Hot on the heels of outlining how it has joined technology peers to develop interconnect technology designed for high-speed connectivity based on Near-Packaged Optics (NPO), Huawei has unveiled a series of announcements aiming to establish an artificial intelligence (AI) computing architecture that will form the base of a series of AI infrastructure products and clusters.
Huawei said that “the rapid approach of an intelligent world” is driving up demand for computing power. As a result, Huawei’s focus will be on developing AI infrastructure based on systems and architecture centred around super points of delivery (SuperPoDs) and SuperClusters to build a silicon foundation for a future intelligent world.
In making its announcements at its Connect 2026 conference, Huawei noted that AI is sweeping the world faster than any previous technological revolution. It calculates that foundation model parameters are rapidly approaching 10 trillion and are projected to exceed 100 trillion by 2030. While current AI agents can now work on the same task continuously for hours on end, it is likely that they will be able to handle tasks that span months by 2030.
Huawei added that in China alone, the average number of inference tokens consumed every day has surged to around 500 trillion and is expected to reach quintillions by 2030. On-device AI is also advancing rapidly, with those for smartphones expanding from three billion parameters in 2024 to 30 billion by the end of 2026 and will push toward hundreds of billions in the near future.
At the heart of the new infrastructure and platforms is the Peerium Computing Architecture, which represents “a computing architecture for the AI era”. Peerium is constructed to enable processors at the million scale to work as one computer, meeting an ever-growing demand for AI compute.
The architecture achieves scaling to the million-processor level through nested parallelism, unified memory addressing and peer interconnect. It introduces Nested BSP (Nested Bulk Synchronous Parallel), extends the von Neumann single-machine architecture and is said to overturn the existing master-slave architecture so that a million processors can act as one larger computer.
UnifiedBus (UB) is cited the key interconnect technology that makes the Peerium Computing Architecture possible. Built on a single open protocol, UB is a high-speed bus that is claimed to be able to scales without limit to connect CPUs, NPUs, memory, SSDs, network interface cards (NICs) and switches. With UB, peer interconnect is achieved across compute, storage and networking.
The Atlas 950 SuperPoD and SuperPoD-based SuperClusters will be the first-generation product built on the Peerium Computing Architecture. An Atlas 950 SuperCluster with 256,000 cards is already being deployed, and the Atlas 960 system based on NPO) is currently under testing.
Huawei said that it was dedicated to building next-generation communications networks to bring readily available compute and intelligence to every person, home and organisation.
Commenting on network developments, David Wang, deputy chairman of the board and rotating chairman, at Huawei said: “We have developed a next-generation optical interconnect product based on NPO, the High-density Optical-interconnect-Node Engine [Hi-ONE]…
“[This] has a multi-physics design for balancing optical, mechanical, electrical, electromagnetic and thermal performance, realising a transmission capacity of 7.2 Tbps per single engine. This is the industry’s first NPO product ready for mass production, delivering the largest transmission capacity. It is also the industry’s first NPO product with a built-in light source.”
Earlier in September 2026, Huawei submitted an implementation agreement (IA) on NPO to the Optical Internetworking Forum (OIF) standards organisation. The company said that it would continue its efforts to further refine the NPO industry ecosystem.
