初创公司尽调
尽调报告 Automotive AI chips / infrastructure-devtools Series A 2026-08-29

Shenji Technology

车载芯片已有真实落地,私募市场定价也有可信锚点;但对外销售经济性和客户多元化仍过于不透明,还不足以给出高确信度买入结论。

Shenji 是真实且具战略价值的汽车芯片平台,但以当前私募估值看,不透明经济性、重度 NIO 依赖和未解的对外商业化风险会封住上行空间。

封面要素

投资建议 06
track [CV019]

公司概况

Shenji Technology(Anhui Shenji Technology Co., Ltd.,品牌也作 GeniTech)是 NIO 剥离出的车载芯片平台,2025 年 6 月在合肥正式注册,承接集团智能驾驶半导体业务。公司目前是一个战略意义很高、但仍由 NIO 控制的芯片平台,覆盖已量产的 NX9031 智驾 SoC、面向具身智能和智能体推理的新变体,以及相邻的感知组件。

成立时间
2025-06-17
创始人
NIO autonomous-driving chip division
创立地点
Hefei, Anhui, China
总部
Hefei, Anhui, China
产品
Shenji 销售并开发以 NX9031 系列为核心的车规级 AI 芯片,主要面向智能驾驶,同时扩展到具身智能、智能体推理和配套边缘 AI 场景的新版算力芯片与平台。
客户
先服务 NIO 和 Onvo 车型项目,同时开始接触第三方 OEM 以及相邻的 AI / 机器人客户。
商业模式
当前业务主要是战略性自用芯片,加上面向 NIO 车型的内部转移定价 / BOM 节省;外部变现仍处早期,形式包括授权和潜在 OEM 芯片供货。
阶段
Series A
融资情况
首轮外部融资 RMB 2.257 billion 于 2026-02-26 完成 / 宣布,投后估值约 RMB 8.27 billion,NIO 仍以 62.7% 持股保持控制。
[CO001, CO003, CO015, CO016, CO017, CO020, CE001, CE021]

执行摘要

主要优势

  • 产品已经在市场中跑通,部署覆盖 NIO 与 Onvo 车型,2026 年出货披露从 150,000+ 台推进到 300,000+ 台。
  • 外部投资人给出了可信独立估值,同时让 NIO 保持控制权,说明 Shenji 被视作可融资的战略资产,而不只是内部成本中心。
  • Shenji 路线图正从单一驾驶芯片扩展到具身智能和智能体推理平台;如果对外商业化成功,确实有可选项。
  • NIO 的全栈软件、整车和系统工程环境,为 Shenji 提供了少见的车规硅内部验证场。

主要风险

  • 独立收入、毛利率、现金消耗,以及按收入计的客户集中度仍未披露,限制估值精度。
  • 可见需求绝大部分仍在 NIO 控制渠道内,客户多元化风险很高。
  • 先进制程制造、存储、封装和出口管制暴露,在公开层面只看到一部分。
  • 除非治理和支持独立性更清楚,外部 OEM 可能不愿依赖一家由竞争车企控制的供应商。
  • 当前估值已经计入相当多的战略可选项;如果对外变现滞后,容错空间更小。

未决问题

  • Shenji 作为独立实体的 P&L、毛利率和现金跑道报告。
  • 按 NIO、Onvo、授权交易对手和潜在外部 OEM 拆分的收入与出货集中度。
  • 面向对外客户的尽调包,覆盖安全验证、质保、支持 SLA 和治理保护。
  • 晶圆代工、封装、存储、EDA 工具和出口管制依赖图谱。
  • 超出授权或管线讨论之外的具名第三方量产客户证据。

