The market never punishes storytelling, only stories that fail to become numbers. On August 19, 2026, Unitree Robotics went public with a stunning debut, surging 629% at the opening bell and closing up 460%, reaching a market value of 341.8 billion yuan. Its online lottery winning rate of 0.018% set a record low for the STAR Market. Just fourteen months earlier, in June 2025, its final private funding round had valued the company at a mere 12.7 billion yuan post-investment.
On July 27, 2026, ChangXin Memory Technologies, which posted a net profit attributable to shareholders of 1.875 billion yuan in the prior year, listed with a closing market value of 3.28 trillion yuan on its first day — meaning the market was willing to pay approximately 1,750 yuan for every yuan of annual profit (or about 309 yuan on a non-recurring basis). Its stock price rallied sharply after listing, briefly touching 4 trillion yuan intraday on July 31, making it the first company in A-share history to surpass the 4-trillion-yuan mark. By August 13, its market value stood at approximately 3.54 trillion yuan, overtaking Tencent to become China's most valuable listed company.
Market value has never been a pure calculation exercise; it is a negotiation between those who tell stories and those who listen to them. Rewinding five years, one can see that China's tech companies have completed a full narrative shift within the 100-billion-yuan market cap club: the object of pricing, the storytellers, and the grammar of valuation have all changed. The object has shifted from "certainty of growth" to "necessity of position"; the storytellers have shifted from foreign capital and star public funds to state capital and patient capital; and the grammar has shifted from discounted cash flow to a sovereign option-style pricing based on "must-exist" status.
Nobel laureate Robert Shiller, in his seminal work "Narrative Economics," proposed that economic narratives spread like viruses, following infection curves and undergoing processes of forgetting. Narratives often appear in "constellations" — "constellations contain not only theoretical stories but also human interest stories," and single narratives have limited credibility; they only become popular by borrowing strength from one another. Using his framework, we can observe how the narrative constellation of China's capital market has been reshaped over the past five years.
To understand what changed in 2026, one must first recall what was being bought in 2021. In February 2021, China's internet platform economy reached its valuation peak: the Hang Seng Tech Index hit an intraday record high of 11,001.78 points in mid-February, corresponding to a trailing P/E ratio of 65-70 times, while the Nasdaq Index traded at a P/E of roughly 50 times at the same period. The Hang Seng Tech Index commanded a valuation premium of about 30%-40%. That same February, Tencent's market value reached 7.3 trillion Hong Kong dollars, Meituan 2.6 trillion, Kuaishou hit approximately 1.4 trillion on its debut day, Pinduoduo 260 billion US dollars, and Baidu 100 billion US dollars.
The narrative grammar at the time was simple and uniform: traffic empires were priced at approximately 35 times forward earnings, loss-making platforms were valued on price-to-sales ratios, but all clearly outlined a path to profitability — the market believed that barriers built on user scale could sustain excess profits over the long term. "Burn cash for scale, scale for valuation" was the prevailing consensus; every new business segment a platform entered was seen as a new engine for market value growth. The storytellers then were northbound capital, star public funds, and foreign investors, with channels through dual listings on US and Hong Kong exchanges. The most popular concepts in the narrative constellation were "user value (MAU×ARPU)," "traffic monetization efficiency," "two-sided platforms," and "track theory."
It should be noted that in the 2021 tech valuation system, hardcore semiconductor and new energy companies like SMIC, CATL, and BYD had already gained attention, though their market caps and market focus still lagged internet giants by orders of magnitude. Their stories were more tied to "early-stage import substitution" and "rising industry penetration rates" rather than today's "AI computing power race." SMIC's then market cap of over 400 billion yuan was already the ceiling for the semiconductor sector, yet it pales in comparison to Cambricon and ChangXin Memory in 2026.
Shiller noted that narratives, like epidemics, collapse faster than they spread. The period from 2022 to 2024 was one of narrative reconstruction: platform companies collectively shrank their balance sheets, "cost reduction and efficiency improvement" replaced "boundless expansion," and a 10x P/E ratio became standard for Chinese internet companies. The secondary market began evaluating internet giants with a framework typically reserved for utilities. Of course, the protagonists of the old story did not truly die — they merely hibernated, waiting for a new constellation.
Interestingly, the new constellation was illuminated by a company that has not yet gone public. In January 2025, DeepSeek released R1, causing Nvidia to drop 17% in a single day and forcing global capital to reprice Chinese AI. The "DeepSeek moment" demonstrated the power of narrative constellations: import substitution had been an old story since 2018, and tech self-reliance was not a new phrase, but when the old narrative of import substitution, the new narrative of AI catch-up, the policy narrative of patient capital, and the institutional narrative of STAR Market reform began endorsing each other, scattered stars connected into a new constellation.
In February 2025, Alibaba announced plans to invest over 380 billion yuan in AI infrastructure over three years, marking a turning point for big tech stories shifting from cost reduction to AI capital expenditure. Under the new narrative constellation, the market is more willing to pay for "position stories" within grand narratives: ChangXin is "national memory," Cambricon is "China's Nvidia," Zhipu and MiniMax are placed in a coordinate system competing with OpenAI, and Unitree carries the banner of global embodied intelligence. From a global perspective, becoming the one that "must exist" in future tech competition is worth betting on more than being "about to turn profitable."
