Skip to main content

13 posts tagged with "Cambricon"

Cambricon AI chips and accelerators

View all tags

国产 AI 芯片 vs 国际 AI 芯片:2026 年终全面对比

· 14 min read
Industry Research Team

2026 年是 AI 加速芯片竞争分水岭式的一年。一边是 NVIDIA 用 Blackwell 产能与 Rubin 首发把「训练算力」的标尺再抬高一截,另一边是中国本土供应链在出口管制常态化下,把「可用」加速推向「好用」——昇腾 960 超节点、寒武纪思元 690、摩尔线程 MTT S5000、阿里真武 V900 等产品密集落地。本文用同一把尺子,把国产与国际主力芯片放到一张桌子上对比。

一、为什么说 2026 是分水岭​

  • 出口管制进入稳态:美国对高端训练卡的禁运已不是新闻,而是采购决策的默认前提。国产芯片的「替代」叙事从口号转为规模部署。
  • 国产份额跨越临界点:据 Counterpoint 口径,中国 AI 芯片市场国产化率从 2022 年近乎为零,升至 2025 年的约 41%,其中昇腾占国产份额约 62%。这意味着「国产 vs 国际」不再是技术爱好者话题,而是采购刚需。
  • 代际差距收窄但仍在:纸面算力上,旗舰国产卡(昇腾 910C 128GB、思元 690 196GB)在显存容量上已不输同代国际卡;但在单卡稠密算力、软件生态与集群扩展效率上,仍有一代左右的差距。

本文对比维度固定为五类可比指标:算力(FP16 稀疏口径,与站内科普一致)、显存容量与带宽、功耗与能效、制程、价格。所有数字取自本站已校验芯片数据库(252 款)与价格库(报价截至 2026-10),官方未公开的一律标注「未公布」,不作推测填充。

二、国产 AI 芯片进展​

2.1 华为昇腾:从 910C 到 960 的三级跳​

型号架构制程显存显存带宽FP16(稀疏)TDP状态
昇腾 910BDa VinciTSMC 7nm+(含 SMIC 国产化)64GB HBM2e600 GB/s320 TFLOPS310W量产中
昇腾 910CDa Vinci(双 die)中芯国际 7nm N+2128GB HBM2e3.2 TB/s800 TFLOPS600W规模部署
昇腾 950PRDa Vinci v5中芯 N+3(等效 5nm 级)128GB HiBL 1.0未公布486 TFLOPS600W已上市
昇腾 950DTDa Vinci v5中芯 N+2144GB4 TB/s547 TFLOPS500W2026 Q4 量产
昇腾 960DT全新 SIMD/SIMT未公布288GB9.6 TB/s未公布未公布2027 Q1
昇腾 960PR全新 SIMD/SIMT未公布192GB2.4 TB/s未公布未公布2027 Q3

昇腾 910C 是当前国产训练主力,CloudMatrix 384 超节点已用于 DeepSeek 等大模型训练验证。950 系列在 2026 年 HC 大会官宣:950PR 渠道报价已超 ¥8 万(年初约 ¥6 万),950DT 超 ¥25 万(较 7 月上涨 20%~50%),HBM 供给挤压是涨价主因。960 超节点公布 NPO Hi-ONE 引擎、4096 卡规模与 99.8% 可用度目标,其中 960DT 主攻训练(FP4 4 PFLOPS)、960PR 主攻推理(FP4 8 PFLOPS),是 2027 年的旗舰预期。

2.2 寒武纪思元 690:纸面规格天花板​

思元 690 采用双 die Chiplet 设计与 MLUarch 4.0 架构,中芯国际 N+2 制程、196GB HBM3、3.35 TB/s 带宽、FP16 700 TFLOPS、TDP 500W。其显存容量与带宽已对齐 H100 SXM 级别,是国产卡中少数在纸面规格上逼近国际旗舰的产品。2026 年 9 月受 HBM 供给挤压,官方指导价 ¥19 万、市场均价约 ¥17.5 万,较两月前上调 20%~30%,客户含字节、阿里、百度。

2.3 其他国产厂商​

厂商 / 型号架构制程显存FP16(稀疏)TDP备注
天数智芯 天垓 300SIMT 通用计算7nm144GB HBM3E超 1,000 TFLOPS600W2026-07 WAIC 发布,对标 Hopper
天数智芯 天垓 150ivcore11TSMC 7nm64GB HBM2e192 TFLOPS350W通用 GPU,兼容 CUDA 生态
燧原 云燧 T21GCU-CARA 2.0格罗方德 12nm64GB HBM2E134.4 TFLOPS未公布邃思 2.0,训推一体
百度 昆仑芯 R200XPU-R7nm16/32GB GDDR6128 TFLOPS150W低功耗推理向
摩尔线程 MTT S5000MUSA 4(平湖)TSMC 6nm80GB400 TFLOPS(BF16)500W全功能 GPU,渠道价 ¥7.5 万
阿里 真武 V900平头哥 PPU未公布216GB未公布未公布2026-09 云栖发布,2027 Q1 量产

注:燧原 T21 的 TDP 官方未公布,本站不作推测;摩尔线程、燧原等卡的 FP8/FP4 精度多数官方未单列,表中以 BF16/FP16 口径统一呈现。天数智芯「智铠」系列未见公开采购确认,暂未纳入主表。

三、国际 AI 芯片进展​

3.1 NVIDIA:Blackwell 量产 + Rubin 首发​

型号架构制程显存带宽FP16(稀疏)TDP状态
H100 SXMHopperTSMC 4N80GB HBM33.35 TB/s1,979 TFLOPS700W上代主力
H200 SXMHopper(refresh)TSMC 4N141GB HBM3e4.8 TB/s1,979 TFLOPS700W大显存主力
B200Blackwell(双 die)TSMC 4NP192GB HBM3e8 TB/s5 PFLOPS1000W2025 起量产
GB300Grace Blackwell UltraTSMC 4NP576GB HBM3e(超节点口径)16 TB/s10 PFLOPS(双芯)3000W超节点主力
Rubin R200Rubin台积电 3nm288GB HBM419.2 TB/s4 PFLOPS1800W2026 H2 已规模出货

NVIDIA 在 2026 年的核心叙事是交付:Vera Rubin NVL72 下半年持续放量,MLPerf Inference v6.1 首秀(Qwen3-VL 较 GB300 最高 3.7 倍)。Rubin R200 官方 FP4 达 50 PFLOPS(稀疏),是本代出货主力;N1X PC 处理器秋季上市,把战场延伸到消费级。

3.2 AMD:MI300X/MI325X 守基本盘,MI455X 冲机架​

MI300X(192GB HBM3 / 5.3 TB/s / 1,307 TFLOPS)与 MI325X(256GB HBM3e / 6.48 TB/s)提供高显存性价比;真正的增量在 Helios 机架(72× MI455X,432GB HBM4 每卡、23.3 TB/s、FP4 40.3 PFLOPS、计算芯粒 TSMC 2nm),绑定 OpenAI 6GW、Meta 6GW 部署。MI450 为其母型号,量产节点排在 2027 年 Q2。

3.3 Intel 与 Google​

  • Intel Gaudi 3:TSMC 5nm、128GB HBM2e、3.67 TB/s、BF16 矩阵算力 1,835 TFLOPS、TDP 900W,24× 200GbE RDMA 集成网卡,以太网组网直接对标 NVL72。Gaudi 3 之后 Intel 路线转向 Crescent Island(推理)与 Jaguar Shores(2027),Falcon Shores 已取消。
  • Google TPU v6(Trillium):v6e(32GB / 918 TFLOPS / 200W)与 v6p(96GB HBM2 / 1,350 TFLOPS / 450W)同代,TSMC 5nm,仅通过 Google Cloud 按小时计费。第七代 Ironwood 已把 HBM 容量推到 192GB、带宽 7.38 TB/s。

3.4 其他赛道​

Cerebras 已于 2026-08 发布机架级 CS-4(3× WSE-3 Turbo,750 PFLOPS),需要注意的是 CS-4 并未采用新一代晶圆,硅片与 CS-3 完全相同;SambaNova SN40L 以三层内存架构服务稀疏场景;Graphcore 已被 SoftBank 收购并退出中国市场。

四、关键指标对比(核心)​

方法学:FP16 统一采用稀疏(sparse)口径,与站内科普及首页热门芯片一致;价格为 pricing.json 市场报价(国际 USD、国产 CNY,截至 2026-10);「未公布」表示官方未披露、本站不作推测填充。