目录

Chapter 01

01公司概况

1.1 身份、来源与战略角色

Shenji Technology 更应被理解为 NIO 多年自动驾驶芯片投入的正式商业化外壳,而不是一家常规从零起步的创业公司。公开报道将主体识别为 Anhui Shenji Technology Co., Ltd.,英文也称 GeniTech,2025 年 6 月在合肥注册,经营范围围绕集成电路芯片设计和销售。无论官方材料还是独立报道,对这次剥离的战略目的表述基本一致:把一个昂贵的内部芯片项目放进能融资、能做员工股权激励、并最终服务 NIO 之外买家的结构里。关键在于,芯片项目最初的合理性是纵向整合、控制成本、降低对 Nvidia 的依赖;到 2026 年,叙事已经升级为横跨自动驾驶、具身智能和智能体推理的更广义 AI 硅平台。因此,公开信息已经支撑后续章节的三个核心身份判断:Shenji 由 NIO 控股,根植于 NIO 车型路线图,并正从隐形研发成本中心转向独立半导体资产。[CO001, CO002, CO003, CO015, CO016, CO020]

KPI 快照表
指标数值 / 状态日期 / 期间可信度缺口 / 备注
法律实体公司名:Anhui Shenji Technology Co., Ltd. / GeniTech2025-06英文和中文报道同时使用 Shenji 与 GeniTech 两种称谓。
注册地 / 总部城市中国安徽合肥2025-06-17与 NIO 中国总部同城,据报道注册地址也相同。
阶段NIO 控股拆分公司,已完成首轮外部机构融资2026-02-26不是从第一天起就独立运营的传统创投公司。
最新外部融资RMB 2.257 billion 首轮外部股权融资2026-02-26NIO 官方披露金额。
投后估值官方结构隐含约 RMB 8.27 billion;部分媒体四舍五入至接近 RMB 10 billion2026-02-26 至 2026-02-27不同媒体的换算和取整略有差异。
融资后持股NIO 62.7%;外部投资人 27.3%;ESOP 管理人 10.0%2026-02-26官方持股拆分。
旗舰芯片状态NX9031 已量产;5nm 车规级 AD 芯片2024-2026官方和独立来源均支持其量产状态。
部署规模截至 2026-02 累计出货 150,000+;截至 2026-07 NX9031X 出货 300,000+2026-02 至 2026-07后一个数字适用于 NIO 和 Onvo 的 NX9031X,不一定覆盖 Shenji 所有 SKU。
未披露的关键指标独立收入、ARR、经审计员工数、外部客户数、毛利率2026-08-29 运行日期这些缺项明显限制独立投资判断。

本表混合官方交易事实、公司自称部署指标,以及用于后续投资测算的明确披露缺口。

[CO001, CO003, CO011, CO012, CO015, CO016]
利益相关方 / 投资人图谱
利益相关方角色控制权 / 经济重要性尽调问题
NIO / NIO 子公司控股股东兼内部锚定客户外部融资后保留 62.7% 持股,并继续财务并表。厘清转让定价、关联方收入,以及 NIO 是否能限制第三方销售。
合肥国资资本 / 合肥海恒本地战略投资人为合肥作为拆分公司融资和产业基地提供背书。确认股权资金之外,支持是否还包括政策、场地或采购。
IDG Capital外部财务投资人为首轮融资增加知名创投品牌背书。了解董事会、预留额度和退出预期。
China Fortune-Tech Capital产业投资人支撑其半导体定位,而不只是汽车板块形象。厘清这家投资人是否带来客户介绍或制造生态入口。
股权激励计划管理人人才留存池持有剩余 10% 激励预留。索取归属安排、按职能划分的池规模,以及融资后稀释机制。
潜在外部 OEM(Leapmotor、Geely)潜在客户,而非已确认投资人据报道的供货接触会影响 Shenji 在 NIO 之外的商业化判断。确认洽谈是否推进到项目定点,还是停留在探索阶段。

本表混合股权持有人和潜在客户类别,因为 Shenji 的公司逻辑与商业逻辑高度绑定。

[CO016, CO019, CO024, CO029, CO040]
FO002: 公司快照逻辑

NIO 持股、Shenji 产品、自用部署、外部野心和融资如何拼在一起。

[CO003, CO015, CO016, CO022, CO029, CO030]
FO003: 快照 KPI

截至报告运行日 2026-08-29 的关键公司与产品公开事实。

出货数据来自公司口径,独立经济性仍未披露。

[CO012, CO016, CO017, CO018, CO040, CO041]