Even the anchor of the narrative constellation has changed. In the previous round of storytelling, the highest praise was "China's version of XX," with valuations discounted against Silicon Valley; this round it's "XX's China," where the gap with global top companies is no longer narrated as risk but as upside potential. In the A-share market in 2021, the way to praise a new company was to compare it to Moutai — "Ning-Moutai" (CATL), "Security-Moutai" (Hikvision). In August 2025, Cambricon's stock price surpassed Moutai, briefly claiming the title of "king of stocks"; in June 2026, Foxconn Industrial Internet's market value also exceeded Moutai. Since then, few in new narratives have compared computing power and chip companies to Moutai.
The list of storytellers has also changed. In 2021, the marginal pricers were northbound capital and star public funds; after 2024, Central Huijin clarified its quasi-national team fund status, the National Venture Capital Guidance Fund entered with a 20-year "patient capital" positioning, and passive index funds surpassed active equity funds in scale. In ChangXin's shareholder list, state capital and national funds dominate, while Unitree's final funding round was led by Tencent, Alibaba, China Mobile, and Geely. When the pricing party's objective function includes not only financial returns but also industrial cultivation and strategic allocation, the market's tolerance for high valuations is structurally amplified.
ChangXin and Unitree are not isolated cases: Cambricon Technologies Corporation Limited once had a P/E ratio exceeding 3,000 times, and its intraday market value briefly surpassed one trillion yuan at the end of June this year, becoming the first trillion-yuan stock on the STAR Market. Zhipu, with annual revenue of approximately 700 million yuan, touched about 1.33 trillion Hong Kong dollars intraday on June 22. In this round of narrative shift, the feedback loop between price and story has become clearer. Tsinghua University's Global Securities Market Research Institute, in its "China Listed Companies Market Value Analysis Monthly Report (July 2026)," noted that ChangXin Memory's listing "reshaped the regional market value landscape," pushing Hefei's securitization rate to 340%, ranking third nationally — a city's asset pricing rewritten by the story of a single company.
The listing channel itself has become part of the narrative. In June 2025, the CSRC launched the "1+6" reform of the STAR Market, restarting the fifth set of listing standards and adding a "tech innovation growth tier" specifically serving hard-tech companies with major technological breakthroughs and broad commercial prospects but not yet profitable. ChangXin Memory was the first case under the pre-review mechanism, going from acceptance (December 30, 2025) to listing in just seven months. Moore Threads took 170 days from acceptance to listing; Unitree took only 104 days from acceptance to registration approval. Zhipu and MiniMax completed listings under Hong Kong's inclusive framework for loss-making large model companies.
Tsinghua University's data provides a "balance sheet" for this shift: the number of A-share companies with market caps above 100 billion yuan has grown from 61 to 205 over the past decade, with the electronics sector surpassing banking with 35 companies to take the top spot, and private enterprises growing from 7 to 67. Approximately 60% of the decade's market cap growth among 100-billion-yuan companies came from earnings expansion and 40% from valuation repair — the ratio of performance to story can be measured with this ruler.
If ChangXin and Unitree represent the present tense of the new narrative, DeepSeek is its future tense. According to reports, it has initiated IPO preparations, planning to file with the STAR Market within the year and list in 2027. Its latest funding round reportedly values the company at approximately 71 billion US dollars pre-investment, equivalent to about 500 billion yuan, with the first tranche reportedly exceeding 50 billion yuan. The investor list includes Tencent, CATL, JD.com, NetEase, and the National AI Industry Fund. Under Shiller's framework, DeepSeek is a textbook "super-spreader," possessing nearly all the high-transmission characteristics on Shiller's narrative checklist.
Consider the celebrity effect. Shiller wrote, "In every event, there is a story that spreads like a virus, and this spread typically leverages some celebrity effect" — Liang Wenfeng's low-profile and mystery is the perfect carrier for this effect. Just as Shiller noted in the Bitcoin narrative, without repeated promotion of the Satoshi Nakamoto mystery, Bitcoin's transmission rate would not have been so high. A founder who refuses exposure is more contagious than ten CEOs on roadshows. Then there's the human interest story. Shiller reminds us that narrative constellations "contain not only theoretical stories but also human interest stories" — DeepSeek's human story is nearly perfect: small team, low cost, open source, defeating a giant — a David and Goliath story that fits seamlessly into the theoretical story of great power competition. And "rigorous criticism typically lacks transmission power" — technical questions about its training cost methodology and computing power sources never truly entered the transmission curve.