4.1 算力对比(FP16 稀疏)​

芯片FP16(稀疏)FP8 / FP4 备注架构
国产
天垓 300超 1,000 TFLOPS支持 FP4 / FP8SIMT
昇腾 910C800 TFLOPSFP8 1.5 PFLOPSDa Vinci 双 die
思元 690700 TFLOPSFP8 约 1.4 PFLOPSMLUarch 4.0
昇腾 950DT547 TFLOPSFP4 2 PFLOPSDa Vinci v5
摩尔线程 MTT S5000400 TFLOPS(BF16)FP8 约 1 PFLOPS 稠密MUSA 4
天垓 150192 TFLOPS384 TOPS INT8ivcore11
云燧 T21134.4 TFLOPS未公布GCU-CARA 2.0
昆仑芯 R200128 TFLOPS256 TOPS INT8XPU-R
国际
GB300(双芯)10 PFLOPSFP4 40 PFLOPSGrace Blackwell Ultra
MI455X10.1 PFLOPSFP4 40.3 PFLOPSCDNA 5
B2005 PFLOPSFP4 20 PFLOPSBlackwell
Rubin R2004 PFLOPSFP4 50 PFLOPSRubin
Gaudi 31,835 TFLOPS(BF16 矩阵)FP8 1,835 TFLOPSGaudi 3
H100 SXM1,979 TFLOPSFP8 3,958 TFLOPSHopper
H200 SXM1,979 TFLOPSFP8 3,958 TFLOPSHopper
TPU v6p1,350 TFLOPSFP8 2,700 TFLOPSTPU v6
MI300X1,307 TFLOPS未单列CDNA 3
MI325X1,307 TFLOPS未单列CDNA 3

读图:单卡 FP16 上,B200/GB300 把国产旗舰甩开约 5~6 倍;但昇腾 910C(800)已接近 H100 的 40%,思元 690(700)紧随,天垓 300(超 1,000)已进入国际旗舰区间。差距从「数量级」收敛到「一代」。

4.2 显存与带宽对比​

芯片显存类型带宽
昇腾 960DT288GBHBM9.6 TB/s
MI325X256GBHBM3e6.48 TB/s
阿里真武 V900216GB未公布未公布
思元 690196GBHBM33.35 TB/s
B200192GBHBM3e8 TB/s
MI300X192GBHBM35.3 TB/s
H200141GBHBM3e4.8 TB/s
H10080GBHBM33.35 TB/s

读图:国产旗舰在显存容量上已经反超——思元 690 的 196GB、昇腾 960DT 的 288GB、真武 V900 的 216GB 都高于 H100/H200。但带宽上,只有昇腾 960DT(9.6 TB/s)与 B200(8 TB/s)进入同一梯队,HBM3e/HBM4 代差是关键。

4.3 功耗与能效对比​

芯片TDPFP16/W(稀疏能效)
昇腾 910C600W1.33 TFLOPS/W
思元 690500W1.40 TFLOPS/W
天垓 150350W0.55 TFLOPS/W
H100700W2.83 TFLOPS/W
B2001000W5.0 TFLOPS/W
MI300X750W1.74 TFLOPS/W
TPU v6e200W4.59 TFLOPS/W

读图:能效是国产最明显的短板。H100 的 2.83 TFLOPS/W 约为昇腾 910C 的 2.1 倍,B200 更是 3.8 倍。低功耗端 TPU v6e(200W / 4.59)与昆仑芯 R200(150W)代表「每瓦算力」的另一条路线。

4.4 制程对比​

厂商代表芯片制程晶体管
华为910C中芯 7nm N+2未公布
寒武纪思元 690中芯 N+2(5nm 级)未公布
NVIDIAH100 / B200TSMC 4N / 4NP800 亿 / 2080 亿
AMDMI300XTSMC 5+6nm 混合1530 亿
AMDMI455XTSMC 2nm(计算芯粒)+ 3nm未公布

读图:制程代差是结构性差距的根源。国产主力仍停留在 7nm 级,而国际旗舰已进入 2nm 计算芯粒时代。这是单卡算力差距的物理下限。

4.5 价格对比(市场报价,2026-10)​

芯片价格币种口径
H100$28,000USD市场均价
H200$38,000USD市场均价
B200$42,000USD市场均价
GB200$62,000USD市场均价
MI300X$13,500USD市场均价
MI325X$16,500USD市场均价
Gaudi 3$16,500USD市场均价
TPU v6e$22,000USDCloud 仅按小时
昇腾 910C¥110,000CNY2026-09 渠道(年初约 ¥9 万,+22%)
昇腾 950PR¥80,000CNY2026-09 渠道(年初约 ¥6 万,+30%)
昇腾 950DT¥250,000CNY2026-09 渠道(较 7 月 +20~50%)
思元 690¥175,000CNY2026-09 渠道(官方指导价 ¥19 万)
MTT S5000¥75,000CNY渠道估算

读图:按汇率约 7.2 折算,昇腾 910C(约 $15.3k)与 H100($28k)价差显著但算力仅其 40%,单位算力成本仍偏高;思元 690(约 $24.3k)价格接近 H100 但显存翻倍。HBM 短缺推高国产报价,是 2026 下半年的一致趋势。

五、生态对比​

厂商软件栈框架支持迁移难度特点
NVIDIACUDA + cuDNN全支持—算子覆盖约 100%,生态壁垒
AMDROCm主流支持中覆盖约 90%,持续补齐
华为CANN主流支持中覆盖约 80%,昇思 MindSpore
寒武纪NeuWare主流支持中覆盖约 75%~85%
天数智芯IXUCA主流支持低(兼容 CUDA)CUDA 兼容路线
燧原DTU Toolkit部分支持高覆盖约 60%
百度昆仑XREA / 飞桨主流支持低(飞桨原生)与 Paddle 深度绑定

关键判断:生态差距比硬件差距更难追。CUDA 的长期积累使 NVIDIA 在算子覆盖、工具链、开发者社区上仍具压倒性优势;国产路线分两派——华为/寒武纪自建栈,天数智芯/摩尔线程走 CUDA 兼容以降低迁移成本。

六、市场表现​

6.1 出货量(2026 预估)​

厂商2026 出货(万颗)同比主要客户
NVIDIA(全球)150–200+30%全球科技巨头
华为(昇腾)20–30+100%政府、国企、科技公司
寒武纪(MLU)10–15+150%字节、阿里、百度
AMD(全球)30–40+80%Meta、微软等
其他国产10–20+200%各类 AI 公司

6.2 中国市场份额​

类型20242026(预估)趋势
国际(NVIDIA/AMD)约 80%约 40%大幅下降
国产(华为/寒武纪等)约 20%约 60%快速提升

6.3 客户与部署​

  • 字节跳动 / 阿里 / 百度:寒武纪思元、昇腾 910C 双线采购;
  • 腾讯:昇腾 910C 为主,探索其他国产;
  • 智算中心:多家国产芯片混合部署,国产化率成硬指标;
  • 超大规模云:NVIDIA 仍主导海外训练,但国内云厂商国产占比快速抬升。

七、未来展望(2027)​

阵营产品时间状态
华为昇腾 960DT / 960PR2027 Q1 / Q3已官宣
华为昇腾 9702028已官宣
NVIDIARubin Ultra2027已公布
AMDMI450 / MI5002027 Q2 起已公布
IntelJaguar Shores2027+已公布
AWSTrainium42027已公布(FP4 为 Trainium3 的 6 倍以上)
寒武纪思元 7902027未公布

差距判断(2027 预期):

  • 算力:从落后 40%~50% 收窄至 30%~40%;
  • 显存:从落后 20%~30% 收窄至 10%~20%,部分型号已反超;
  • 制程:从落后 1~2 代收窄至约 1 代;
  • 生态:从落后 35 年收窄至 23 年。

八、结论与选型建议​

核心观点:

  1. 2026 年国产与国际旗舰的差距从「代际」缩小到「一代」,显存容量局部反超;
  2. 单卡稠密算力、能效、制程仍是结构性短板,集群扩展与软件生态是最大变量;
  3. 国产化率快速攀升(2025 约 41%),采购从「尝鲜」变「刚需」;
  4. 出口管制常态化下,国产替代已从政策驱动转为市场驱动。

给不同角色的建议:

  • 企业采购:训练为主且合规无碍,选 NVIDIA B200/GB300/Rubin;国产化硬指标,选昇腾 910C/950DT 混合部署;推理高密度,关注昆仑芯 R200 与 TPU v6e。
  • 开发者:优先 CUDA 保泛用性;若锁定国产,选 CUDA 兼容栈(天数智芯/摩尔线程)降低迁移成本。
  • 投资者:关注 HBM 供给、国产先进制程良率,以及昇腾/寒武纪的客单与份额变化。
  • 政策制定者:延续对先进制程、HBM、EDA 的投入,缩短生态差距比单点算力更关键。

数据来源与方法:本文全部性能与价格取自 MirrorFrog 已校验芯片数据库(chips.json,252 款)与价格库(pricing.json,报价截至 2026-10),数据核对日期 2026-10-13。FP16 统一稀疏口径;官方未公开字段以「未公布」标注,不作占位填充。文中 Rubin R200 带宽(19.2 TB/s)与 TDP(1800W)、Gaudi 3 算力(BF16 矩阵 1,835 TFLOPS)、昇腾 950 系列制程与报价等,均为 2026-10 核对后的修正值。

