by XU Meihui

Shanghai-based AI photonics chipmaker Lightelligence surged more than 380% on its Hong Kong trading debut on April 28, valuing it at HK$80.9 billion and highlighting growing investor interest in next-generation computing.

The company sold about 13.8 million H shares at up to HK$183.20 each, with a 15% greenshoe option, according to its prospectus.

Founded in 2017, Lightelligence develops photonic-electronic hybrid systems, focusing on optical interconnect and optical computing. Founder and chief executive SHEN Yichen, a physics PhD from the Massachusetts Institute of Technology (MIT), was among the first to experimentally demonstrate optical computing in a 2017 paper in Nature Photonics.

Shen said traditional electronic computing is nearing its physical limits, with power and data bottlenecks becoming more pronounced. Optical systems, he said, could offer lower latency and higher bandwidth as AI workloads scale. He said the company aims to play a disruptive role in computing, likening its ambitions to reshaping the industry in the way Tesla did in autos.

The shift has gained urgency with the rise of large AI models, which are driving demand for increasingly large computing clusters. In the US, companies such as OpenAI and chipmakers including Nvidia and Advanced Micro Devices (AMD) have built ecosystems around GPU-based computing. Optical computing, by contrast, remains at an early stage, with no clear global peers.

AI infrastructure spending is surging. Research firm Gartner forecasts global semiconductor revenue will exceed US$1.3 trillion by 2026, with AI chips accounting for about 30%, while investment bank Morgan Stanley estimates cumulative AI infrastructure investment could approach US$3 trillion by 2028.

As gains in single-chip performance slow, companies are increasingly linking hundreds of processors to scale computing power. Shen said matching high-end systems such as Nvidia's NVL72 rack system could require more than 500 Chinese-developed GPUs linked into a single cluster, placing heavy demands on data bandwidth and latency.

He likened copper-based interconnects to "slow trains", compared with optical links that function more like "high-speed rail", allowing large clusters to function more efficiently as a single system.

Lightelligence's optical interconnect products have already been tested in thousand-GPU clusters, while its PACE series of photonic accelerators remains under development, with a next-generation version in the pipeline. At AWE 2026, a major consumer electronics and technology expo in Shanghai, the company and its partners unveiled a 128-GPU optical "supernode" system capable of scaling to thousands of processors, signalling early-stage commercial deployment.

Despite strong growth expectations, the company remains in an investment-heavy phase. Revenue rose from 38.2 million yuan in 2023 to 106 million yuan in 2025, while research spending reached 479 million yuan last year, leading to widening net losses.

Shen said continued investment is necessary given the technical barriers, adding that revenue growth should eventually outpace research spending as commercial adoption expands.

Industry forecasts point to rapid expansion in optical interconnect, with optical computing also expected to grow from a low base. Still, the company faces risks including shifting demand, supply chain uncertainty and reliance on a small number of customers.

Shen said the biggest near-term challenge is not competition from rival photonics firms, but entrenched reliance on traditional electronic systems.

The company's rise has benefited from Shanghai's semiconductor ecosystem, one of China's largest clusters and a key base for China's chip industry, offering access to talent, manufacturing and downstream AI applications. Shen said the proximity of chipmakers and AI firms helps speed up testing and integration, shortening the path from research to deployment.

He added that China could become an early large-scale market for optical computing, as constraints in conventional chip development push the industry toward alternative architectures.