{
  "name": "China Domestic AI Chip Scoreboard",
  "url": "https://chinaaidispatch.com/trackers/domestic-chips/",
  "publisher": "China AI Dispatch",
  "datePublished": "2026-07-16",
  "dateModified": "2026-07-18",
  "asOf": "2026-07-18",
  "description": "A sourced scoreboard of China’s domestic AI accelerators: maker, chip, fabrication node, production status and the key public numbers, each row attributed to a named source.",
  "note": "Process nodes are outside estimates, not maker disclosures. Contested figures are hedged in place.",
  "rowCount": 14,
  "rows": [
    {
      "id": "huawei-ascend-950pr",
      "chip": "Ascend 950PR",
      "maker": "Huawei",
      "kind": "Training + inference",
      "node": "Not disclosed; SMIC-class domestic process (est.)",
      "status": "Announced Mar 2026; volume ramp targeted 2026",
      "numbers": "Claims 2.87x the single-card compute of Nvidia’s China-market H20; first China chip to support 4-bit float (FP4); pairs with Huawei’s own HiBL 1.0 memory (112GB, 1.4 TB/s). ByteDance and Alibaba plan orders; Huawei targets 750,000 units in 2026. About 812,000 Ascend cards shipped in 2025.",
      "source": "TrendForce; Reuters; IDC",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#huawei-ascend-950pr",
      "companyUrl": "https://chinaaidispatch.com/companies/huawei-ascend/",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-co-design"
    },
    {
      "id": "huawei-ascend-950dt",
      "chip": "Ascend 950DT",
      "maker": "Huawei",
      "kind": "Inference",
      "node": "Not disclosed; SMIC-class domestic process (est.)",
      "status": "On the 2026 roadmap (Q4 2026)",
      "numbers": "144GB of in-house HiZQ memory at 4 TB/s. A SemiAnalysis teardown finds part of DeepSeek V4’s architecture was co-designed for it. Series codenamed \"David\"; roadmap runs 960 in 2027, 970 in 2028.",
      "source": "SemiAnalysis via InfoQ; Reuters",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#huawei-ascend-950dt",
      "companyUrl": "https://chinaaidispatch.com/companies/huawei-ascend/",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-co-design"
    },
    {
      "id": "huawei-ascend-910c",
      "chip": "Ascend 910C",
      "maker": "Huawei",
      "kind": "Training + inference",
      "node": "SMIC advanced line (~7nm-class, DUV, est.)",
      "status": "In volume; deployed in national clusters",
      "numbers": "Shenzhen’s 10,000-card cluster runs entirely on 910C, rated 11,000 petaflops with ~92% pre-booked. A third-party full-parameter post-training of DeepSeek V4-Pro (1.6T params) ran on ~1,000 910C cards; hardware utilization was 34.9% vs 50 to 55% on Nvidia.",
      "source": "SCMP; Leiphone",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#huawei-ascend-910c",
      "companyUrl": "https://chinaaidispatch.com/companies/huawei-ascend/",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/closed-in-code"
    },
    {
      "id": "cambricon-accelerators",
      "chip": "AI accelerators",
      "maker": "Cambricon",
      "kind": "AI accelerator",
      "node": "SMIC N+2 (~7nm-class, est.)",
      "status": "In production; publicly listed",
      "numbers": "First pure-play AI chip designer past 1 trillion yuan market cap. Shipped ~116,000 cards in 2025, a 2.9% share of the ~4M China bought, against Nvidia’s 55%. Holds an estimated 9 to 11% of SMIC’s advanced N+2 capacity.",
      "source": "IDC; tmtpost",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#cambricon-accelerators",
      "companyUrl": "https://chinaaidispatch.com/companies/cambricon/",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-harder-half"
    },
    {
      "id": "alibaba-thead-m890",
      "chip": "T-Head M890 (Zhenwu)",
      "maker": "Alibaba (PingTouGe)",
      "kind": "AI accelerator",
      "node": "Domestic process; node not disclosed",
      "status": "In production; deployed in-house",
      "numbers": "144GB HBM, 128-chip supernode. About 560,000 T-Head chips shipped. Alibaba’s own 10,000-chip data center runs on earlier Zhenwu-line parts.",
      "source": "Reuters; CAD reporting",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#alibaba-thead-m890"
    },
    {
      "id": "moore-threads-s5000",
      "chip": "MTT S5000",
      "maker": "Moore Threads",
      "kind": "General-purpose GPU",
      "node": "Domestic process; node not disclosed",
      "status": "In volume; listed on Shanghai’s STAR Market",
      "numbers": "Trained a 27B coding model end-to-end on its own cluster; its GPU backend was merged into the mainline SGLang serving framework. Contests SMIC N+2 capacity with Biren and Alibaba’s T-Head. Pre-announced first-half 2026 revenue of 1.65 to 1.75 billion yuan, up 135 to 149% year on year from about 702 million yuan; the company did not disclose how much of its output is on advanced versus mature process nodes.",
      "source": "Moore Threads STAR Market pre-announcement via Sina Finance; tmtpost; CAD reporting",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#moore-threads-s5000",