延伸阅读:完整芯片对比表、交互式对比工具、NVIDIA H100 规格卡、华为昇腾 910C 规格卡、寒武纪思元 690 规格卡。

寒武纪 MLU690 规模部署与订单交付最新进展

· 5 min read
Industry Research Team

国产 AI 芯片市场对 MLU690 的期待,从 2025 年一路延续到 2026 年。但据公开报道,MLU690 在 2026 年并没有形成大规模出货计划——寒武纪全年晶圆资源优先保障已成熟量产的 MLU580 与 MLU590。摩根士丹利 2026 年 10 月的研报也确认了这一点:MLU690 系列(特别是 MLU692)收入确认延迟,构成三季度业绩不及预期的主因。本文梳理 MLU690 的真实交付进展。

MLU690 交付现状:产能优先给成熟型号​

据公开报道,寒武纪 2026 年的产能分配策略已经明确:全年晶圆资源优先供给成熟量产的 MLU580、MLU590 两款芯片,MLU690 中芯版本暂不规划批量出货,也没有主动向客户推广。核心制约仍是产能瓶颈,即便拿到客户订单也无法按时交付,因此公司索性收缩 MLU690 的市场投放节奏。

摩根士丹利的研报从财务角度印证了同一件事:预计寒武纪 2026 年第三季度收入为 36.5 亿元人民币,低于此前预期的 47.7 亿元。不及预期的主要原因在于 MLU690 系列(特别是 MLU692)的收入确认延迟以及 MLU580 的交付量有限。寒武纪在第三季度已将 MLU692 交付给主要的 CSP(云服务提供商)进行测试,但预计该产品的大部分收入将在第四季度才能确认。

这就是 MLU690 当前的真实状态——不是"卖不动",而是"没有足够的晶圆去卖"。

型号2026 年产能地位交付状态
MLU580主力,产能倾斜交付量有限,三季度提价后毛利率仍偏低
MLU590主力,产能倾斜上半年内可交付
MLU690 / MLU692产能优先级靠后已交付主要 CSP 测试,收入多在四季度确认
MLU790(下一代)尚未推进受 690 验证进度影响,节奏更谨慎

据公开报道,MLU690 的合同虽在上半年签订,但交付预计在第三季度。这与摩根士丹利所述"大部分收入在第四季度确认"基本一致。

央国企市场:可见的订单空间​

在消费级与互联网市场之外,央国企与训练场项目是 MLU690 的主要落地方向。据公开报道,2026 年央国企市场的可见商机空间预计在 10 万至 20 万卡,其中上半年预计可落地订单约为 5 万卡。这 5 万卡中,有 3 万卡将用于与京能合作的国家训练场项目,另外 2 万卡来自其他央国企计算项目扩容(如浙江台州、昆山的项目)以及上海、安徽合肥等地的训练场和电信运营商项目。

值得注意的是,央国企订单的竞争格局并不宽松。据公开报道,在央国企市场寒武纪并非绝对的第二供应商,市场地位大约能排进前五;在信创领域华为是绝对的第一供应商,占据超过 60% 的市场份额。对寒武纪构成最大竞争压力的是平头哥和昆仑芯——这两家公司背靠阿里云和百度云,具备算力消纳能力,能通过云服务输出算力,而寒武纪无法做出此类承诺。

超节点布局:停滞状态​

超节点是 2026 年国产算力的主战场,但寒武纪在这条线上基本缺席。据公开报道,寒武纪自己没有做超节点,也没有明确的自建超节点计划。当前主要是给一些做超节点的厂商提供产品做验证和适配,2026 年还没有客户明确要求寒武纪去部署超节点,而且产能也不支持这么做。

现有产能仅能勉强满足字节的八卡基础服务器订单,字节自身也没有超节点部署计划,相关计划更偏 2027 年。腾讯和阿里虽有超节点方向的规划,但寒武纪没有办法保证供货,所以没有真正推进到客户部署层面。目前更多是与 OEM 厂商做平台适配,比如阿里的盘古、浪潮、新华三等平台已经做过适配,但没有量产安排。

还有一个被忽视的连锁反应:MLU690 的延迟量产会进一步拖累下一代 MLU790 的流片进度。MLU690 本身承担 790 前置大规模商用验证任务,缺少足量客户真实场景数据迭代,下一代芯片的研发与流片节奏会持续延后。

研判​

寒武纪当前的处境可以用一句话概括:需求不缺,晶圆不够。据公开报道,摩根士丹利预计寒武纪 2027 年向 CSP 客户的出货量将接近 35 万颗,主要得益于 MLU692 的产能爬坡和下一代 MLU580 产品的推出;代工方面主要依赖中芯国际的 N+2/N+1 产能。HBM3e 仍是关键供应瓶颈,但供应链调研显示寒武纪拥有一定库存,并有望随国内产能爬坡获得长鑫存储的一部分 HBM3e 产量。

财务侧的压力同样真实:HBM 成本上升对单位经济效益造成压力,MLU580 即使三季度提价,毛利率仍将保持在相对较低水平。摩根士丹利维持增持评级,目标价从 1408 元上调至 1700 元,但明确将三季度称为"业绩低谷"。

对客户而言,MLU690 在 2026 年不是可规模化规划的选项;2027 年随着产能爬坡与 790 产品线就位,它才会真正进入放量期。这一年多的窗口期,寒武纪要证明的是晶圆产能能否转化为可持续的交付能力。

寒武纪思元 590 深度解析:承前启后的关键一代

· 4 min read
Industry Research Team

2025 年寒武纪交出了营收 64.97 亿元、同比增长超过四倍、上市以来首次年度盈利的成绩单,背后的功臣正是思元 590。作为公司第三代云端训练与推理一体芯片,它既收割了出口管制带来的历史性窗口,也为新一代旗舰思元 690 铺平了道路。

规格速览:7nm Chiplet 的诚意之作​

思元 590 采用全新 MLUarch05 架构与 7nm Chiplet 封装技术,核心规格综合官方披露与公开报道整理如下:

参数思元 590思元 690(新一代,对照)
制程7nm,Chiplet 封装更先进制程,双 die 设计
FP16 算力256 TFLOPS(部分来源标注 345 TFLOPS)700 TFLOPS 以上
INT8 算力512 TOPS未完整公开
显存96GB HBM2e(另有 80GB 版本报道)196GB HBM3
显存带宽约 2 TB/s约 3.35 TB/s
标称功耗350W400 至 500W
状态2026 年全场景规模出货2026 年量产

能效比方面,公开资料显示思元 590 达到约 5 TOPS/W;其实测训练性能约为英伟达 A100 的八成水平,且功耗更低。

三年长跑:从送样到放量​

思元 590 的商业化时间线值得细看。2022 年 9 月世界人工智能大会上,董事长陈天石首次披露这款芯片的研发信息;2023 年,百度成为首批尝鲜者并完成内部测试;2024 年芯片问世并通过互联网大客户测试;2025 年起批量进入字节跳动、百度等公司的智算集群,切入大模型推理与训练供应链;2026 年初实现全场景规模出货。

这条路径的关键转折是 2025 年:在英伟达高端芯片对华供应持续受限的背景下,互联网大厂将算力预算切给国产供应商,寒武纪凭借第三方中立的厂商身份与相对完善的软件栈成为最大受益者之一。公开数据显示其 2025 年出货约 11.6 万颗,同比增长约三倍半,字节跳动一度成为占比超过四分之一的第一大客户。其间也出现过波折,2025 年 12 月寒武纪官方否认了"将推出约 50 万张含思元 590 加速卡"的市场传闻,提醒投资者以公告为准。

与思元 690:两代人的分工​

需要强调的是,思元 590 与思元 690 是两代产品,规格不可混为一谈。590 是 2025 年的营收主力,7nm 制程、HBM2e 显存、350W 功耗,定位于大规模推理与中等规模训练;690 则是 2026 年 7 月正式亮相的新旗舰,采用更先进制程与双 die Chiplet 设计,FP16 算力跃升至 700 TFLOPS 以上,显存升级到 196GB HBM3,带宽约 3.35 TB/s,对标的性能目标直指更高世代的国际旗舰。软件层面,两代产品共享 NeuWare 全栈与 MLU-Link 互联体系,客户从 590 迁移到 690 的成本相对可控。

小结:一代产品改变一家公司的命运​

思元 590 的意义超越了规格表本身:它证明了国产云端训练芯片可以在真实大客户的大模型负载中规模化商用,也让寒武纪从"八年亏损王"变成盈利标的。对于计划建设国产智算集群的团队,590 仍是当前供应稳定、生态验证充分的选择,而追求更高算力密度的用户则可以关注已经量产的思元 690。承前启后,正是这一代芯片最好的注脚。

国产 AI 芯片的订单积压:交付周期与产能爬坡

· 4 min read
Industry Research Team

2026年的国产AI芯片市场出现了一个教科书级的卖方市场特征:客户不问价格、不挑型号,只问一件事——什么时候能到货。据央视财经援引的行业调研,全年需求约400万颗,实际交付约300万颗,百万级缺口把订单周期一路推到三年开外。

