Sales Nexus CRM

Huawei Unveils First NPO Optical Engine with Built-in Light Source, Paving Way for Greener AI Computing

By Advos
Huawei's new Hi-ONE optical engine, the industry's first NPO optical engine with a built-in light source, promises to slash power consumption and latency in AI data centers while driving industry standards.
Huawei Unveils First NPO Optical Engine with Built-in Light Source, Paving Way for Greener AI Computing

At HUAWEI CONNECT 2026, Huawei introduced the industry's first near-packaged optics (NPO) optical engine with a built-in light source, named Hi-ONE, capable of a total capacity of 7.2 TB/s. The company says it is the first to mass-produce such an optical engine, marking a significant step in high-speed optical interconnect technology for AI computing infrastructure.

The Hi-ONE engine has already been deployed in the Ascend 960 SuperPoD, which is built on a "UnifiedBus plus Hi-ONE" architecture. According to Huawei, this integration has reduced the number of 800G optical modules within the SuperPoD by 48,000, significantly lowering system power consumption and enhancing reliability. The announcement was made via Media OutReach Newswire.

NPO technology keeps optical engines close to switching chips, shortening the distance between optical and electrical signals while maintaining the independence of the optical engine. This approach balances innovation with ecosystem continuity, allowing customers to preserve existing usage habits. Hi-ONE is the first NPO product to adopt a built-in light source architecture, achieving high-density, high-speed optical interconnect while reducing power consumption and latency.

Huawei reports that Hi-ONE integrates 36 channels of optical transceivers, reducing discrete components and lowering failure rates. By incorporating compound optical chips, silicon photonic chips, and optoelectronic RF chips, and through a balanced design of optical, mechanical, electrical, magnetic, and thermal elements, the engine achieves a transmission capacity of 7.2 TB/s on a single optical engine. This results in a tenfold improvement in overall reliability, ensuring stable operation for long-duration training tasks.

Power efficiency is another key advantage. With a linear direct-drive architecture, Hi-ONE eliminates the need for high-power oDSP chips, cutting power consumption by approximately 66% and latency by about 90%. This reduces energy costs and boosts communications efficiency, addressing a critical challenge for AI data centers as demand for computing power continues to surge.

Beyond product innovation, Huawei is actively contributing to industry standards. In May 2026, a standard proposal for a "12.8 TB/s NPO Module," jointly submitted by the China Academy of Information and Communications Technology (CAICT), Huawei, and other companies, was officially initiated by the Optical Internetworking Forum (OIF). The proposal has received support from more than 40 global industry chain vendors, signaling broad consensus on the direction of NPO technology.

The launch of Hi-ONE marks the beginning of NPO technology's transition from industry consensus to large-scale commercial use. For the AI computing industry, this could mean more efficient, reliable, and sustainable data center operations. As AI models grow in size and complexity, the demand for high-speed, low-power interconnects becomes ever more pressing. Huawei's advancements, coupled with international standards efforts, may accelerate the adoption of NPO solutions worldwide, shaping a new landscape for AI infrastructure.

Advos

Advos

@advos