At UBBF 2026, Huawei's data communication VP Cai Jun said over 60 operators worldwide have deployed AI WAN, and Huawei's new-generation AI WAN solution has reached commercial scale, moving the industry from technical verification to full deployment.
As of October 22, ForgeKernel, an operator set fully written by AI with no manual line-by-line tuning, ranked first in both NVIDIA's SOL-ExecBench FlashInfer-Bench collection and Ascend's CANNBench A3 hardware leaderboard, with a baseline speedup of 1.835x, 25% ahead of the second place, and outperformed DeepSeek v4.1 Flash operators.
DeepSeek is working closely with Huawei on AI chip design, with Huawei engineers embedded at DeepSeek to feed large-model training needs into chip design, as part of efforts to reduce reliance on Nvidia chips.
Huawei's Security Response Center announced that vulnerability reports assisted by AI or automated tools must undergo manual verification before submission, after receiving many false positives from AI-generated reports. Accounts submitting three invalid reports will be warned, and five may be banned. Vulnerabilities in Huawei terminal tools (e.g., DevEco) will be temporarily out of scope from September 30, 2026.
In a video released on Oct 4, 2026, Huawei's Yu Chengdong said Huawei Semiconductor has designed and mass-produced 381 τ chips in areas including phones, AI, general-purpose computing, networking, and intelligent vehicles. He also introduced the "time (τ) scaling" Tao law and LogicFolding core technologies, with high-end chip transistor density expected to reach 1.4nm-equivalent levels by 2031.
On October 1, Huawei held its Mate 90 series and all-scenario new product launch event, where Richard Yu announced that the Mate 90 standard edition is equipped with the Kirin 9030 flagship τ chip. Compared with the Kirin 9020, the chip delivers a 13% CPU increase, a 54% GPU increase, and a 13% NPU increase, supports FFRT software-hardware coordinated scheduling for a 100% faster image loading speed, and includes a mobile security processor MSP supporting post-quantum cryptographic algorithms.
At the Mate 90 series launch event on October 1, Huawei's Richard Yu announced that the Mate 90 Pro Max flagship phone is powered by the Kirin 9050 Pro logic-folding τ chip, featuring a transistor density of 238 million per mm², up 28%. The Kirin 9050 Pro processor was previously released on September 7.
On October 1, Huawei held its Mate 90 series and all-scenario new product launch event, where Richard Yu announced that the Mate 90 Pro debuts with the Kirin 9035 flagship chip. Compared to the Kirin 9030 in the Mate 80 Pro 12GB version, the new chip brings an 11% CPU increase, 10% GPU increase, and 51% NPU increase, with FFRT software-hardware collaborative scheduling and 100% faster image loading.
DeepSeek has officially open-sourced infrastructure components for Huawei's Ascend computing platform, including the TileLang compiler, compute libraries, and distributed communication libraries, all corresponding to its previously open-sourced components for NVIDIA platforms.
China Mobile Beijing and China Mobile Cloud announced the deployment of Huawei's Ascend 950DT 1,024-card supernode at the Mobile Cloud Beijing node in Changping Info Harbor Data Center. The supernode delivers 1 EFLOPS (FP8) per node, with 256 TB unified memory and microsecond latency, supporting large-model training and agent development.