满产满销:订单不是问题,产能才是​

多位产业链人士在公开交流中给出的判断高度一致:2026年国产算力芯片全行业满产满销,订单完全不愁,核心变量只剩上游产能与良率。头部互联网大厂与头部芯片公司签订保供框架协议、锁定绝大部分产能的做法,已成为行业标配——先签长约者得产能,后来者排队。

业绩数据佐证了供不应求的成色:2026年上半年,寒武纪营收59.96亿元,同比增长108.1%;摩尔线程营收17.36亿元,同比增长147.4%;壁仞科技营收12.36亿元,同比暴增约20倍。下游智算中心同样如此,有大型智算园区负责人公开表示在手订单已排到三年后,中游集成商的客户数量一年增长10倍,甚至催生了"集装箱式算力中心"——把传统近一年的建设周期压缩到24小时上线。

产业链环节2026年现状
上游芯片需求约400万颗,交付约300万颗,缺口百万级
下游智算中心头部园区订单排至三年后
中游集成部分厂商客户数量一年增长10倍
备货(寒武纪)存货82.48亿元,较上年末增长66.8%
备货(摩尔线程)存货从13.3亿元增至35.5亿元

积压的结构:从晶圆到 HBM 的逐层紧张​

订单积压并非单点瓶颈,而是一条传导链。最前端是先进制程晶圆:国产AI芯片高度依赖国内代工产能与有限的海外先进制程通道,产能分配向头部客户倾斜,行业普遍采用"第一顺位"机制。第二层是HBM与存储:国产HBM供给能力有限,叠加大环境存储涨价周期,先进存储成了最紧的物料之一。第三层是先进封装:CoWoS类2.5D封装产能全球性紧张,国产封装厂扩产需要时间。逐层卡脖子叠加,最终表现为交付周期从数周拉长到数月乃至更长。

厂商的应对是"以存货换交付":2026年上半年,国产AI芯片公司集体大规模备货,寒武纪存货达82.48亿元,海光信息75.18亿元,天数智芯从7.1亿元猛增至19.68亿元。提前锁料锁产能,本质是用资产负债表为交付承诺背书——在涨价预期下,早买还等于省钱,这笔账算得过来。

价格信号:供给短缺的隐性定价​

短缺正在改写价格体系。据公开调研信息,市场主流国产AI芯片的单卡价格受HBM涨价与产品组合变化影响持续上移,多家厂商上调了收入指引;同时晶圆代工端,台积电确认2027年起上调代工价格3%至6%,三星先进制程涨幅最高达15%,内地代工厂也纷纷提价。成本端的全面上移,最终会传导到算力服务价格——这也是国产算力租赁市场持续走强的底层原因之一。

研判:观察产能爬坡,而非订单数字​

对国产AI芯片行业,2026年的正确观察指标不是订单总额,而是三个爬坡速率:晶圆投片量的季度环比、HBM国产供给的实际到位节奏、以及先进封装的良率学习曲线。只要这三个指标持续向上,百万级缺口就会逐步收敛,收入兑现只是时间问题;反之若某环节卡壳,订单积压将转化为客户流失与替代方案回流。当下的国产芯片产业正处在从"能用到好用"再到"够用管饱"的中场——订单积压是这份成绩单上最耀眼也最脆弱的一栏。

寒武纪 MLU690 深度解析:700+ TFLOPS FP16 与国产训练卡的寒武纪时刻

· 6 min read
Industry Research Team

在华为昇腾与英伟达的双雄叙事之外,寒武纪正在写下国产 AI 芯片的第三条故事线:思元690 以 700+ TFLOPS FP16 登顶国产纸面算力,公司实现首次年度盈利,下一代芯片未发先涨 20%–30%。这个曾经靠情怀撑估值的"AI 芯片第一股",第一次同时握住了技术、商业与定价权三张牌。

一、思元690 核心规格:196GB HBM3 撑起的算力天花板​

按本站芯片卡收录的规格,思元690(MLU690)采用双 die Chiplet 封装,2026 年初量产,核心参数如下:

项目参数
架构新一代 MLUarch 架构
制程中芯国际 N+2(5nm 级),亦见台积电 4nm 双 die 报道
显存196 GB HBM3
显存带宽3.35 TB/s
FP16/BF16700+ TFLOPS
INT42800+ TOPS
互联MLU-Link 890 Gbps/卡
TDP400–500 W
单价¥14–15 万

与上代思元590 相比,这是一次全面换代;与 H100 相比,则是典型的"错位竞争":

指标思元690思元590H100 SXM对比 H100
显存196GB HBM396GB HBM2e80GB HBM32.45×
显存带宽3.35 TB/s~2 TB/s3.35 TB/s持平
FP16700+ TFLOPS345 TFLOPS989 TFLOPS~71%
INT8 推理反超(INT4 2800+ TOPS 口径)1380 TOPS1979 TOPS部分场景反超

196GB 显存是思元690 最锋利的差异化武器:单卡装下更大的模型与 KV Cache,在推理场景性价比突出;纯推理场景可达到 H100 的 80%–90%。短板同样清晰——FP16 训练算力与千卡以上分布式训练的互联工程成熟度。为此,WAIC 2026 上寒武纪同步落地了 256 卡规模超节点集群方案,已在多地智算中心部署,从"单卡算力"走向"系统级算力"。

二、架构演进:MLUarch 换代与 HBM3 上量​

思元690 的代际跃迁体现在三处:其一,MLUarch 架构换代(市场披露口径称新一代芯片采用 MLUarch 04),配合双 die Chiplet 封装,把晶体管预算投向张量算力与显存通道;其二,显存从 HBM2e 一举上量到 196GB HBM3,3.35 TB/s 带宽与 H100 持平——在国产供应链里率先解决"内存墙";其三,MLU-Link 890 Gbps 卡间互联已接近 NVLink 4 的 900 GB/s 量级,为 256 卡超节点铺路。软件侧的 NeuWare 是国内少有的云边端统一栈:PyTorch/TensorFlow 适配版、NeuBLAS、NeuDNN,以及全球首批商用 MLIR 图编译推理框架 MagicMind。

供应链上,台积电因实体清单停供后,寒武纪全线转向中芯国际 N+2,成为中芯先进制程的最大客户之一;N+2 产能已锁定至 2027 年。

三、涨价 20%–30%:国产芯片第一次掌握定价权​

路透社 2026 年 9 月 10 日报道:寒武纪下一代芯片(市场暂称"690")重新定价,较两个月前上涨 20%–30%。同期昇腾 950DT 报价突破 25 万元、沐曦、天数智芯等国产卡也集体上调 20%–30%——这不是寒武纪一家的孤例,而是 HBM 成本飙升叠加供不应求下的全行业现象。

值得注意的是官方态度:寒武纪董秘办回应"不要听信市场消息,一切以官方公告为准"。渠道价与官方定价之间存在明显温差——涨价是市场现实,但公司不愿为渠道炒作背书。这背后是国产 AI 芯片定价权的历史性变化:过去国产卡必须以"替代品折价"换取订单,如今供给稀缺让定价逻辑转向"稀缺溢价"。在华为、寒武纪、摩尔线程构成的国产三强格局中,寒武纪第一次拥有了"不降价也能卖动"的底气。

四、首次年度盈利:商业模式的拐点​

如果说涨价反映的是市场热度,盈利则验证了商业模式的成色。按 2026 年上半年行业复盘口径,寒武纪已实现首次年度盈利。配套数据同样亮眼:2026 年 Q1 营收 28.85 亿元,同比增长 160%;2026 年 6 月市值突破万亿元。客户结构上,字节跳动是最大客户(思元580/590/690 合计采购超 50 万颗,约占其采购的 65%),阿里、百度、中国移动相继进入订单名单。

"首次盈利"对国产 AI 芯片行业的标志意义在于:这个赛道第一次出现了不依赖补贴与融资、靠卖芯片本身赚钱的商业模式。研发投入可以自我造血,意味着产品迭代不再以资本耐心为限——这对动辄三年一代的芯片研发周期,是决定性的正循环。

五、命名之辨:思元系列与市场传闻的"690"​

最后回到一个细节:市场上热议的"690",官方从未正式发布命名。本站芯片卡以"思元690"收录其规格(已量产口径),而路透报道中的"690"实为市场对下一代芯片的暂称,寒武纪对具体命名、规格与定价均保持谨慎,一切以官方公告为准。这种"市场抢跑、官方缄默"的组合本身就是一个信号:在供给稀缺的卖方市场里,厂商不再需要靠预热造势——把名字留到正式发布,反而是一种定价权。

小结​

思元690 代表的"寒武纪时刻",本质上是三件事同时发生:技术上,700+ TFLOPS FP16、196GB HBM3 与 256 卡超节点让国产训练卡首次形成完整的系统级方案;商业上,首次年度盈利证明了国产 AI 芯片可以自我造血;市场地位上,20%–30% 的重新定价宣告国产芯片开始掌握定价权。当然,考验依然存在:千卡以上训练的互联工程成熟度、NeuWare 与 CUDA 的生态差距、以及下一代芯片能否如期放量。但至少在 2026 年秋天,国产 AI 芯片第一次同时拥有了技术、利润与议价能力。

