The story of Zhihu began with a legendary aura from the primary market: this company, labeled as "China's AI first stock," has seen its valuation increase by approximately 130 times from the angel round to the IPO. After landing in Hong Kong in early 2026 with the stock code 02513.HK, its stock price continued to rise within just a few months, accumulating a rise of about 24.6 times by June 22. A total of 57 external investors invested approximately 8.36 billion yuan, and now its corresponding book value in the secondary market is about 770.8 billion Hong Kong dollars, with an overall return of about 85 times — this is the extreme answer given by the capital market when a few stand at the forefront of the AI wave, and it represents the most vivid scene of China's AI wealth creation myth. Meanwhile, on the other end of the screen, there is a completely different picture: on June 23, before the opening of the U.S. stock market, the NASDAQ 100 futures fell about 2.5%, Intel fell about 7.6% before the opening, AMD fell more than 6%, and Qualcomm fell about 5.4%, with the chip sector as a whole under pressure; previously, SpaceX had accumulated a decline of about 23% over three trading days, and before the opening, it further dropped about 2.3%. The short-term adjustment of America's star tech assets exposed the risks and fragility of this wave of enthusiasm under the spotlight. However, behind the dramatic price swings, the underlying logic of supply and demand as well as technology has not shifted: according to a TrendForce report, the DRAM shortage has already spread from mainstream products to older memory types like DDR2, and prices are expected to continue rising in Q3 2026; Samsung Electronics has just launched the UFS 5.0 flash solution, with sequential read speeds of about 10.8GB/s and write speeds of about 9.5GB/s, over twice the performance of UFS 4.1, with plans for mass production in Q4 to create a broader "data expressway" for AI applications at the edge and mobile end; Jefferies believes that China's AI is entering a period of Token consumption expansion, with low-cost models reshaping user behavior, and OpenRouter's statistics show that Token consumption increased by about 4.7% compared to the previous week as of June 22. Between the explosive growth curve of Zhihu and the pre-opening adjustment of U.S. tech stocks, storage prices, interface specifications, and Token data all point in the same direction: while short-term sentiment can change rapidly, the computing power, storage, and usage intensity supporting AI continue to rise globally.

130 Times Valuation and 85 Times Return: The Feast of China's AI First Stock

If we compress the financing history of Zhihu 02513.HK into a curve, the valuation amplification of about 130 times from the angel round to the IPO is not just a textbook-like entrepreneurial myth, but also a reflection of the collective sentiment in China's primary AI market. The entry of early players like Lei Jun and Meituan can be seen as marking coordinates on a roadmap that has not yet taken shape; subsequently, each increase in valuation verifies a simple and brutal logic: as long as one bets on the "AI infrastructure" card, the market is willing to pay much higher prices than for traditional internet. By the time Zhihu was labeled as "China's AI first stock," the primary market had already completed the pricing of risks and imagination along this curve, behind the hundredfold valuation increase lies a concentrated bet on the future demand for computing power, models, and data.

The real feast was ignited after the IPO. Since its debut on the Hong Kong Stock Exchange at the beginning of the year, Zhihu's stock price has risen approximately 24.6 times as of June 22, pushing the stock price curve in the secondary market to extreme levels, continuing the story of the primary market. Initially, 57 external investors collectively invested about 8.36 billion yuan, and now the corresponding book value has reached approximately 770.8 billion Hong Kong dollars, with an overall return of about 85 times. These numbers alone carry enough narrative tension: someone condensed a decade's venture capital returns into just a few years during the first wave of China's AI. For entrepreneurs, names like Lei Jun and Meituan connected with "China's AI first stock" create a strong demonstration — as long as selected by top-tier funds, valuations and stock prices can be amplified in multiple markets and stages; for subsequent capital, this is a clear signal: in the context of global AI demand and infrastructure upgrades, local projects in China are fully capable of telling an international wealth creation story in the Hong Kong market. This wealth creation model built jointly by top projects and leading investors is profoundly changing the capital logic and participant expectations in China's AI innovation ecosystem.