1.2 管理层线索与治理透明度

管理层能见度既是 Shenji 的强项,也是缺口。公开来源中最明确的具名经营者是白剑,他被识别为 NIO 芯片与智能硬件负责人,也是 Anhui Shenji 的法定代表人。William Li 仍是战略赞助人,因为 Shenji 仍由 NIO 控股,并反复被放在 NIO 的资本配置和产品战略里解释。NIO 品牌与传播副总裁马麟也在公开场合放大过芯片叙事,尤其围绕 WAIC 2026。缺失的信息同样重要:本报告审阅的来源没有给出清晰的 Shenji 独立 CEO 披露、董事会名单、委员会结构,或外部投资者的详细治理权利。对尽调而言,这意味着投资者可以确认技术和公司层面的支持真实存在,但还无法确认治理厚度。Shenji 因此更像一家被战略性隔离、外部资本化程度上升的子公司,而不是一家完全透明的独立芯片公司。[CO004, CO016, CO019, CO029, CO032, CO040]

管理层与创始人表
人物职务背景 / 公开角色职能覆盖关键人物或治理含义
Bai Jian法定代表人;NIO 芯片与智能硬件负责人OFweek 和 WAIC 前后报道将其列为 Shenji 背后的公开芯片负责人。负责芯片项目的技术推进和实体层面运营。重要性高,因为 Shenji 其他领导层披露很少。
William LiNIO 创始人、董事长兼 CEO公开为芯片战略背书;披露研发成本对比和盈利逻辑。资本配置、战略叙事和母公司支持。重要性高,因为 Shenji 仍由 NIO 控股。
Ma LinNIO 品牌与传播副总裁公开发声者,将 Shenji 定位为中国唯一横跨三个通用 AI 领域并已出货的公司。外部叙事、品类定位和公开市场沟通。中等;更偏传播角色而非运营控制,但在市场定位中能见度高。

公开领导层能见度仍以母公司为中心,独立公司透明度不足。

[CO004, CO029, CO032, CO038, CO040]

1.3 产品基础、部署与规模信号

Shenji 档案里最强的证据是产品已经落地。NIO 和多家独立媒体把 NX9031 描述为 5nm 车规级智能驾驶芯片,拥有超过 50 billion 个晶体管、32 核 CPU 架构、LPDDR5x 支持、546 GB/s 带宽,NIO 称其算力相当于四颗 Nvidia Orin-X。这些说法不只是概念;公开来源把该芯片与 ET9 的真实部署相连,随后又连接到 NIO 和 Onvo 车型的扩展。规模披露在方向上很重要:NIO 称截至 2026 年 2 月,NX9031 累计出货超过 150,000 颗;WAIC 期间的报道则把 NIO 和 Onvo 车型上 NX9031X 的披露装机量提高到 300,000 颗以上。Shenji 还把叙事从驾驶扩展到 NX9031U、NX9031C 和 NX6031。这还不能证明广泛的第三方变现,但足以证明 Shenji 建在真实量产芯片项目之上,而不是只停留在演示文稿里。[CO005, CO006, CO007, CO008, CO009, CO010]

里程碑表
日期事件类型金额 / 估值 / 状态参与方 / 备注含义
2024-07-27NIO 宣布 NX9031 成功流片产品NIO 称为全球首款 5nm 车规级 AD 芯片NIO IN 2024 / 7 月更新确立了后来成为 Shenji 核心的技术资产。
2024-12-21NIO ET9 搭载自研技术栈正式发布产品ET9 发布;计划 2025 年 3 月交付NIO Day 2024把自研芯片战略绑定到旗舰车型商业化。
2025-04-23NIO 在 Auto Shanghai 2025 重点展示包括 Shenji NX9031 在内的全栈技术合作品牌和技术展示NIO 官方活动芯片从后台研发走到前台产品叙事。
2025-06-17Anhui Shenji Technology Co., Ltd. 在合肥注册创立10 million CNY 注册资本Shenji / NIO 公司架构为融资和外部业务创建独立法律实体。
2025-11-19据报道 Shenji NX9031 对外授权规模化技术授权市场打开LatePost,经 CarNewsChina 和 CnEVPost 转引芯片项目从内部成本中心转向收入机会。
2026-02-26宣布首轮外部股权融资融资RMB 2.257 billion;隐含投后估值约 RMB 8.27 billionNIO、Shenji、中国投资人在保留 NIO 控制权的同时,验证了独立融资吸引力。
2026-03-18据报道接触潜在外部 OEM 供货合作推介 M97 >700 TOPS 芯片;目标 Q3 发布Shenji、Axera、Leapmotor、Geely显示其有意在母公司生态之外商业化。
2026-04-11Onvo L90 确认采用 Shenji NX9031规模化子品牌铺开Onvo / NIO把内部部署基盘扩展到 NIO 品牌车型之外。
2026-04-27据报道 Onvo L80 高配车型搭载 Shenji NX9031规模化更多子品牌集成Onvo / NIO巩固其在更新产品周期中的重复部署。
2026-07-17Shenji 首次独立亮相 WAIC,并展示更广产品线治理展示 NX9031X/U/C 和 NX6031;提及近 RMB 3 billion 融资WAIC 2026 上的 Shenji公司从内部汽车芯片团队重新定位为全场景 AI 芯片平台。