HBM短缺引爆国产AI芯片涨价潮:昇腾950DT报价超25万元、寒武纪690涨20%-30%

· 5 min read
Industry Research Team

据路透社报道,出口管制推高HBM成本,华为、寒武纪等国产AI芯片厂商集体上调价格——正如戴尔CEO所言,"比涨价更糟的是没货"。

路透社爆料:三位知情人士描绘的涨价全景​

据路透社2026年9月10日援引三位知情人士的报道,因出口管制导致HBM成本飙升,华为、寒武纪等中国AI芯片厂商已大幅上调现有及下一代AI处理器的价格。这不是单一产品的调价,而是一轮覆盖多代产品线的系统性涨价。

产品年初价格当前价格涨幅
华为昇腾950DT加速卡未公开指示性价格25万元人民币以上(约37255美元)两个月内报价最多上涨50%
华为昇腾950PR约6万元/单卡突破8万元约30%
华为昇腾910C板卡(老款)约9万元11万元以上—
寒武纪下一代芯片(暂称690)未公开未公开较两个月前上涨20%-30%

其中最引人注目的是昇腾950DT:其指示性价格已调高至25万元人民币以上(约37255美元),两个月内报价最多上涨50%。年初单卡约6万元的昇腾950PR,现已突破8万元;老款昇腾910C板卡也从年初的约9万元涨至11万元以上——连上一代产品都在涨价,说明短缺是全方位的,而非某一型号的局部现象。

寒武纪官方回应与厂商动态​

在涨价消息流传之际,寒武纪董秘办作出回应:"不要听信市场消息,一切以官方公告为准。"这一表态既未证实也未否认。据上述报道,寒武纪下一代芯片(暂称690)已重新定价,较两个月前上涨20%-30%。

与此同时,涨价并非华为与寒武纪的"专属":沐曦、天数智芯亦有类似的涨价动作。也就是说,这轮价格上行已经从头部厂商扩散至更多国产AI芯片公司,成为行业性现象。对采购方而言,"再等等看会不会降"的观望策略,正在面临价格持续上行的反向风险。

涨价根源:HBM短缺与出口管制双重挤压​

这轮涨价的逻辑链条并不复杂,可以拆成三层:

  1. 需求侧:AI算力需求持续爆发,国产AI芯片订单饱满,供不应求是常态;
  2. 供给侧:全球HBM持续紧缺,高带宽显存作为AI加速卡最关键的物料,产能增量远远跟不上需求增量;
  3. 成本侧:出口限制卡住了国产芯片的上游供应,HBM获取成本大幅飙升,直接推高单卡物料成本。

上游成本抬升叠加供不应求,价格上调几乎是必然结果。对厂商而言,涨价既是转嫁成本的被动选择,也是在新的供需格局下重新校准产品定价的主动机会——毕竟在"有货就能卖掉"的市场里,定价权掌握在供给方手中。

戴尔CEO的警告:2027年可能更严重​

需求侧的信号同样值得警惕。戴尔CEO在高盛大会上表态:半导体结构性短缺短期难以缓解,2027年可能更严重;并且留下一句点睛之语——"比涨价更糟的是没货"。

这句话道破了当前AI算力市场的真实状态:价格上行尚可通过预算调整来消化,而拿不到货则直接卡死算力建设进度,延误的是整个AI战略的落地节奏。HBM作为AI加速卡的瓶颈物料,其短缺正在向上传导为加速卡、服务器乃至智算中心建设层面的系统性不确定性。

对采购决策的影响​

对于正在规划AI算力建设的企业来说,这轮涨价潮至少意味着三件事:

  • 预算规划要留弹性:单卡成本模型需要动态更新,参考昇腾950PR两个月上涨约30%的斜率,静态预算很容易失真;
  • 交期比价格更关键:昇腾910C这类老款产品都在涨价,说明现货稀缺,早锁定意味着早交付、早形成算力;
  • 多供应商组合价值凸显:华为、寒武纪、沐曦、天数智芯全线涨价的环境下,国产+海外、多品牌组合的采购策略更有抗风险能力。

小结​

HBM短缺与出口管制双重压力下,国产AI芯片正在经历一轮罕见的全线上调:华为昇腾950DT报价突破25万元(约37255美元)、两个月最多涨50%,昇腾950PR突破8万元,昇腾910C板卡涨至11万元以上,寒武纪下一代芯片690较两个月前上涨20%-30%,沐曦、天数智芯跟进调价。戴尔CEO"2027年可能更严重"的判断提示我们:这场短缺或许不是短期波动,而是未来数年的行业常态。对采购方而言,与其等待价格回落,不如尽早锁定供货。

国产 AI 芯片半年报大检阅:寒武纪净赚 23 亿、壁仞营收暴涨 20 倍、燧原登板,"抢芯大战"白热化

· 6 min read
Industry Research Team

8 月底至 9 月初,国产 AI 芯片厂商半年报密集披露,加上燧原科技 9 月 2 日启动 IPO 申购,被称为"国产四小龙"的摩尔线程、沐曦、壁仞、燧原全部完成上市,加上早已在科创板的寒武纪——国产 AI 芯片的资本市场拼图就此补齐。更重要的是,财报数字第一次集体印证了一件事:国产算力正从"能用"跨向"好用",商业化拐点已现。


1. 半年报成绩单:增长是主旋律,盈利是分水岭​

厂商2026H1 营收同比盈利状态技术路线
寒武纪59.96 亿元+108.1%归母净利 23.11 亿自研 MLU 指令集(DSA)
摩尔线程17.36 亿元+147.4%净亏 1156 万(收窄)全功能 GPU(兼容 CUDA 路线)
壁仞科技12.36 亿元+1997.6%未盈利通用 GPU
沐曦13.24 亿元+44.7%净利 6.12 亿(首次扭亏)通用 GPU
燧原11.20 亿元+279.1%未盈利DSA 专用架构(TopsRider)

几个值得注意的细节:

  • 沐曦率先跨过盈利线:8 月 31 日披露的半年报显示净利润 6.12 亿元、同比扭亏(上年同期亏损 1.86 亿);不过扣非净利润仍为 -4900 万(亏损收窄 75.8%)——含金量仍在爬坡。
  • 壁仞低基数暴增:近 20 倍的同比增速来自上年同期极低的收入基数,但 12.36 亿的绝对体量已与燧原、沐曦同量级,第二梯队座次重新洗牌。
  • 研发强度惊人:沐曦研发费用占营收 39.7%、摩尔线程 44.3%、壁仞高达 65%——高研发投入是全员未完全盈利的根本原因,也是未来竞争力的来源。
  • 摩尔线程毛利率承压:从 78% 降至 57%,成本增速(+245%)远超营收增速(+147%)——大规模量产期的品控与爬坡成本开始显现。

2. 燧原登板:四小龙资本拼图补齐​

燧原科技 9 月 2 日启动公开申购,发行 4303 万股新股(约占上市后总股本 10%),募资目标 60 亿元,采用 DSA 专用架构 + 自研 TopsRider 软件平台(不兼容 CUDA)。至此:

  • 科创板:寒武纪(2020 年)、摩尔线程、沐曦、壁仞(2026 年)
  • 燧原:2026 年 9 月完成上市

国产 AI 芯片第一梯队全部进入公开市场,融资通道打开后,研发投入的"军备竞赛"将进一步升级。

3. 需求端:百万卡缺口,订单排到三年后​

财报爆发的另一面,是需求端的极度饥渴:

  • 产能缺口:行业调研显示,2026 年国产 AI 芯片需求规模约 400 万颗,实际交付约 300 万颗,存在百万级缺口;
  • 订单周期:内蒙古乌兰察布远景星河基地(规划支撑百万卡级并行算力)表示"在手订单已排到三年后";
  • 供给创新:算力紧缺催生"集装箱式算力中心"——20/40 英尺标准集装箱为载体,插电接水即可在 24 小时内完成部署,把传统"一年工期"压缩到"一天上线";
  • 市场空间:IDC 数据显示 2025 年中国 AI 加速卡出货约 400 万片,国产占约 41%(165 万片);CIC 预测中国 AI 加速器市场 2028 年将超万亿元,其中国产方案份额约 90%。

4. 两种路线的两种赌注​

市场研究机构预测 2026 年中国高端 AI 芯片市场中,国产方案份额将接近 90%,其中华为约 62%、寒武纪约 14%,剩余由摩尔线程、沐曦、壁仞等共同占据。而寒武纪与摩尔线程这对"双龙头",正在押注两种截然不同的未来:

寒武纪:用现金买确定性。 半年报资产负债表上,存货 82.48 亿 + 预付款 29.14 亿,两项合计 111.6 亿、占总资产 61%——在实体清单限制下"产能即订单",提前锁定未来 12-18 个月的晶圆产能,代价是经营现金流净额同比下滑 66%,且上半年已计提存货跌价损失 3.97 亿。底气来自订单确定性:57 亿元股权激励计划的考核目标是 2026 年营收不低于 135 亿、2026-2028 年累计不低于 1000 亿。

