News

Optical computing has reached a turning point in commercialization, and Chinese companies have begun to create their own “hardware lottery.”

3 min read
Source: zhidx.com
Wisdom and Things Author | Chen Junda Editor | Mo Ying Wisdom and Things reported on July 27 that at WAIC 2026, which concluded last week, optical computing became one of the new technology directions that attracted much attention. However, unlike previous years that focused on demonstrating technological breakthroughs, this year the focus of industry discussions has shifted to commercial implementation. At the meeting, Xizhi Technology, the leading domestic optical and electrical hybrid computing power company, held the first special forum in WAIC history that focused entirely on AI optical computing power. The optical computing part began to focus on more practical issues: which scenarios are suitable for the first deployment of optical computing, which industries need optical computing the most, and how to promote large-scale implementation. Outside the forum, Xizhi Technology demonstrated for the first time the world's first all-in-one optoelectronic computing machine "Tianshu Optical Cube" and built a real-time visual perception system based on this product for the audience to experience at the WAIC site. This system has also completed a "stress test" in Xizhi's own office building. "Tianshu Light Cube" is used to drive access control systems, run hundreds of target detection tasks every day, and accurately complete image recognition tasks in milliseconds. ▲ "Tianshu Optical Cube" This is the first time that optical computing has entered a real high-frequency civilian scenario and achieved long-term stable operation. It also means that optical computing has begun to have a business case for sustainable operation, verifiable effects, and replicable promotion. It is also worth noting that Dr. Shen Yichen, founder and chairman of Xizhi, revealed that Xizhi’s latest generation PACE 3 optical computing chip was successfully produced in June this year. The photon matrix size of this generation of chip exceeds 256×256, which is mainly designed for low-latency large-model inference design and is expected to unlock richer usage scenarios. From discussing "whether optical computing can be done" to thinking about "where optical computing should be implemented first", this year's WAIC released a clear signal: optical computing is moving from verifying technical feasibility to a critical stage of verifying commercial feasibility, and the industrialization window is opening. 1. The theoretical advantages are significant, but what are the obstacles in the commercialization of optical computing? In the past few years, as silicon photonics technology has gradually matured and the demand for AI computing power has continued to explode, optical computing has once again become an important direction in the field of global AI computing power. Photons have high bandwidth,