Huawei launches 7.2 Tbps NPO

Technologies and Products

Huawei has announced the launch of the world’s first mass-produced Near-packaged optic (NPO) optical engine as it seeks to set the standard for next-generation AI infrastructure.

The Hi-ONE NPO delivers transmission capacity at 7.2 Tbit/s and is currently the industry’s only NPO with a built-in light source.

An NPO is an architectural design where light-based data converters are placed right next to a main computer chip on a circuit board, instead of far away at the edge of the device. This helps to reduce latency by bringing the optical connectivity closer to a processor.

At HUAWEI CONNECT 2026, the company also announced that the NPO will be used – alongside its UnifiedBus – to power the Atlas 960E SuperPoD, the industry’s first NPO-based SuperPoD.

A SuperPoD is an advanced enterprise AI infrastructure concept where multiple physical computers or processors are connected to act as a single machine.

As artificial intelligence systems continue to grow to an unprecedented scale, nearing 10 trillion parameters, they require thousands of interconnected AI supercomputing chips working together.

Huawei’s SuperPoD design approach centres on connecting all cluster components under a single protocol with peer-to-peer interconnect, supporting cross-physical-server memory access, and adopting short-reach copper alongside long-reach optics.

With unified memory addressing enabled by UnifiedBus, a single SuperPoD can scale to 4,096 NPUs, delivering 8 EFLOPS at FP8 precision and 16 EFLOPS at FP4 precision.

By integrating 5,500 Hi-ONE units, this SuperPoD doesn’t need the 48,000 800G optical modules that would traditionally be required to connect all the NPUs.

This cuts power consumption by over 550 kilowatts, while doubling the system’s fault-free operating time, achieving 99.8% system availability. All told, the system greatly accelerates innovation in training and inference for frontier models.

“At Huawei, we are committed to building powerful AI infrastructure and going open source and open system to build out computing ecosystems,” said David Wang, Deputy Chairman of the Board and Rotating Chairman at Huawei. “We will continue to support native training for mainstream foundation models on Ascend and actively embrace a wide range of models and applications. Together, let’s lead the charge into the intelligent world.”