摩尔线程:用亏损买时间。 全功能 GPU 的"大而全"路线(AI 计算 + 图形渲染 + 物理仿真 + 视频编解码)前期投入巨大,需要在现金流耗尽之前跨过规模化的门槛。上半年亏损已收窄至千万级,IPO 募资到位后,时间窗口正在变宽。

5. 软件生态:被忽视的胜负手​

需求外溢(英伟达 H20 系列在中国市场遇冷)推动 DeepSeek、智谱 GLM、月之暗面 Kimi 等头部模型纷纷适配国产加速器(昇腾、沐曦、摩尔线程等),这给了国产芯片难得的"真实负载打磨机会"。但正如行业共识:芯片可以三年一代,生态需要十年之功。寒武纪自研指令集的封闭高效、摩尔线程 MUSA 对 CUDA 生态的兼容路线、燧原 DSA 的专用极致——哪条路能沉淀出真正的开发者生态,才是决定 2030 年格局的关键变量。


相关链接​

参考资料​


本文基于上市公司半年报、Pandaily 与央视财经等公开报道整理。财务数据为公司披露口径,市场份额为研究机构预测,不构成任何投资建议。

Domestic Big Three 2026 H2: Localization Rate Crosses 40% Toward 60%, Ascend 960 Roadmap, MLU690 and S5000 Ecosystems Ramp Up

· 6 min read
Industry Research Team

In 2026, China's AI chip market landscape has shifted from "NVIDIA unipolar dominance" to "overseas vendors leading, domestic multi-route catch-up." According to industry research, China's overall AI accelerator market was ~4M units in 2025, of which 1.65M were domestic, with share first breaking 40%; as products iterate and fabs follow up, the localization rate is expected to rise to 60%-70% by 2027. This article focuses on the latest H2 2026 progress of Huawei Ascend, Cambricon, and Moore Threads — the domestic "Big Three."


1. Huawei Ascend: 950 Capacity Fully Booked, 960 Roadmap Unveiled​

Ascend's core advantage is "architecture + full-stack ecosystem synergy," with ~800K units shipped in 2025, capturing 50% of the total domestic vendor share. The product iteration cadence is clear:

TimeProductNote
2025 Q1Ascend 910CMain transitional model
2026 Q1Ascend 950PRInference flagship
2026 Q4 (planned)Ascend 950DTTraining flagship, drives domestic HBM iteration
2027-2028Ascend 960 / 970Roadmap products

950 series capacity has entered a "fully booked" state: 950PR entered mass production in April 2026; June monthly capacity jumped to 500K-600K units (nearly 10x MoM), with a full-year target of 1.2M units at 100% certainty; ByteDance locked in 350K units for $5.6B, while Tencent / Alibaba / Baidu combined locked in 400K units.

Ascend 960 roadmap specs (per roadmap disclosure):

MetricAscend 960
ArchitectureAscend 6th gen (Da Vinci v6)
FP8 compute~4 PFLOPS
Memory288GB
Memory bandwidth9.6 TB/s
Super-nodeAtlas 960 SuperPoD, 15,488 cards, Lingqu optical-electrical converged bus
Debut2027 Q4 (roadmap)

The previous-gen Ascend 384 super-node has cumulatively shipped over 750 sets, deployed across 20+ industries including internet, operators, finance, education, and healthcare — Huawei calls it "the only domestic super-node that has trained a SOTA model."


2. Cambricon MLU690: H2 Mass Production, Entering ByteDance Bidding Window​

Cambricon is the core domestic compute leader in the absence of an Ascend IPO, with the technology gap continuously narrowing:

  • Siyuan 590 (7nm): Performance equivalent to 80% of A100, already supports DeepSeek, continuously adapting to mainstream large models like Qwen 3 and GLM
  • Siyuan 690 series: Will enter mass production in H2 2026, expected to achieve order scale-up during ByteDance's H2 bidding window
  • Revenue certainty: Equity incentive targets show >100% revenue growth for the next 3 years: 2026 revenue target 13.5B RMB, 2027 27B RMB, 2028 60B RMB

Cambricon fully benefits from the industry dividend of "domestic CSP capex + full adaptation of domestic large models and domestic chips," making it the most direct elasticity play on rising localization rate.


3. Moore Threads MTT S5000: Full-Function GPU + Ecosystem Breakthrough​

Moore Threads takes a differentiated "full-function GPU" route, with the flagship MTT S5000 based on the 4th-gen "Pinghu" MUSA architecture:

MetricMTT S5000
Dense AI compute1000 TFLOPS
Memory80GB
Memory bandwidth1.6 TB/s
Inter-card interconnect784 GB/s
PrecisionFP8 to FP64 full precision (training + inference)
SecurityFirst batch to pass national "Safe and Reliable Evaluation" (Level I)

Its engineering capability is verified: the Kuae (KUAE) intelligent computing cluster based on S5000 achieves 95% training linear scaling efficiency, with compute efficiency loss within 5% at ten-thousand-card scale; supports checkpoint-resume training with effective training time ratio >90%; and has trained a MoE-236B base model with >25 trillion tokens of corpus from scratch.

The ecosystem is Moore Threads' deepest moat: MUSA has achieved 100% core math library compatibility, 3000+ PyTorch operator compatibility, covers 55 categories of core AI operators, has official vLLM and SGLang support, Day-0 adaptation of mainstream models, and 800K+ developers. Its PD heterogeneous-disaggregation solution achieves equivalent replacement of international high-end GPUs at a 2:1 ratio with S5000, significantly reducing inference cost.

The 5th-gen "Huagang" architecture (released 2025-12) supports FP4 to FP64 full precision, with 50% higher compute density and 10x better energy efficiency than the previous gen, supporting 100K+ card clusters; cumulative R&D investment in the "Huashan" (train-infer integrated) and "Lushan" (graphics rendering) new chips based on this architecture exceeds 900M RMB.


4. Software Ecosystem Decides: Day-0 Adaptation Becomes Routine​

Beyond hardware, software ecosystem realization is the watershed for domestic compute in 2026:

  • Huawei's CANN heterogeneous computing architecture and MindSeries suite are fully open-sourced, with the community incubating 67 projects, 12.44M+ lines of code, and 3,500+ monthly active developers
  • The "release-and-adapt" closed loop between domestic large models and domestic chips has basically formed: Tencent Hunyuan T3 (295B), DeepSeek-V4, and GLM-5.2 all completed Day-0 adaptation
  • 2026 is regarded as the "first year of domestic super-nodes"; Huatai Securities estimates China's super-node architecture market will reach 341.4B RMB by 2028, with a 2026-2028 CAGR of 194%

5. Industry Judgment: From "Can It Be Built" to "Can It Be Used Well"​

The domestic Big Three are converging along three paths:

  1. Huawei: Locks government/enterprise and internet big customers with super-node system-level capability + full-stack software
  2. Cambricon: Impacts the revenue inflection point by narrowing the training-side gap + scaling up via big-customer bidding
  3. Moore Threads: Covers cloud-edge-end full scenarios with full-function GPU generality + mature CUDA-compatible ecosystem

The common shortcoming of all three remains advanced process and HBM supply — precisely the core link of overseas controls. But as domestic HBM iterates and fabs follow up, a realistic path to 60%-70% localization by 2027 exists.

References​


This article is compiled from public industry research, broker views, and corporate announcements as of August 2026. Some shipment and market-share figures are third-party estimates, not officially confirmed data.

Domestic GPU IPO Wave: The "Four Little Dragons" Assemble on Capital Markets, Moore Threads MTT S5000 Benchmarks Against H100

· 5 min read
Industry Research Team

From December 2025 to July 2026 — just half a year — at least 6 AI chip companies have listed or are about to list on capital markets. Together with already-listed Cambricon, Hygon, and Iluvatar, the domestic GPU corps' total market cap is approaching ¥2 trillion. This marks the critical climb from domestic GPUs being "usable" to "useful."

1. The "Four Little Dragons" assemble on capital markets​

CompanyListing statusRaise / issue priceSponsor
Moore ThreadsListed (STAR Market sh688795, 2025-12-05)Issue price ¥114.28, raised ¥8BCITIC Securities
MetaXIPO accepted (2026-06-30)¥3.904B (total investment ¥5B)Huatai United
EnflamePassed review (2026-06-15)¥6B—
BirenHKEX / sprinting——

Already-listed camp: Cambricon (sh688256, STAR Market 2020-07-20), Hygon, Iluvatar (HKEX). Moore Threads turned a book profit of ¥29.35M in Q1; MetaX narrowed losses 57.7% and gave a 2026 breakeven timeline.

