Skip to content

Huawei News: Latest Releases & Events

In November, an Epoch AI analysis found Huawei's Horinger campus delivers about 7% of the compute of SpaceX's Colossus 2. In October, SemiAnalysis found Ascend hardware roughly five years behind Nvidia but Huawei's software at the frontier. In April, DeepSeek released V4 Pro and Flash with support for Huawei's Ascend NPUs.

All 3 Huawei entries of the AI News timeline, newest first.

2026

November 2026

2026-11-06

  • Huawei's Horinger data center: 7% of Colossus 2
    An Epoch AI analysis using satellite imagery and chip sales estimates finds that Huawei's Horinger campus in Inner Mongolia, likely one of China's largest AI data centers, runs about 105,000 Ascend 910C chips delivering roughly 81,000 H100-equivalents of compute — just 7% of SpaceX's Colossus 2, the world's largest AI data center. The report attributes the gap to Ascend chips being less efficient than Nvidia's export-controlled chips, projects the campus will add over 70% more compute (with next-gen Ascend 950 chips) by early 2028, and concludes China's compute constraints are unlikely to ease in the next two years despite abundant energy. (epoch.ai)
    SpaceX Nvidia

October 2026

2026-10-21

  • SemiAnalysis Deep Dive: Huawei-DeepSeek Symbiosis
    SemiAnalysis publishes a technical deep dive into the Huawei-DeepSeek partnership, anchored by DeepSeek's September 30 release of Ascend support across six open-source repositories (TileLang, DeepGEMM, DeepEP, TileKernels, FlashMLA, DeepSelect) plus new Ascend 950PR microbenchmarks. The analysis finds that while Ascend hardware remains roughly five years behind Nvidia, Huawei's software ecosystem is at the frontier, and DeepSeek will likely run primarily on Huawei's stack by 2027-2028 as pressure builds for a fully domestic Chinese AI ecosystem. (newsletter.semianalysis.com)
    DeepSeek Nvidia

April 2026

2026-04-24

  • DeepSeek releases V4 Pro and Flash models
    DeepSeek released DeepSeek-V4-Pro and DeepSeek-V4-Flash, two Mixture-of-Experts models featuring a 1-million token context window and optimized for long-running agentic tasks. The models utilize a hybrid attention mechanism to drastically reduce inference costs, with Simon Willison noting it offers frontier performance at a fraction of the price and The Register highlighting support for Huawei's Ascend NPUs. (huggingface.co)
    DeepSeek