里程碑同时覆盖公司设立、产品就绪、融资和生态扩张,因为 Shenji 的身份离不开这条时间线。

[CO001, CO005, CO010, CO015, CO021, CO022]
FO001: 公司里程碑时间线

从 NX9031 流片到融资,再到 WAIC 重新定位的时间线。

日期采用公开里程碑的发布日期或公告日,而非公司内部实际完成节点。

[CO001, CO005, CO015, CO022, CO024, CO026]

1.4 资本化、里程碑与未决风险

2026 年 2 月融资是 Shenji 从内部实验走向外部定价资产的最清晰证明。NIO 官方披露显示,RMB 2.257 billion 新资金注入 Shenji 新发行股份,NIO 保留 62.7% 持股,外部投资者获得 27.3%,另留 10% 股权激励池用于留住人才。二级报道把这换算为约 RMB 8.27 billion 投后估值,一些英文报道则把数字四舍五入到 RMB 10 billion 附近,或约 US$1.5 billion。战略逻辑很直接:一个据报道耗费数十亿人民币的芯片项目,现在可以分担融资压力并寻找外部客户。风险也同样清楚。外部授权到 2025 年底才开始,2026 年 3 月关于与 Leapmotor、Geely 洽谈的报道仍属前瞻;更怀疑的分析认为,如果收入仍主要由 NIO 内部消化,估值可能跑在独立经济性之前。因此,Shenji 进入后续章节时呈现出异常强的技术和战略动能,但独立商业证明仍不完整。[CO015, CO016, CO017, CO018, CO019, CO020]

Chapter 02

02市场分析

2.1 Shenji 实际切入的市场边界

分析 Shenji 时,应把它放在高算力车载 AI 栈里,而不是拿整个汽车半导体市场作参照。边界包括自动驾驶和 ADAS SoC、域控制器、传感器融合算力、具备功能安全能力的车载 AI 处理器,以及软件定义汽车所需的系统集成逻辑。信息娱乐专用 SoC、电池管理芯片、车身控制器、通用 MCU 和大多数非 AI 车规硅都不在边界内。这个区分很关键,因为宽口径汽车 SoC 市场报告常把座舱、连接、动力总成和 ADAS 需求混在一个数字里,而 Shenji 目前的产品证明集中在智能驾驶算力,以及相邻的具身智能雄心。NIO 自身的全栈材料也强化了这一点:公司卖的是系统级承诺,包括自研智能驾驶芯片、高性能感知、域融合和软件集成。因此,Shenji 的可触达市场有意义,但明显窄于泛化的车载芯片 TAM 叙事,并且仍高度取决于中国 OEM 是否想要 Nvidia 级自动驾驶算力的国产替代。[CM001, CM002, CM004, CM011, CM017, CM024]