2. Moore Threads MTT S5000: benchmarking against H100​

Moore Threads announced its flagship AI train+inference GPU MTT S5000 successfully completed full-pipeline adaptation validation of Zhipu's new-generation large model GLM-5 — measured performance "breaks the domestic compute ceiling":

MetricMTT S5000
Architecture4th-gen "Pinghu" architecture
FP8 compute1 PFLOPS (1,000 TFLOPS)
Memory bandwidth1.6 TB/s
PositioningFull-function train+inference GPU, benchmarks against NVIDIA H100
ProductionMass-produced; clusters online supporting trillion-parameter training

Deployment validation: jointly completed full-pipeline training of embodied-brain model RoboBrain 2.5 with BAAI; partnered with SiliconFlow for high-performance DeepSeek-V3 inference, single-card speed near international top products. IPO funds go to three directions: next-gen AI train+inference chip, next-gen graphics chip, next-gen AI SoC chip.

WAIC 2026 new progress: Moore Threads showcased the MTT C256 SuperNode (first-of-its-kind single-layer Scale-up 256-card full interconnect, sub-microsecond latency) and three AI-factory solutions — "model training factory / token production factory / agent production factory"; the company pre-announced H1 2026 revenue of ¥1.65B-1.75B, up 135%-149% YoY.

3. Cambricon: dual flagships MLU590/690​

ChipProcessComputeMemoryCustomer / status
MLU590 (思元590)7nm ChipletINT8 512 TOPS / FP16 345 TFLOPS96 GB HBM2eByteDance inference mainstay, ~80% of A100 overall, mass shipments early 2026
MLU690 (思元690)5nm-class (SMIC N+2)FP16 700+ TFLOPS / INT8 2800+ TOPS196 GB HBM3 (3.35 TB/s)Dual-die packaging, MLU-Link 890 Gbps; ~70% of H100 (80-90% pure inference); ByteDance largest customer, mass production early 2026

Cambricon is the only domestic AI chip vendor with a "unified edge-cloud architecture" — one MLU instruction set spans 思元 220 (edge) → 370 (border) → 590/690 (cloud), with one NeuWare toolchain across compute tiers.

Capital and performance double explosion: Cambricon's total market cap exceeded ¥1 trillion on June 30, 2026, becoming the STAR Market's first "trillion-yuan stock," up 75%+ YTD. On performance, Q1 2026 revenue ¥2.885B (+160% YoY), deducted net profit ¥934M; full-year 2025 revenue ¥6.497B (+453% YoY), net profit attributable to parent ¥2.059B, ending long-term losses. ByteDance has cumulatively deployed over 100k 思元 590/690, its largest customer.

4. DeepSeek-V4 effect: changing the expectation coordinate system​

On April 24, 2026, DeepSeek released the trillion-parameter flagship DeepSeek-V4. Unlike a year earlier when V3's launch sparked debate over "can domestic chips even run large models," this time multiple domestic chips — Huawei Ascend, Cambricon, Hygon, MetaX, Moore Threads, Kunlun, T-Head, Iluvatar — completed adaptation on launch day.

The evaluation coordinate system is shifting: from "what percentage of NVIDIA's same-generation product performance" to "can it carry the real workloads of top-tier large models."

Industry interpretation​

  1. Capital ammunition in place: dense IPOs provide ample funding for domestic GPU R&D iteration and capacity expansion, moving from "technology breakthrough" to "commercial virtuous cycle."
  2. Train+inference becomes the mainstream route: Moore Threads takes the full-function GPU route (graphics+AI+general compute), differentiating from Huawei Ascend's "AI-focused."
  3. Software ecosystem is the decider: Day-0 adaptation and the maturity of unified software stacks (MUSA / NeuWare / MXMACA) are replacing raw peak compute as the core yardstick of domestic GPU "usability."

References​


This article continuously tracks the domestic GPU listing process and product iteration.

2026 H1 AI Chip Industry Review: Blackwell Ultra, the Domestic Big Three, and the Inference Era

· 11 min read
Industry Research Team

In the first half of 2026, the AI chip industry underwent a historic turning point — the center of gravity shifted from the "training race" to "inference efficiency," domestic chip market share broke 40% for the first time, NVIDIA built higher barriers with Blackwell Ultra, and the inference-specific chip track bloomed in diversity.


I. Compute Doubles Again: NVIDIA Blackwell Ultra Launch (June 1)​

On June 1, 2026, NVIDIA CEO Jensen Huang unveiled the new-generation AI chip Blackwell Ultra at Computex 2026 (Taipei), setting a new starting line for the AI infrastructure race over the next two years.

Key Specs​

MetricBlackwell UltraB200Improvement
FP8 compute20 petaFLOPS~10 petaFLOPS100%
ArchitectureBlackwell UltraBlackwellUpgrade
Expected delivery2027 Q12026 Q1—
PositioningHyperscale training + inferenceTraining + inferenceFlagship

Industry Significance​

  1. Direct impact of doubled compute: 20 petaFLOPS FP8 means training time for hundred-billion-parameter models drops sharply; trillion-parameter model training moves from "scientific experiment" to "engineering routine"
  2. System-level balance: Blackwell Ultra is not just a chip but a system-level engineering breakthrough across NVLink, HBM, cooling, and power delivery
  3. Roadmap certainty: The Q1 2027 delivery timeline lets cloud vendors and AI labs plan infrastructure budgets 18 months ahead

Challenges​

  • Energy crisis: Doubled performance comes with sharply higher power; datacenter power and cooling design face extreme challenges
  • Accessibility: Top-tier compute goes first to top cloud vendors; how smaller developers and research institutes reach compute at reasonable cost via cloud services
  • Software stack adaptation: New hardware needs matching CUDA versions and framework support; software ecosystem maturity becomes the key bottleneck for compute conversion

II. Domestic AI Chips: The Tipping Point from "Usable" to "Good"​

On June 16, 2026, Xinchuang World published "2026 China Domestic AI Chip Vendor Capability Quadrant", clearly outlining the overall domestic landscape.

2.1 Capability Quadrant Ranking​

QuadrantRepresentative Vendors
Leader quadrantHuawei Ascend, Hygon, Cambricon, Alibaba T-Head, Moore Threads
Visionary quadrantBaidu Kunlunxin, Biren, Enflame, Iluvatar, HardyVision
Contender quadrantTSINGMICRO, Black Sesame, SemiDrive, Lisuan, Houmo
Challenger quadrantDenglin, Zhicun, VeriSilicon, Rockchip, Intellifusion

2.2 Huawei Ascend: The Anchor of Domestic Compute​

Market Position​

  • In 2025, Ascend series shipped 812,000 units, capturing 49% of the domestic AI accelerator card share, firmly No.1 domestically
  • Ascend 950PR single-card FP8 compute reaches 1P (PetaFLOPS), FP4 compute reaches 2P
  • Inference performance is about 2.87x that of NVIDIA H20, priced at only 72,000-75,000 RMB, a significant price/performance advantage

Full-Stack Advantage​

Huawei's "device-network-cloud-chip" integrated strategy is Ascend's core moat:

  • Chip design: Da Vinci 3.0 architecture iterating continuously
  • OS: HarmonyOS/Euler OS deeply optimized
  • Networking: Euler network protocol stack
  • Cloud: Huawei Cloud ModelArts platform seamlessly integrated

Latest Progress​

  • On June 5, 2026, Shenzhen Hetao College, together with HIT (Shenzhen) and Huawei, completed full-parameter post-training of a 1.6-trillion-parameter DeepSeek V4 Pro model on an Ascend 910C cluster
  • This is the first time domestic AI chips completed trillion-parameter-level model training, marking "domestic substitution" moving from inference to training

2.3 Cambricon: The First Profitable Domestic AI Chip Benchmark​

Performance Explosion​

MetricFull-year 20252026 Q1YoY Growth
Revenue6.497B RMB2.885B RMB+453% / +160%
Net profit2.059B RMB (first annual profit)1.013B RMB— / +185%

Core Product: Siyuan 590​

  • In DeepSeek R1 inference scenarios, TPS reaches 942, about 50% higher than H20
  • Years of joint optimization with ByteDance; strongest short-term cloud inference deployment capability
  • Of 2.885B RMB Q1 2026 revenue, Siyuan 590 contributed over 70%

Potential Risks​

Absent from the 2nd 2026 "Safe and Reliable Evaluation Results Announcement"; the reason is unclear and will affect its domestic government/enterprise market performance.

2.4 TSINGMICRO: The "Third Route" of Reconfigurable Chips​

Technical Route​

TSINGMICRO adopts a reconfigurable dataflow architecture同源 with Groq LPU, finding a balance between GPU generality and ASIC extreme efficiency.

MetricTSINGMICRO TX81Traditional GPUAdvantage
Inference costBaseline+100%Reduced 50%
Energy efficiencyBaselineBaseline3x improvement
ArchitectureReconfigurable dataflowSIMT/SIMDBetter for inference

Deployment Progress​

  • Cumulative shipments of reconfigurable chips exceed 30 million units
  • Scaled deployment in a dozen-plus thousand-card-scale intelligent computing centers nationwide
  • Has begun A-share IPO tutoring; likely to become the "first reconfigurable chip stock"

III. The Inference Chip Track: Core Signal of the Industry Shift​

On June 4, 2026, TrendForce published a deep report "The Era of Inference Economy: The Rules of AI Chips Are Being Rewritten," pointing out that the compute competition center of gravity is shifting from training to inference.

