What happened
Huawei Technologies unveiled a 7.2-terabit-per-second (Tbps) near-packaged optics (NPO) module at the China International Optoelectronic Exposition (CIOE) in Shenzhen. The company described the product as the industry's first of its kind, designed to replace traditional copper wiring with light-based signals for high-speed data transfer between AI processors. Huawei stated that the module combines 36 channels operating at 200 gigabits per second each. While the company is in the product development stage, it expects 3.2Tbps NPO modules to see real-world deployment this year, with 7.2Tbps versions gaining traction over the next two years.
Huawei Technologies showcased a high-speed optical module at the China International Optoelectronic Exposition (CIOE) in Shenzhen, claiming it is the industry's first 7.2-terabit-per-second (Tbps) near-packaged optics (NPO) product. The module is designed to eliminate bottlenecks in artificial intelligence infrastructure by replacing traditional copper wiring with light-based signals. According to Man Jiangwei, director of Huawei’s advanced optoelectronics laboratory, the module combines 36 channels operating at 200 gigabits per second each to achieve the aggregate bandwidth.
The technology aims to address the physical limits of conventional copper connections in bandwidth, transmission distance, and power efficiency as AI clusters grow to include thousands of interlinked processors. By bringing optical components closer to network switching chips, NPO shortens electrical connections, thereby reducing signal loss and power consumption. Huawei stated that it has fully entered the product development stage and expects to move quickly into volume production once the supply chain and manufacturing system are in place.
Man Jiangwei indicated that 3.2Tbps NPO modules are expected to start seeing real-world deployment this year, with 7.2Tbps versions gaining traction over the next two years as supply chains mature. He noted that a lack of unified industry standards remains a major hurdle to commercializing NPO at scale, as component makers, equipment suppliers, and chipmakers currently use different designs. Huawei is actively stepping up efforts to shape these emerging specifications, noting that strong industry participation is accelerating the standard-setting process, which typically takes 18 to 24 months.
Why it matters
As AI clusters expand to include thousands of interlinked processors, conventional copper connections are reaching physical limits in bandwidth, transmission distance, and power efficiency. This optical module addresses these bottlenecks by bringing optical components closer to network switching chips, which reduces signal loss and power consumption. The development is significant for the AI infrastructure sector because it offers a more energy-efficient alternative to current interconnect standards. Additionally, Huawei's move highlights the competitive landscape in optical interconnects, where Chinese manufacturers already dominate the global supply chain, supplying nearly two-thirds of global optical transceivers by volume. The technology could influence emerging industry standards for next-generation data center connections, potentially shifting the balance of power in AI hardware infrastructure.
The shift from copper to optical interconnects is critical for the scalability and energy efficiency of large-scale AI training and inference clusters. As AI models require more data movement between processors, the power consumption and heat generation of copper wiring become significant operational costs and technical constraints. Huawei's NPO module offers a pathway to mitigate these issues, potentially lowering the total cost of ownership for AI data centers.
The development also underscores the dominance of Chinese manufacturers in the global optical transceiver supply chain. According to Counterpoint Research data cited in the report, domestic firms supplied nearly two-thirds of global optical transceivers by volume and accounted for around 60 percent of optical data communications transceiver revenue in 2025. This market position allows Chinese companies like Huawei to influence the direction of next-generation AI infrastructure standards.
The competition in this space is intensifying, with US chipmaker Broadcom developing 6.4Tbps optical engines for co-packaged optics. The race to set standards for high-density optical interconnects will likely determine which technologies become the default for future AI hardware, with significant implications for global supply chains and technological sovereignty.
What to watch next
The timeline for volume production of the 7.2Tbps module and the deployment of 3.2Tbps versions in real-world AI data centers. Industry standardization efforts for near-packaged optics, as Huawei and other manufacturers work to unify specifications. Competitive responses from other optical module manufacturers, including Broadcom and Chinese rivals like InnoLight and Eoptolink, who are also developing next-generation optical engines.
Monitor the actual deployment of 3.2Tbps NPO modules in commercial AI data centers this year, as this will provide the first real-world performance data for the technology. The success of these early deployments will be a key indicator of the viability of NPO as a replacement for copper interconnects.
Track the progress of industry standardization efforts for near-packaged optics. The ability of Huawei and other stakeholders to agree on unified specifications will be crucial for the widespread adoption of the technology. Delays in standardization could fragment the market and slow down the transition to optical interconnects.
Observe competitive responses from other major players in the optical module market, including Broadcom, InnoLight, and Eoptolink. The release of competing products with similar or superior specifications could impact Huawei's market share and the pace of NPO adoption.