Looking beyond China, the narrative constellations of global tech companies are also undergoing dramatic shifts. This summer, Alphabet announced plans to raise 85 billion US dollars for AI investment, and SK Hynix set an ADR issuance record with 29.4 billion US dollars. In June, SpaceX priced its IPO at 135 US dollars per share, raising 75 billion US dollars with a valuation of 1.77 trillion US dollars, far surpassing Saudi Aramco's 25.6 billion US dollar record that had stood for seven years, setting dual records for IPO fundraising and valuation. But it barely had time to settle in before challengers appeared at the door: on August 21, Bloomberg reported that Anthropic could publicly file its IPO documents as early as late August, with fundraising expected to match or even surpass SpaceX. Investors are discussing valuations starting at 2 trillion US dollars, with a possible listing as early as October. Combined with OpenAI, which has already filed confidentially, the fall of 2026 may see global capital markets settling accounts for three "story companies" simultaneously.
Expanding the ledger reveals the differences and interconnections between Chinese and American narrative grammars. Once OpenAI and Anthropic list, the public market will for the first time answer the question "what is a pure AI company worth" at a price approaching 2 trillion US dollars — and that answer will not stay on the other side of the Pacific.
In the same summer's waiting list, there is also a company with a very different style. According to Bloomberg's June report, Xiaohongshu has hired Goldman Sachs and CICC and confidentially submitted a listing application to the Hong Kong Stock Exchange, potentially becoming one of Hong Kong's largest tech listings in recent years. Its secondary share transfer valuation has reached 50 billion US dollars (approximately 350 billion yuan), with listing valuations reportedly ranging up to 70 billion US dollars (approximately 500 billion yuan) — under either figure, its debut day would place it in the front row of the 100-billion-yuan club. According to the report, Xiaohongshu turned profitable for the first time in 2023 (with revenue of 3.7 billion US dollars and net profit of 500 million US dollars); in 2025, revenue is expected to reach 6-8 billion US dollars with net profit of approximately 3 billion US dollars; monthly active users exceed 400 million, and advertising revenue accounts for 76%.
Placing it into the new narrative constellation yields an almost nostalgic reading: at a 50-70 billion US dollar valuation against approximately 3 billion US dollars in annual profit, the P/E ratio is just 17-23 times — far below Tencent's peak forward P/E of about 35 times in 2021. It is still telling the previous round's story: monthly active users, daily usage time, the "seed-grass to pull-grass" conversion loop, and the monetization path of advertising plus e-commerce. In a market where Zhipu once commanded a price-to-sales ratio exceeding 1,200 times, it can only use the previous round's narrative grammar.
This is what Shiller calls "recurrence and mutation" in action. He found that long-term narratives never completely go silent; they become popular again after mutating — just as influenza resurges after mutating to overcome acquired immunity. The consumer internet narrative has been dormant for five years, and to return, it must undergo two mutations. The first is the human interest story — Shiller says human interest stories and theoretical stories borrow strength from each other for a constellation to take shape. This is exactly what Xiaohongshu excels at: it has always told the story of a new lifestyle community, a template that never existed in the old narrative era. The second is the AI story it lacks but must add: it needs to frame its search and content ecosystem as an entrance to the AI era, not a victim — Bloomberg's same report noted that AI applications like MiniMax are threatening the traffic and business models of platforms like Xiaohongshu.
A company with old grammar must now defend itself using the vocabulary of the new constellation. Xiaohongshu thus becomes a natural controlled experiment — its offering price will answer a reverse question: when buyers are accustomed to paying for "position," is there still demand for "certainty of growth"? If it meets a cold reception, it means the old grammar has been expelled by the new constellation; if it is warmly received, it suggests the market is pricing both narratives simultaneously. The significance of the story constellation lies precisely in this: no single star can illuminate the entire sky alone.
One cannot overlook what Shiller calls "narrative repair": earnings reports, lock-up expirations, and placements can at any time create "antibodies" against commercial narratives, causing premiums to quickly give back gains. Several star companies in the new narrative camp have already demonstrated this. Zhipu fell about 20% in a single day after completing a 31.4 billion Hong Kong dollar placement, with its market value retreating over 60% from the 1.33 trillion Hong Kong dollar peak. MiniMax has fallen more than 80% from its March peak, hovering near 100 billion Hong Kong dollars. Unitree's stock plummeted 18.7% on its second trading day, closing with a market value of approximately 278 billion yuan, with opening-day chasers facing paper losses of about 37% within two days.
Of course, DeepSeek has not yet listed, meaning its narrative lacks corresponding antibody mechanisms: no quarterly reports to disappoint, no lock-up expirations to release shares, no price to betray. It remains in the most brilliant phase of the narrative constellation — everyone talks about it, no one trades it. It is being priced according to current enthusiasm, which may have raised the ceiling of this narrative cycle.
Starting in 2027, members of the new 100-billion-yuan club will be re-examined. In 2021, the market asked: how many users do you have? Now it asks: how irreplaceable are you? The question has changed, but the rules for answers have not — stories must ultimately be translated into numbers. This does not mean the bubble is on the verge of bursting: Cambricon Technologies Corporation Limited saw revenue grow 453% in 2025 and turned profitable for the first time; ChangXin's revenue grew to 61.8 billion yuan with its first profit; Foxconn Industrial Internet's net profit grew 52% — their stories are being caught up by the numbers. The market never punishes stories; it only punishes stories that fail to become numbers.
Comments