Amid the technology sector, the Hong Kong stock market's enthusiasm in 2026 is self-evident, with AI large models, computing power chips, intelligent driving, AI pharmaceuticals and other sub-sectors successively welcoming representative targets. In this wave of hard tech listings, XIZHI TECH-P (01879), with its rare positioning as the "world's first AI silicon photonics chip stock," has become one of the most recognizable targets. The eye-catching effect arrives fiercely, but what truly determines how far a frontier tech stock can go is never the initial listing gains, but whether its technological positioning can be gradually verified in subsequent operating cycles. For XIZHI TECH, the price elasticity in the early listing period has given way to a calmer examination amid valuation digestion: the market's focus is shifting from "story pricing" to "realization pricing."
After a stunning opening: valuation digestion and shareholder structure rebalancing. XIZHI TECH's listing debut can be called the most extreme sample of 2026 Hong Kong IPO newcomers. On April 28, the first day of listing, XIZHI TECH's share price surged as much as 445%, still closing with a gain of approximately 383.62%. On the sixth trading day after listing, the share price hit a historical high of HK$1,050, up more than 473% from the issue price, with total market capitalization exceeding HK$90 billion. However, the company's share price subsequently entered a sustained digestion channel, with the lowest price dropping to HK$228.40, a pullback of nearly 80% from the peak. This trajectory is not difficult to understand: the initial public offering received over 5,785 times subscription, combined with cornerstone and international placement locking up the vast majority of circulating shares, leaving an extremely small actual tradable float. The "thin float" characteristic naturally amplified price elasticity. And when multiple narratives such as AI optical interconnect, domestic computing power, and optical computing simultaneously pressed onto a single scarce target, capital quickly gave "story pricing." But on the underlying fundamentals, XIZHI TECH is still in the early stage of large-scale commercialization, with revenue scale unable to fully support high valuations. The share price has also declined from its high to consolidate around HK$300 per share. This is not the result of a single negative factor, but a common valuation digestion process for high-volatility new stocks.
On October 28, XIZHI TECH-P (01879) will welcome the unlocking of cornerstone investors after its Hong Kong listing. In terms of total volume, 20 cornerstone investors collectively hold approximately 8.9746 million shares of XIZHI TECH (accounting for 9.54% of total share capital after exercise of the over-allotment option), constituting a major test for the company's short-term share price performance. Prior to the unlock, the company's share price was affected by the market and has already experienced further decline. In fact, for listed companies, share price fluctuations before unlocking are normal, reflecting that market capital has already begun gaming in advance. Moreover, the scale of the unlock does not equal actual selling pressure; the real impact still depends on shareholder attributes and exit methods. From the structure of XIZHI TECH's cornerstone investors, they are mainly divided into four categories: sovereign funds, long-term asset managers, industrial capital, and top-tier PE. Among them, top-tier PE mainly makes investment decisions based on comprehensive holding costs, the company's commercialization progress, and track value. The remaining three types of cornerstone investors tend to have longer-term investment intentions. For example, sovereign funds and global long-term asset managers such as GIC, Temasek, BlackRock, Fidelity, and Baillie Gifford usually have assessment cycles measured in years, focusing more on track space, technological position, and medium-to-long-term realization, rather than short-term price differentials after unlocking. Similarly, computing power industry chain capital such as Alibaba, ZTE, and Lenovo has investment logic rooted more in technological synergy, ecosystem positioning, and supply chain security, with investment cycles significantly longer than pure financial arbitrage funds.
Short-term focus on chips, medium-to-long-term focus on three certainty mainlines. The cornerstone group, with fully gamed capital in the earlier period and stronger long-term holding willingness, means that XIZHI TECH's unlocking this time is more like a "stress test" at the level of chip structure and liquidity, rather than an "inflection point" in the company's long-term value curve. In fact, at the fundamental level, it has already formed an inverse contrast with short-term share price fluctuations. In the first half of 2026, XIZHI TECH's revenue increased 284% year-on-year to 79.78 million yuan, with optical interconnect revenue growing 275% and optical computing revenue growing 313%. The adjusted net loss of 147 million yuan was basically flat compared to the same period last year — high revenue growth without a simultaneous increase in losses indicates that the company is crossing the early-to-mid-stage technology verification phase, and the improvement in operating leverage from large-scale commercial delivery has begun to emerge. For instance, CICC gave XIZHI TECH a target price of HK$362 before the unlock, further raising it to HK$426 in August, maintaining an "outperform industry"/"overweight" rating. This institutional judgment will certainly not consider short-term market disturbances before and after the unlock, but rather anchor on the company's medium-to-long-term revenue scale and track positioning.