3.1 Why Now?​

Cost Structure Changed​

  • Training is a one-time cost: Once a model is trained, marginal cost approaches zero
  • Inference is a recurring cost: Every API call, every generated token represents compute consumption and gross-margin pressure
  • Per-unit inference cost and energy efficiency directly affect gross margin and scale-expansion capability

Model Compression Tech Matured​

  • 1.58-bit quantization and weight pruning let models maintain inference accuracy at extremely low memory footprint
  • MoE (Mixture of Experts) architecture activates only a few expert sub-networks per inference via "partial wake-up," greatly reducing actual computation
  • The rise of slimmed models provides commercial viability for hard-wired inference chips

3.2 NVIDIA's $20B Bet: Acquiring Groq (December 2025)​

On December 24, 2025, NVIDIA acquired Groq's inference technology license and core team for $20 billion, one of NVIDIA's largest M&A/tech acquisitions ever.

Strategic intent:

  1. Fill the inference gap: NVIDIA GPU is unshakable in training, but inference efficiency was never its strongest suit
  2. Counter specialized inference chips: Cerebras, Taalas, SambaNova and other startups are eroding the inference market
  3. Position for Agentic AI: Agentic AI needs extremely low-latency, high-throughput inference

3.3 Taalas HC1: Proof of Concept for Hard-Wired Inference​

On February 20, 2026, Canadian AI chip startup Taalas launched inference chip Taalas HC1, directly etching Meta's open-source AI model Llama 3.1 8B into the chip.

Key Metrics​

MetricTaalas HC1NVIDIA B200 (throughput optimized)Advantage
Inference rate16,960 tokens/s/userBaseline~4-5x
Cost per million tokens0.75 cents3.79 centsReduced 80%
Power~250W~700WReduced 64%
ProcessTSMC N6TSMC 4nmMore mature
HBM❌ Not used✅ HBM3eLower cost

Technical Principle​

Taalas HC1 uses an aggressive Computing-in-Memory (CIM) implementation:

  • Model weights directly固化 in Mask ROM (fully hardware-defined)
  • On-chip SRAM handles dynamic data (KV cache and LoRA fine-tuning weights)
  • Only 2 mask layers need modification to produce a dedicated chip for another AI model; turning an AI model into a physical chip takes only 2 months

Limitations​

  • Lack of flexibility: Hard-wiring cannot cope with rapidly iterating model updates
  • Ecosystem barrier: The current cloud market still relies on general-purpose platforms; customers may prefer flexible solutions that upgrade with models
  • NRE cost: High one-time engineering cost, requiring sufficient deployment scale to amortize

3.4 Cerebras: The IPO Path of Wafer-Scale Integration​

On May 14, 2026, Cerebras Systems officially listed on NASDAQ, becoming the first wafer-scale AI chip company to go public.

Core Technology: Wafer-Scale Integration (WSI)​

  • WSE-3 (third-gen wafer-scale engine): An entire 12-inch wafer as a single chip
  • 44GB on-chip SRAM: No external HBM, eliminating the memory bandwidth bottleneck
  • 21 PB/s bandwidth: On-chip communication bandwidth, thousands of times that of GPUs
  • Partnership with OpenAI: Signed a 3-year, 750MW, $20B+ compute cooperation agreement

IPO Significance​

Cerebras's listing marks the maturation of the inference-specific chip track:

  1. Capital markets begin pricing such companies
  2. Proves "non-GPU" technical routes have commercial viability
  3. Provides valuation references for other inference chip startups (Groq, SambaNova, Taalas, etc.)

3.5 Inference Chip Landscape: Multiple Technical Routes Coexist​

CompanyTechnical RouteCore AdvantageRepresentative Product
TaalasHard-wired (Mask ROM)Extreme inference efficiency, low costHC1
CerebrasWafer-scale integration (WSI)Ultra-high bandwidth, large-model inferenceWSE-3
GroqSRAM-first architectureDeterministic latency, high throughputLPU (acquired by NVIDIA)
d-MatrixDigital in-memory compute (DIMC)More flexible than hard-wiringCorsair
EtchedHard-wired TransformerWorks for all Transformer modelsSohu
Axelera AIDigital in-memory compute (D-IMC) + RISC-VHigh energy efficiencyMetis AIPU

TrendForce predicts:

  • General-purpose GPUs still dominate training and multi-model environments
  • But in mature, predictable scenarios, general-purpose GPU profit margins will be compressed
  • The industry shifts from general compute monopoly to a dual-track structure of general + specialized coexistence

IV. Overall Domestic AI Chip Landscape in H1 2026​

4.1 Industry Enters Scale-Up Phase​

Metric20252026 Q1Trend
Domestic AI accelerator shipments1.65M units (41% share)—Rising
Total China AI accelerator shipments~4M units——
Hygon revenue growth—Doubled↑
Cambricon revenue growth—+160%↑
Moore Threads revenue growth—Doubled↑

Leading vendors collectively entered the revenue realization channel, moving from "technical validation" to "scale commercialization."

Trend 1: Capitalization Wave Reshapes the Landscape​

  • Late 2025 to early 2026: Moore Threads, Iluvatar listed on the STAR Market
  • Biren listed on the Hong Kong stock exchange
  • Enflame STAR Market IPO accepted
  • Kunlunxin, T-Head initiated listing processes
  • TSINGMICRO, HardyVision and others advancing IPOs

Capitalization brings dual effects:

  • ✅ Positive: Supports R&D and ecosystem building
  • ⚠️ Negative: Valuation bubbles and revenue realization pressure

Trend 2: Capacity Becomes the Biggest Constraint Variable​

The contradiction between explosive domestic AI chip demand and limited advanced-process capacity is sharpening:

VendorAdvanced-process capacity needActually obtained
Huawei Ascend15K wafers/month (7nm-class)Priority guaranteed
SMIC total capacity~20K wafers/month (7nm-class)—
Other vendors~5K wafers/month combinedExtremely tight

Whether stable wafer capacity can be secured directly determines vendor survival. Cambricon's 75.4% inventory-to-revenue ratio is essentially a lock on capacity.

Trend 3: Competition Shifts from "Usable" to "Good"​

Early competition focused on "can it run the model"; now it's about "runtime efficiency, deployment cost":

Dimension"Usable" era"Good" era
Hardware performanceCan it run the modelRuntime efficiency, energy efficiency
Software stackBasic adaptationMaturity, framework breadth
EcosystemExistenceDeveloper community activity
Deployment costInsensitiveCore competitive factor

V. H2 2026 Outlook​

5.1 Upcoming Key Events​

TimeEventImpact
2026 Q3NVIDIA Rubin architecture details revealedNext-gen flagship specs unveiled
2026 Q3Huawei Ascend 950PR/950DT formally launchedNew benchmark for domestic inference chips
2026 Q4AMD MI350X scaled deliveryNVIDIA Blackwell competitor
2026 Q4Cambricon Siyuan 690 launch (est.)New-gen training chip
2027 Q1NVIDIA Blackwell Ultra deliveryNew compute benchmark lands

5.2 Key Competitive Factors Over the Next Three Years​

  1. Wafer capacity access: Advanced-process capacity is a scarce resource; vendors tied to SMIC and TSMC have inherent advantages
  2. Capital operation efficiency: The IPO window is limited; raising enough capital on the market determines R&D sustainability
  3. Software ecosystem depth: Hardware performance is only the entry ticket; software stack maturity, framework adaptation breadth, and developer community activity are the core moat

VI. Conclusion: A Diverse Ecosystem Will Eventually Form​

In H1 2026, the AI chip industry is undergoing a historic transition from "one dominant player" to "pluralistic coexistence."

  • NVIDIA builds higher training barriers with Blackwell Ultra while laying out inference efficiency via the Groq acquisition
  • Huawei Ascend holds the domestic compute baseline with full-stack capability; 950PR begins to surpass H20 in inference
  • Cambricon proves the commercial viability of domestic AI chips by turning profitable first; Siyuan 590 surpasses international rivals in specific scenarios
  • Cerebras, Taalas and other inference-specific chip companies opened a "non-GPU" third route
  • TSINGMICRO's reconfigurable architecture provides a diversified technical route choice for China's AI chips

Over the next three years, the domestic AI chip endgame will form a pluralistic ecosystem where GPU, ASIC, and reconfigurable computing three technical routes coexist, with cloud and edge developing in coordination. "Domestic substitution" is no longer a slogan, but an industrial reality happening now.


Data sources:

  • Xinchuang World "2026 China Domestic AI Chip Vendor Capability Quadrant" (2026-06-16)
  • TrendForce "The Era of Inference Economy: The Rules of AI Chips Are Being Rewritten" (2026-06-04)
  • RayByte "Compute Doubles! NVIDIA Blackwell Ultra Chip Launched" (2026-06-02)
  • Official financial reports and announcements of each company

Related reading: