Lightelligence installed an optoelectronic computing box of its own making at the company entrance, using it for employee facial recognition access. This move speaks more to a hard-tech company's current stage than any financial report.
On November 3, Lightelligence filed for A-share listing guidance with the Shanghai Securities Regulatory Bureau. It has been just six months since its listing on the Hong Kong Stock Exchange in April. The company raised net proceeds of HKD 2.741 billion from its HK IPO, and as of June 30, 2026, approximately HKD 2.69 billion remained largely unused. The company explained that returning to the A-share market is not merely about financing, but rather leveraging the domestic capital platform to stay closer to the industrial chain, customers, and talent, thereby driving the large-scale deployment of next-generation computing products.
From an industry perspective, this choice is unsurprising. When hard-tech companies reach the commercialization stage, they often face two challenges simultaneously: whether the technology works, and whether the industrial ecosystem can support it.
| From "Usable" to "User-Friendly"
Lightelligence's PACE series optical computing accelerator cards have iterated rapidly within the optical computing sector, progressing from 64×64 to 128×128, and now to the already taped-out 256×256 PACE 3. However, impressive chip specifications do not guarantee stable product operation at customer sites.
The "Tianshu · Optical Cube," equipped with PACE 2, has served as the company's access control system at the entrance since June 1, 2026. Public data shows that by October, it had operated cumulatively for 144 days, with recognition accuracy improving from 64.09% to 96.21%, availability reaching 99.9%, and the self-healing system automatically recovering from seven anomalies.
This set of data is more meaningful than peak metrics showcased at launch events. It demonstrates that optical computing has begun operating continuously in real-world, high-frequency, and uncontrolled environments. It also highlights that significant engineering gaps remain between laboratory usability and commercial viability: recognition rate, throughput, stability, anomaly recovery, and software adaptation all require time to resolve.
For hard-tech companies, this is not a weakness but an inevitable phase. Any new computing architecture must first prove its reliability internally before entering customer production environments.
| Optical Interconnects Secure the Present, Optical Computing Bets on the Future
Lightelligence reported revenue of CNY 79.78 million in the first half of the year, a year-on-year increase of 284%. Of this, optical interconnect revenue accounted for CNY 60.70 million, while optical computing revenue was CNY 19.08 million. This structure essentially outlines the division of labor between the two business lines.
Optical interconnects address the certain demand driven by the expansion of AI computing clusters. As training and inference cluster scales grow, pressure on bandwidth, power consumption, and latency for data transfer increases, making optical interconnects easier to find tangible procurement scenarios for. Optical computing, however, is more long-term oriented, aiming to solve energy efficiency and latency issues in AI inference, edge computing, and scientific computing scenarios, though its commercial validation remains in early stages.
Thus, Lightelligence currently appears to be "walking on two legs": one leg focuses on building revenue, customer base, and delivery capabilities, while the other opens up future possibilities for computing architectures. Both business lines share underlying capabilities such as silicon photonics design, optoelectronic packaging, and heterogeneous integration, which constitutes its platform value.
| Returning to A-Shares Is Not Just About Money
The HK market provided Lightelligence with exposure to international capital markets and completed its first round of market-based pricing. The A-share market is closer to the domestic computing industry chain. Its customers, partners, application scenarios, and talent resources are mostly located domestically.
For a hard-tech company that is not yet profitable but possesses strong technological scarcity, the A-share and HK markets do not simply offer a "better or worse" choice; instead, they provide different valuation perspectives: one side focuses more on cash flow and profitability pace, while the other has relatively higher tolerance for technical barriers, domestic substitution, and industrial synergy. If dual-listing proceeds smoothly, it will help the company connect with both international capital and domestic industrial resources.
Of course, returning to A-shares is only the beginning. Subsequent steps include acceptance, inquiries, and registration. Whether it ultimately lands depends on the pace of large-scale commercialization and performance realization.
| Optical Computing Needs More "Access Control Gates"
This year, interest in the optical computing and optical interconnect sectors has risen significantly, with increased activity in financing, M&A, and IPOs. However, what this industry lacks most is not hype, but verifiable uptime.
Optical computing is not a software product; engineering issues cannot be masked by rapid iteration. It requires end-to-end integration from chips, packaging, systems, and software to application scenarios. Access control is just the first real-world scenario; many others, such as chip design-assisted inference, edge intelligence, and low-latency inference, still need validation.
For companies like Lightelligence, the key is not how grand the story is told, but how long the product runs in real-world scenarios. Running for one day is more important than one demo; running for one year proves commercial capability better than publishing a hundred papers.
The true value of returning to A-shares may lie here: it secures a capital environment closer to the industrial chain for the company and allows the market to observe the process of optical computing moving from technical validation to large-scale commercial use over a longer cycle.
Disclaimer: This article is written based on public information, serving only as industry observation and personal opinion sharing. It does not constitute investment advice, legal opinions, or guarantees regarding company operations. Information concerning companies, finances, technology, and listing progress mentioned herein shall be subject to official disclosures by relevant companies and announcements by regulatory authorities. Markets involve risks; decisions should be made with caution. Please contact us for corrections or deletions if any content is inaccurate or involves rights issues.