市场定义表
细分 / 类别纳入口径剔除口径买方 / 付款方对 Shenji 的意义
自动驾驶 SoC用于感知、规划、控制和域融合的车载 AI 处理器通用 MCU、车身控制器、仅座舱芯片OEM ADAS 工程、车型项目、采购NX9031/NX9031X 当前核心市场
ADAS SoC用于 L2/L2+ 安全、传感器融合和驾驶辅助的算力缺乏实质 AI 算力的低端安全芯片OEM 平台团队、安全工程重要重叠市场,也可能是更近的外部路径
软件定义汽车中央计算域 / 区域整合和高性能整车计算非集中式传统 ECUOEM 架构和平台负责人有意义,因为 Shenji 作为全栈域集成的一部分销售
具身智能计算机器人脑的感知、规划和本地推理没有边缘芯片的纯云端 AI 软件机器人 OEM、工业自动化买方NX9031U 释放了相邻信号,但证据仍早期
通用汽车 SoC座舱、连接、信息娱乐和混合车载芯片电芯、电力电子、充电桩广义汽车电子团队对 Shenji 直接锚定 TAM 过宽;只适合作外围上限

本章把投资测算锚定在 ADAS / 自动驾驶计算,而不是最宽泛的整车半导体篮子。

[CM001, CM002, CM004, CM011, CM024, CM034]
FM001: 市场规模测算视角

Shenji 位于更广汽车计算市场逐层收窄的切片中。

这个金字塔是概念性的:中间层保留已发布市场定义之间的冲突,不假装它们可以相加。

[CM006, CM007, CM008, CM009, CM011, CM034]

2.2 规模口径:市场很大,定义差异也大

这一类别的公开市场规模估计差异很大,因为不同发布方测量的东西不同。Future Market Insights 把广义汽车 SoC 市场估为 2026 年约 USD 22.8 billion;DIResearch 对自动驾驶 SoC 的口径约为 2026 年 USD 33.0 billion;Semiconductor Insight 更窄的自动驾驶 SoC 口径只有 USD 6.8 billion。另一个 Semiconductor Insight 口径则把 ADAS SoC 子市场估为 2026 年 USD 9.2 billion。这些数字本身并不必然错误,只是范围定义不同。因此,默认情况下不应拿最宽口径数字给 Shenji 做基准。真正可投资的问题不是汽车算力总量是否巨大,而是未来几年 Shenji 现实能服务哪一块。今天公司明确服务 NIO 和 Onvo 内部的高端 EV 项目,也表达了进入更高算力相邻场景和外部 OEM 授权的兴趣。公开证据尚不足以支撑干净的自下而上 SOM,因此本章保留区间,而不是制造虚假的精确度。[CM006, CM007, CM008, CM009, CM010, CM011]

TAM/SAM/SOM 或规模测算视角表
发布方 / 口径年份地区数值CAGR / 趋势方法口径可信度局限
Future Market Insights – 汽车 SoC2026全球USD 22.8B至 2036 年 7.96%跨应用的广义车载 SoC 类别口径过宽,难以直接映射 Shenji 当前覆盖范围
Future Market Insights – 汽车 SoC 预测2036全球USD 49.0B7.96%长期广义车载 SoC 市场涵盖 Shenji 今天尚未切入的细分
DIResearch / MarketResearch – 自动驾驶 SoC2026全球USD 33.0B至 2033 年 6.14%更宽的自动驾驶 SoC 竞争格局定义似乎比狭义 ADAS 算力口径更宽
Semiconductor Insight – 自动驾驶 SoC2026全球USD 6.8B至 2034 年 11.3%更窄的自动驾驶 SoC 口径公开页面的方法细节有限
Semiconductor Insight – ADAS SoC2026全球USD 9.2B至 2034 年 11.8%ADAS 专用 SoC 市场公开摘要偏营销层,非完整报告
Shenji 现实 SAM(作者推断)2026中国主导的高端 EV / ADAS 计算低于广义汽车 SoC TAM;最接近高算力 ADAS / AV 切片受产品证据和买方集合约束基于 Shenji 当前产品和部署证据推导没有公开的自下而上细分收入或第三方项目定点数据
Shenji 当前 SOM(作者推断)2026NIO / Onvo 内部基盘加早期 OEM 接触公开资料无法量化装机基盘真实,第三方份额不清基于内部车型部署和据报道的外部接触无法用公开数据精确估算