Specifically at the business level, XIZHI TECH has also left three major logics: The first is optical interconnect, which has the highest certainty and is closest to revenue explosion. Currently, AI clusters continue to Scale-up, and the bottleneck is shifting from single-card computing power to the power consumption and latency of moving data between chips and between racks. Electrical interconnect's physical ceiling at higher speeds, shorter distances, and larger expansion domains is becoming increasingly clear. "Replacing copper with light" is no longer one of the optional technology routes, but a direction forced by the system's own constraints. In this direction, XIZHI TECH has mastered a complete technology route layout from EPS, distributed OCS to NPO, and CPO, with commercialization progress in segmented fields also at the forefront of the industry. From the commercialization rhythm, the company adopts a layered layout of "current volume ramp + long-term positioning": NPO undertakes current industrial demand that can be scaled up, while CPO positions for a more long-term optical interconnect form. Currently, although NVIDIA has proven the CPO route through mass production delivery, NPO, with its faster commercialization progress and small disruption to the industry chain ecosystem, has become a form with high engineering adaptation maturity and scalable volume in current domestic GPU super-node scenarios. In this regard, the company's 3.2T NPO product has passed testing in end-customer scenarios and completed tape-out, while 6.4T NPO is being jointly developed with customers. In more long-term technology options, the company's CPO has also completed prototype engineering verification. CPO requires deep collaboration with main chip manufacturers and is regarded by the market as a more long-term, more future-oriented option. The independence of NPO today comes from its realistic adaptability under current industry constraints, rather than a stopgap measure of "never needing CPO."
The second mainline is optical computing. It still needs time before large-scale commercialization, but the certainty of long-term market space is precisely the thickest among the three mainlines. What optical computing needs to solve is not the cost of data movement, but the energy efficiency and latency of matrix operations themselves; if inference loads continue to grow, low-latency, low-energy scenarios will form long-term demand. Therefore, in addition to optical interconnect, optical computing is also a high-certainty technology route for computing power in the post-Moore era. Optical interconnect solves the cost of data "movement," while optical computing needs to solve the energy efficiency and latency of matrix operations themselves. The continuous expansion of large model inference loads is turning low latency and low power consumption from "optimization items" into "constraint items." Traditional silicon-based electronic chips, after entering the 1-nanometer node, are constrained by quantum tunneling effects, encountering physical bottlenecks in performance improvement, and Moore's Law tends to fail. Photons naturally possess the physical advantages of high parallelism, low latency, and low power consumption. Using photons instead of electrons to complete AI core operations such as matrix multiplication is one of the few paths in the post-Moore era that can bypass this constraint from the physical level. According to Frost & Sullivan data, China's optical computing product market is expected to grow from 63.7 million yuan in 2025 to 1.462 billion yuan in 2030, with a compound annual growth rate of 87.2%; and from 2031 to 2036, the market is expected to further expand from 2.546 billion yuan to 34.759 billion yuan, with a CAGR still reaching 68.7%. XIZHI TECH currently has launched the PACE series of optoelectronic hybrid computing chips, being one of the few photonic computing hardware products that use silicon photonics for computing, complete productization, and generate sales revenue. Currently, the 256×256 matrix has advanced to the tape-out stage. In the first half of the year, this business revenue was less than 20 million yuan, with absolute scale still small, but its strategic significance lies in synergy and positioning: the two complete narratives of optical interconnect and optical computing mean that if future optical computing chips need to cooperate with cluster interconnect, XIZHI TECH has the innate conditions for system-level synergy. Once low-latency inference forms a reusable software stack and interface, first movers have the opportunity to upgrade from "selling chips" to "selling reference architectures," with positioning value far higher than single-device sales. It should be noted that the option value of optical computing is not an indefinitely deferred narrative: PACE 3 (256×256 matrix)'s PIC and EIC have both completed tape-out, and the industry is expected to enter a true market penetration period in 2027 — this constitutes the first time anchor for optical computing moving from "long-term narrative" to "verifiable variable," undertaking a market with greater long-term space than optical interconnect.
The third mainline is OCS, which XIZHI TECH places within the optical interconnect segment and is a direction relatively easily underestimated. OCS does not perform or performs little photoelectric conversion, directly scheduling paths in the optical domain, naturally suitable for high-traffic, relatively stable path super-node scenarios. In the earlier period, XIZHI TECH has built the "LightSphere X" distributed optical circuit switching super-node and completed the deployment of a thousands-card-level GPU cluster, adapting to mainstream domestic large models and domestic GPU platforms, and jointly promoting domestic super-node solutions with partners such as ZTE and Biren. This also means that XIZHI TECH is no longer selling a single optical switching device, but a "super-node optical switching solution." Once the networking topology is deployed, migration costs are high and customer stickiness is strong, forming a system-level binding with domestic GPUs and cloud vendors, rather than single-point procurement. From selling devices to selling architecture, the corresponding valuation coordinate system is inherently different.
Returning to the unlocking event itself. Short-term sentiment disturbances do exist, and the release of 9.54% of chips needs to be digested by the market through trading. But the unlocking also increases the free float ratio, gradually returning pricing power from the "thin float game" in the early listing period to a broader group of market participants.
Comments