Chip Stocks and SpaceX Drop Together: Wall Street Presses the Brake on AI

On June 23, before the opening of the U.S. stock market, the camera pans to the other side of the ocean: NASDAQ 100 futures fell about 2.5%, Intel fell about 7.6% before the opening, AMD fell more than 6%, Qualcomm fell about 5.4%, and technology and chip-related stocks like TSMC, Baidu, Broadcom, and NVIDIA also generally declined. As core suppliers of large model computing power and communication infrastructure, this series of names is under pressure, lacking clear single adverse messages, appearing more like a repricing of the overall risk for technology and AI assets — Wall Street chose to discount the previously optimistic expectations with real prices before the opening.

At the same time, another iconic story was also pressed on the pause button: over the previous three trading days, SpaceX had accumulated a decline of about 23%, and on June 23, it fell further by about 2.3% before the opening. This company, once seen as a symbol of "human ascent to the sky, satellites weaving a net," has left behind only a steep downward curve in the short-term capital market picture. Placing this scene next to the continuous valuation and price amplification of Zhihu in the Hong Kong stock market reveals a stark contrast: on one side, Chinese investors are willing to pay higher valuation premiums for the long-term imagination of AI under the stimulus of wealth creation models; on the other side, American investors, through concentrated selling pressure, remind all technology narratives that they must endure periodic retracements and validations. Standing at the timeline of 2026, the same long-term main line of AI has been interpreted into two different risk curves with different rhythms and tolerances in the Chinese and American markets.

DRAM Shortages and UFS 5.0: Accelerating AI Devices

While the emotions of the capital market ebb and flow, another firmer curve is being set straight at the deeper level of the supply chain. TrendForce's latest tracking shows that the DRAM shortage is no longer just a tight situation for several major types used in servers and mainstream PCs, but has spread even to older memory generations like DDR2—merchandise once viewed as "inventory burdens" in warehouses is now being cleaned out by system manufacturers and spare parts channels. More critically, this institution has provided a timeline: by Q3 2026, DRAM prices will continue to rise. This means that while stock prices can plummet within a day, true computing power and data demand are steadily moving up, pushing the often-overlooked storage segment into a state of full tension.

This overall tension is not simply a periodic replenishing cycle but is structurally driven by AI and various data-intensive applications. Large model inference requires high-bandwidth memory support, and multimodal content production and real-time interaction keep the memory usage of end devices consistently high. Cloud, edge, and client devices act like three synchronously rising reservoirs, forcing DRAM manufacturers to allocate production capacity based on longer demand curves, regardless of stock prices. This is echoed by Samsung Electronics' announcement of the UFS 5.0 flash solution: with sequential read speeds of about 10.8GB/s and sequential write speeds of about 9.5GB/s, this represents more than double the performance of UFS 4.1, with plans to begin mass production in Q4 2026. For mobile phones, tablets, and various edge AI devices, this is not just a parameter upgrade but transforms storage from a bottleneck that "cannot keep up with the model" into an accelerator capable of carrying larger models and higher concurrent inferences, further illustrating that even with fluctuating stock prices, the hardware infrastructure upgrades surrounding AI are steadily advancing.

Token Consumption Expansion: Low-Cost Models Boosting China's Computing Power

If storage and chips are the tracks laid on the ground, then Jefferies’ perspective of "China's AI entering a period of Token consumption expansion" is where the real vehicle starts running on the track. The report states that "low-cost models are reshaping the competitive landscape," meaning very directly: as local large models continuously lower prices and access thresholds, competition is no longer just about model rankings and funding stories, but rather who can make it more frequent and natural for users to integrate models into their daily work and life. Increased invocation frequency, longer context, long documents, and multi-turn dialogues — these seemingly minute changes in usage habits technically point to one thing: the sustained rise in Token consumption.