有公开数字的行均使用 USD billion;Shenji SAM/SOM 为定性,因为强行给出伪精确数字会夸大证据质量。

[CM006, CM007, CM008, CM009, CM010, CM011]
FM002: 市场估算区间

已发布的 2026 市场估计,从狭义自动驾驶计算口径到广义汽车计算包络,跨度很大。

只量化已发布市场值;由于公开设计定点和定价数据缺失,Shenji 特定 SAM/SOM 刻意不量化。

[CM006, CM007, CM008, CM009, CM010, CM035]

2.3 买方分层、预算归属与采用路径

Shenji 的近端买方版图异常清晰,因为它从一个受控母公司起步。NIO 和 Onvo 车型项目是即时经济锚点,芯片部署与 ET9、L90、L80 等旗舰和改款车型发布绑定。下一层买方是评估国产高算力 ADAS 芯片的其他中国 OEM 或芯片合作伙伴,Leapmotor、Geely 的接触报道以及 Axera 牵头的 M97 项目体现了这一点。第三层包括机器人、智能体推理和智能计算用户,Shenji 从 NX9031U 和 WAIC 2026 前后开始讨论这些方向,但从收入证明角度看,这些市场仍偏概念。在三层客户里,经济买方都不是零售车主,而是车型项目负责人、ADAS 工程、采购、财务和平台软件负责人混合组成的组织,他们关心算力密度、热约束、功能安全、工具链成熟度和供应安全。这使车载 AI 芯片成为周期长、定点重的市场,组织摩擦远高于消费半导体。[CM012, CM013, CM014, CM015, CM016, CM017]

细分 / 买方图谱
细分买方用户付款方工作流预算负责人采用触发因素
NIO 旗舰车型NIO 车型平台团队ADAS 栈、车辆软件,终端司机间接使用NIO平台更新和发布周期中完成设计导入车型项目 + 研发负责人需要自研算力、成本控制和功能集成
Onvo 子品牌车型Onvo / NIO 共享平台团队大众市场子品牌车型上的 ADAS 栈NIO / Onvo从 Nvidia 栈迁移到 NX9031 栈项目管理 + 采购母公司主导新智能驾驶硬件和软件铺开
外部中国 OEMOEM ADAS 工程 + 采购寻源整车工程团队OEM 资本和 BOM 预算试点、验证、项目定点、SOPCTO / 平台 / 采购性能和安全经济性足够的国产替代
芯片 / 系统合作伙伴半导体或控制器合作方,如 Axera 关联项目联合开发和车型项目集成团队合作伙伴研发 / OEM 项目预算联合开发、流片、验证合资 / 产品负责人需要补齐性能或本地化缺口
具身 AI / 机器人机器人 OEM 和智能设备厂商机器人自主性和控制团队设备制造商或工业客户边缘 AI 评估和集成产品 + 前沿工程需要汽车以外的本地高算力边缘推理

买方、用户和付款方都嵌在漫长的项目定点周期里,这会拖慢采用,但通过资质认证后切换成本也更深。

[CM012, CM013, CM014, CM015, CM016, CM017]
FM003: 买方 / 细分市场图

不同买方细分看重的性能、安全、软件、成本和自主可控组合不同。

[CM013, CM014, CM015, CM017, CM024, CM027]
FM004: 采用漏斗或价值链图

汽车 AI 芯片要先跑完漫长的设计定点和验证步骤,才能形成规模收入。

[CM012, CM013, CM014, CM026, CM028, CM035]

2.4 增长驱动与采用约束

需求驱动力很扎实。多个市场来源都描述了 AI 算力需求的上升:车辆从传统 ADAS 走向软件定义、传感器更丰富的 L2+、L3,并最终迈向 L4 架构。安全监管、域集中化、更复杂的感知栈,以及 OEM 对本土供应链的兴趣,都会推动车规级 AI 算力需求。同时,约束同样重要。汽车芯片要经历多年认证周期,流片和验证成本高,在电动车内部还要承受苛刻的热和功耗要求。Nvidia、Mobileye、Horizon 等既有玩家不只比 TOPS,也比软件工具链、安全认证历史、OEM 集成和支持生态。Shenji 因此享受真实宏观顺风,但市场转化要看其平台能否为 NIO 自身项目之外的买家跨过这些非硅壁垒。这也是为什么公开第三方 SOM 证据仍是市场分析的主要缺口,而不是原始市场规模本身。[CM003, CM018, CM019, CM020, CM021, CM022]

