Terms

Neural Processing Unit

Follow the latest coverage, related explainers and connected technology stories.

Latest coverage

CN
Huawei Reveals Its Vision for AI Infrastructure and Announces SuperPoD Simulation Results

Huawei Reveals Its Vision for AI Infrastructure and Announces SuperPoD Simulation Results

During HUAWEI CONNECT 2026, Huawei presented its vision for building AI infrastructure based on SuperPoDs and SuperClusters, noting that a simulation of a group comprising 100,000 NPUs achieved 2.75 times higher utilization efficiency compared with a traditional server architecture. It also announced the deployment of more than 1,000 Atlas 900 A3 SuperPoD systems.

CN
Google Unveils Googlebook: Laptops Designed to Work with Android Phones

Google Unveils Googlebook: Laptops Designed to Work with Android Phones

Google announced the premium Googlebook line of laptops, featuring direct integration with Android phones, Gemini tools, and hardware that includes Intel and Qualcomm processors and NPUs exceeding 45 TOPS. Preorders start at $899, with availability in the United States beginning October 4, 2026.

CN
Microsoft Reframes Its Windows AI Strategy Around Local Execution and Agents

Microsoft Reframes Its Windows AI Strategy Around Local Execution and Agents

Microsoft is seeking to reduce Windows’ reliance on direct Copilot buttons and push the system to become a platform for running AI models and agents locally on users’ devices. The concept of “unmetered intelligence” does not mean eliminating the cloud or making Copilot free without limits; rather, it means shifting some everyday inference workloads to the computer’s CPU, GPU, and NPU.

CN
Why NPUs Need Redesigning as Edge AI Shifts to LLMs

Why NPUs Need Redesigning as Edge AI Shifts to LLMs

Sharad Chol of Expedera analyzes how the transition of edge devices from traditional vision networks to LLM and VLM models is shifting the nature of the bottleneck from computational capacity to memory traffic. He outlines the role of packet-based processing in reducing external data transfers and improving model execution inside vehicles and embedded devices.