Chinese display maker Leyard has told investors it is cooperating with the Chinese Academy of Sciences on research into MicroLED optical modules and has already supplied prototypes intended to replace traditional optical transmission schemes. The company said the effort is still at the research-and-validation stage and carries uncertainty, asking investors to note the risks.
The push comes as demand for AI compute surges and power consumption has become a critical constraint for hyperscale data centres and chip designers. Leyard cited recent estimates from research groups that a MicroLED-based co‑packaged optics (CPO) approach could cut per‑unit transmission energy to roughly 5% of an equivalent copper-cable solution, an efficiency gain that would materially alter infrastructure power economics if realised at scale.
MicroLED is better known for next‑generation displays, but engineers and researchers have been exploring its use as a compact, highly efficient light source for short‑reach optical interconnects. A MicroLED CPO architecture would place optical transmitters much closer to switching ASICs and accelerators, reducing electrical I/O losses and enabling denser, lower‑power links that are attractive to operators building AI clusters.
Significant technical and commercial hurdles remain. MicroLED manufacturing is still a challenging, yield‑sensitive process; packaging, driver integration, thermal management and alignment tolerances for co‑packaged modules are non‑trivial; and standards and supply ecosystems for CPO are still evolving. Leyard’s disclosure underscores both the potential upside and the high uncertainty: the modules are experimental, and broader adoption would require further validation and industrialisation.
For Leyard, the project signals a strategic diversification from its core LED‑display business into photonics and datacentre hardware — an attractive adjacency if the technology proves viable. For China’s broader technology agenda, domestic development of low‑power optical interconnects aligns with national priorities to bolster indigenous capabilities in critical infrastructure for AI and cloud computing.
Commercial timelines are likely measured in years rather than months. Even with dramatic energy savings on paper, customers will demand reliability, interoperability and economies of scale that typically follow successful pilot deployments and supply‑chain maturation. The near term, therefore, is one of cautious R&D advancement rather than immediate market disruption.