增长驱动与约束表
驱动 / 约束方向时间窗口影响尽调问题
L2+ 到 L4 算力需求抬升驱动因素现在到 2030 年后支撑更高 TOPS 车规级芯片需求核查 TOPS 和内存需求上行时,Shenji 路线图能否守住竞争力
软件定义汽车走向集中式架构驱动因素现在到 2030 年后集成平台厂商比单功能芯片更占优评估 Shenji 能否卖出系统价值,而不只是裸芯片
中国供应链本土化驱动因素当前为本土算力平台带来政策和采购顺风核查外部 OEM 视 Shenji 为战略选择,还是仅限 NIO 专属
功能安全和验证负担约束持续拖慢定点速度,也抬高认证成本审查 ASIL 证据、验证历史和工具链成熟度
功耗、散热和电动车续航取舍约束持续可能限制超高算力芯片采用按车型索取每瓦性能和热设计边界
既有厂商软件生态约束持续Nvidia、Mobileye 和 Horizon 拥有成熟 OEM 集成和软件栈测试 Shenji 在 NIO 内部项目之外能否降低迁移摩擦

各行把需求驱动和执行约束放在一起,因为汽车芯片采用速度不只由市场规模决定。

[CM003, CM018, CM019, CM020, CM021, CM022]

2.5 市场证据仍不能证明什么

现有市场证据足以证明方向,但还不能精确证明变现。它显示软件定义汽车正在提高算力强度,中国 OEM 有战略理由摆脱对外国供应商的依赖,Shenji 也已经通过 NIO 和 Onvo 拿到相当大的内部基础。它还没有证明 Shenji 能在这个生态之外赢下多大比例的更广义可触达市场,也没有说明未来需求中会有多少留在信息娱乐或低算力安全系统等 Shenji 甜区之外的场景,更无法证明具身智能和智能体推理相邻业务会在短期内成为收入池,而不是叙事延伸。做估值时,正确结论是有纪律的乐观:用宽口径市场数字说明品类顺风,但用更窄的 ADAS 和自动驾驶算力切片锚定投资判断。在定点、价格和外部客户转化更清晰之前,Shenji 的市场机会可证明很大,但商业化仍只完成了一部分。[CM010, CM011, CM012, CM030, CM034, CM035]

Chapter 03

03竞争格局

3.1 竞争格局:谁真正与 Shenji 竞争

Shenji 不是在与一个单一的“AI 芯片市场”竞争。它面对的格局至少可以拆成五类有意义的竞争者。第一类是 Nvidia,这个全球既有玩家拥有从训练、仿真到车载算力最完整的端到端栈。第二类是 Mobileye,它用高度集成、低功耗的 ADAS 和自动驾驶硅,与深厚 OEM 关系竞争。第三类是 Horizon Robotics,中国本土最醒目的车载算力专家,正在把 Journey 6 系列推向多个算力档位和量产合作。第四类是 Qualcomm,它通过 Snapdragon Ride 和更广义的 Digital Chassis 切入市场,强调跨域集成。第五类是 Black Sesame,这家中国挑战者依靠 Huashan 系列推动开放生态,并逐步强化全球化定位。Shenji 与这些玩家并列,但来源是垂直整合的内部项目,目标是把 NIO 内部成功转化为更广泛的平台可信度。因此,正确比较不只是 TOPS,而是买方信任、验证历史、软件工具链和可规模化 OEM 采用证据。[CP001, CP002, CP003, CP004, CP005, CP006]

