promoting genuine corporate governance reforms. He hedged currency risk by issuing yen-denominated bonds and used his own reputation to endorse Japanese assets. By 2025, Berkshire's stake in the five trading houses approached 10%.

Buffett provided a new valuation language for Japanese assets. The old language was "Japan will dominate the global economy." The new language was "low valuation + high dividends + corporate governance reform."

Where is the "New Language" for Chinese Internet?

Placing Japan's timeline alongside the experience of Chinese internet reveals several structural similarities that cannot be ignored.

The old valuation framework is dead. The failure of the "US counterpart discount" model is analogous to the collapse of the "Japan will dominate the world" narrative. The fundamentals of companies on both sides did not completely deteriorate; what was negated were the macro assumptions supporting the valuation premium. For Chinese internet, the macro assumption was "the deep integration of the Chinese market with global capital markets will continue." For Japan, it was "the Japanese model represents the most efficient form of capitalism." Both assumptions have been disproven.

A new valuation framework has not yet been established. The current market pricing of Chinese internet assets essentially applies a discount amidst the ruins of the old framework. Like Japan in 1995, the market knows the old price was wrong, but it doesn't know what the new price should be.

Based on Japan's experience, this vacuum period could be much longer than most people expect. It took Japan about 25 years from the bubble's burst to having a new valuation framework accepted by the market. The systematic dismantling of the Chinese internet valuation framework began around 2020; it's been only six years. If Japan's timescale offers any reference, the current position might just be the early stage of the revaluation process.

However, key differences exist between China and Japan. Japan's asset revaluation coincided with long-term deflation and population shrinkage, and corporate profitability did indeed deteriorate significantly after the bubble burst. Leading Chinese internet companies are still profitable. Tencent's annual net profit exceeds 220 billion RMB, and Alibaba's core e-commerce cash flow remains robust. This suggests that if a new valuation language can be constructed, the speed of revaluation could be faster than Japan's.

What could become the "new valuation language" for Chinese internet?

AI is the most obvious candidate, but also the most contradictory one.

Over the past two decades, the underlying business model of global internet companies has been highly convergent: capture users' attention, aggregate traffic on platforms, and monetize through advertising, e-commerce commissions, or in-game purchases.

AI is shaking the foundation of this business.

When AI agents can compare prices, place orders, and plan itineraries for users, users no longer need to open Taobao and scroll through pages themselves. When AI can recommend or even generate content based on preferences, the time users spend "scrolling" on a single platform decreases. Attention shifts from human eyes to the interface of AI agents; the gateway for traffic changes, and the platform's strategic position as an intermediary is hollowed out. This poses a threat to almost every core internet sector – e-commerce, search, social, content, and gaming.

If any Chinese internet company can successfully transition from an "attention platform" to an "AI infrastructure and service provider," it could potentially acquire a new valuation language.

The cruelty of this path lies in the fact that proactive disruption means dismantling the most profitable legacy businesses with one's own hands.

Taobao's advertising revenue is built on merchants bidding for rankings. If an AI agent bypasses the rankings to help users select products, this revenue stream will shrink. Each step of the transition erodes existing profits, while the profitability of the new model remains unproven.

If you chase AI, you must endure the crushing impact of massive capex on free cash flow. Microsoft's PE dropping from 34 times to 22 times is precisely the outcome of this story. If you don't chase AI, the market will judge you as obsolete.

Microsoft is betting $190 billion on rewriting its revenue architecture. If it wins, it becomes the infrastructure of the new era. If it loses, it's the biggest capital misallocation in history.