This expansion is not an abstract judgment; OpenRouter's data provides a snapshot: as of the week before June 22, 2026, Token consumption increased by about 4.7% month-on-month. For a platform with a considerable base already, a single-week increase approaching 5% is enough to outline the contours of an expansion phase: it is not a spike-like short-term event, but a gradually increasing slope along higher usage frequency and thicker context. Of course, the brief also reminds that this is just a sample from OpenRouter, subject to platform structure and user profile deviations, and cannot be simply extrapolated as an accurate scale for the entire Chinese market. However, as a trend signal, it still indicates that users are becoming accustomed to the new normal of "using a bit more" and "speaking a bit more detailed."

At the downstream of the technology stack, such a growth in Token consumption will directly transmit to cloud computing or storage demands. Every model call must be completed on a GPU, and each additional segment of context or retained chat record occupies space in memory and on disks; as businesses and individuals in the Chinese market continue to increase their usage of local low-cost models, cloud service providers and data centers will have to expand their computing clusters, add high-bandwidth memory, and faster storage to accommodate the accumulating flow of Tokens. Thus, beyond the extreme amplification of Zhihu's stock price and the drastic fluctuations in Hong Kong's market sentiment, another main line quietly forms: the physical chain among workloads, Tokens, and computing power is extending more steadily and stubbornly than the stock price curve; from Jefferies' judgment to OpenRouter's growth data, the so-called Token consumption expansion period is shifting the competitive focus of China's AI from storytelling and market capitalization towards genuine computing and storage capacity.

Behind the Wealth Creation Myth: Long-Term AI Track and Short-Term Volatility

From Zhihu 02513.HK's rapid completion of its rise from the angel round to the IPO and then the secondary market, delivering roughly an 85 times return to 57 external investors, to the pre-opening decline of about 2.5% in NASDAQ 100 futures on June 23, 2026, placing pressure on tech assets like Intel, AMD, Qualcomm, and SpaceX, alongside the DRAM shortage spreading to older memory types like DDR2, the expectation of price increases in Q3, and Samsung planning to begin mass-producing the UFS 5.0 with double the read-write performance in the fourth quarter of this year, narratives and reality tug at the same timeline: on one side is the valuation myth of "China's AI first stock" and the surging sentiment in the Hong Kong stock market, while on the other side is Wall Street's severe repricing of tech assets. Meanwhile, the underlying hardware and Token consumption data silently prove that this remains a long-term track of rising demand and ongoing infrastructure expansion, though the price curve is more emotional than the industry upgrade. Jefferies' report and OpenRouter's approximately 4.7% month-on-month increase in Tokens as of June 22 hint that Chinese AI applications have entered a consumption period of "calculating more and storing heavier," but the rhythm of storytelling in the primary market, the impulse of chasing highs in the secondary market, and investors' understanding of risk premiums vary across regions, causing wealth effects and deep corrections to occur simultaneously within the same week. What truly needs to be closely monitored is not the birth of the next “first stock,” but several calmer clues: when will Chinese AI companies begin to systematically realize profits, how will the sentiment of U.S. tech stocks recover from severe adjustments, whether the supply-demand dynamics of DRAM and UFS will gradually ease, and whether the current Token consumption expansion can evolve into sustainable business growth; these are key coordinates for assessing the transition of AI from a wealth creation narrative to a sustainable industrial cycle.

Join our community, let's discuss together and become stronger!
Exclusive Hyperliquid benefits for AiCoin: https://app.hyperliquid.xyz/join/AICOIN88
Exclusive Aster benefits for AiCoin: https://www.asterdex.com/zh-CN/referral/9C50e2
On-chain Telegram community: https://t.me/AiCoinWhaleData
On-chain community: https://www.aicoin.com/link/chat?cid=N6OVMor5g
AiCoin's on-chain Twitter: https://x.com/aicoinwhaledata

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。