      "companyUrl": "https://chinaaidispatch.com/companies/moore-threads/",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-toolmakers-cut"
    },
    {
      "id": "hygon-dcu",
      "chip": "DCU accelerators",
      "maker": "Hygon (海光信息)",
      "kind": "AI accelerator (x86-and-accelerator house)",
      "node": "Domestic process; node not disclosed",
      "status": "In volume; publicly listed on Shanghai’s STAR Market",
      "numbers": "Supplies a large share of China’s domestic AI servers. Pre-announced first-half 2026 revenue of 8.5 to 9.3 billion yuan, up 55.56 to 70.20% year on year, with net profit up 41.5 to 52.3%. The company did not disclose the advanced-versus-mature process-node split behind that output.",
      "source": "Hygon STAR Market earnings pre-announcement via Sina Finance",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#hygon-dcu",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-toolmakers-cut"
    },
    {
      "id": "baidu-kunlunxin",
      "chip": "Kunlunxin",
      "maker": "Baidu",
      "kind": "AI accelerator",
      "node": "Domestic process; node not disclosed",
      "status": "In production; being spun off for listing",
      "numbers": "Being spun off for a Hong Kong listing at a target near 50 billion dollars; Tencent is a customer.",
      "source": "CAD reporting",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#baidu-kunlunxin"
    },
    {
      "id": "iluvatar-corex",
      "chip": "CoreX (Tianshu Zhixin)",
      "maker": "Iluvatar CoreX",
      "kind": "Inference GPU",
      "node": "Domestic process; node not disclosed",
      "status": "In production; scaling with anchor buyer",
      "numbers": "ByteDance is finalizing a deal for at least 50,000 inference chips, which would make Iluvatar its third domestic GPU supplier behind Huawei and Cambricon.",
      "source": "Reuters; SCMP",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#iluvatar-corex",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-third-supplier"
    },
    {
      "id": "enflame",
      "chip": "Enflame",
      "maker": "Enflame (Tencent-backed)",
      "kind": "AI accelerator",
      "node": "Domestic process; node not disclosed",
      "status": "In production; STAR Market IPO cleared",
      "numbers": "Cleared a STAR Market IPO targeting about 6 billion yuan; Tencent owns roughly 20%.",
      "source": "CAD reporting",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#enflame"
    },
    {
      "id": "biren",
      "chip": "Biren",
      "maker": "Biren",
      "kind": "General-purpose GPU",
      "node": "SMIC N+2 (~7nm-class, est.)",
      "status": "In volume production",
      "numbers": "In volume production; a named member of the domestic-AI-infrastructure alliance SenseTime convened at WAIC. Contests SMIC N+2 capacity with Moore Threads and T-Head.",
      "source": "tmtpost; CAD reporting",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#biren",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-doorman"
    },
    {
      "id": "metax",
      "chip": "MetaX",
      "maker": "MetaX",
      "kind": "General-purpose GPU",
      "node": "Domestic process; node not disclosed",
      "status": "In production; domestic IPO track",
      "numbers": "In production and on the domestic IPO track; a named member of the SenseTime-convened domestic-compute alliance alongside Cambricon, Hygon, Moore Threads and Biren.",
      "source": "CAD reporting",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#metax",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-doorman"
    },
    {
      "id": "sugon",
      "chip": "Sugon accelerators",
      "maker": "Sugon (CAS-affiliated)",
      "kind": "HPC / AI accelerator",
      "node": "Domestic process; node not disclosed",
      "status": "Deployed in a national computing hub",
      "numbers": "60,000 units run the Zhengzhou national computing hub with zero chips from Nvidia, AMD or any Western supplier.",
      "source": "CCTV",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#sugon",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/60000-chips-zero-from-nvidia"
    },
    {
      "id": "deepseek-inference",
      "chip": "DeepSeek inference chip (in development)",
      "maker": "DeepSeek",
      "kind": "Inference (in-house, unannounced)",
      "node": "Unknown",
      "status": "Reported in development; not shipping",
      "numbers": "Reported to be designing its own chip aimed at inference, not training, to own the recurring inference bill rather than out-Nvidia Nvidia. Reuters reports it is talking to chip-design firms, foundries and memory suppliers. No specifications disclosed.",
      "source": "Reuters (7 Jul 2026); The Information via CAD",
      "url": "https://chinaaidispatch.com/trackers/domestic-chips/#deepseek-inference",
      "companyUrl": "https://chinaaidispatch.com/companies/deepseek/",
      "cadIssueUrl": "https://chinaaidispatch.substack.com/p/the-convert"
    }
  ]
}