竞争对手画像表
竞争对手类别规模 / 融资 / 证明目标细分市场差异化与 Shenji 相比的短板或反向对比
Shenji起于内部配套的本土挑战者NX9031X 在 NIO 和 Onvo 累计出货 300,000+;NIO 持有多数股权高端乘用车 ADAS 及相邻 AI 算力真实内部量产、垂直整合、中国本土化外部 OEM 定点和公开工具链可见度仍有限
NVIDIA DRIVE全球既有全栈平台获头部车企和 robotaxi 车队选择;覆盖 L2++ 到 L4 栈高端 ADAS、自动驾驶、robotaxi软件栈深,端到端平台覆盖广OEM 寻求本土替代时,对其依赖度更高
Mobileye EyeQ车规专用既有厂商拿下美国大型订单后,未来 EyeQ6H Surround 交付量预计 19M+主流到高端 ADAS 和自动驾驶低功耗车规专长,OEM 嵌入面广相比本土供应商,中国自主可控吸引力较弱
Horizon Robotics Journey中国本土量产专家10+ 家 OEM / 品牌合作;Journey 6 覆盖 10+ 到 560 TOPS中国大众市场到高阶智能驾驶软硬件协同优化,本地部署履历内部整车栈绑定弱于 NIO/Shenji
Qualcomm Snapdragon Ride平台集成商Digital Chassis 关系广,Ride 平台家族完整ADAS 叠加座舱 / 跨域架构跨域整合和生态撬动公开量产定价和外部证明仍有限
Black Sesame Huashan本土开放生态挑战者Huashan A2000 通过美国评审;出现 BYD 采用信号中国智能驾驶项目,潜在走向全球开放生态,本土性能定位进攻性强公开全球量产证明仍落后于 Nvidia 和 Mobileye

各行比较的是战略位置,而不是完全一致的收入规模,因为多数厂商没有公开私有公司和项目级经济性。

[CP001, CP003, CP004, CP005, CP006, CP007]
FP001: 竞争定位图

证据支撑的顺序尺度图:X 轴是汽车 AI 计算的外部商业化验证广度,Y 轴是系统集成深度。

顺序评分汇总公开证据,包括设计定点、集成广度和部署可见度,不等同于经审计市场份额。

[CP003, CP004, CP005, CP006, CP007, CP019]

3.2 既有玩家优势与 Shenji 当前位置

Nvidia 仍是最难跨越的基准,因为它的价值主张在系统层,而不只是芯片层。DRIVE 栈覆盖模型开发、仿真、验证和车载算力,已经被领先车企和 robotaxi 车队选用。Mobileye 的竞争方式不同,它优先押注高效、可扩展的车规级算力,并拿下可能延伸到未来数千万辆车的大型 OEM 项目。Horizon 在中国与 Shenji 最直接竞争,因为它把本土定位、量产规模、软硬件协同优化和覆盖 10+ 到 560 TOPS 的 Journey 6 路线图结合在一起。Qualcomm 则借 Snapdragon Ride / Digital Chassis 框架内座舱和 ADAS 工作负载的融合施压;Black Sesame 以国产开放生态、激进性能宣称和 BYD 线索证明推进。面对这一组对手,Shenji 的核心优势是具体部署:NX9031 系列不是假设,公开来源显示它已在 NIO 和 Onvo 上真实装车。核心弱点是能见度:与既有玩家相比,它在软件栈成熟度、具名外部客户和量产级封装方面提供的公开证据更少。[CP003, CP004, CP005, CP006, CP007, CP008]

功能 / 能力矩阵
采购标准ShenjiNVIDIA DRIVEMobileye EyeQHorizon Journey 6Qualcomm RideBlack Sesame Huashan A2000
内部量产证明NIO/Onvo 内部高全球高全球高中国高公开证据中等公开证据中等
公开软件 / 生态可见度低到中中到高
中国本土自主可控姿态
多域平台叙事
已审阅来源中的具名外部 OEM 定点
公开算力区间披露

单元格是有公开证据支撑的序数判断;衡量的是可见度和竞争姿态,而非绝对技术优越性。

[CP011, CP014, CP015, CP016, CP017, CP018]
FP002: 功能广度 / 证明缺口图

能力图突出 Shenji 在公开外部验证和生态可见度上仍落后于同业的环节。

[CP020, CP021, CP023, CP024, CP027, CP028]

3.3 切换成本、分发能力与定价透明度

这个市场最大的竞争壁垒不是裸硅性能,而是组织切换成本。车载 AI 算力通过漫长设计周期、安全验证、整车平台集成和支持生态销售,买方无法轻易替换。Nvidia 受益于广泛的软件环境和系统级开发工作流,Mobileye 受益于长期嵌入 OEM 项目,Horizon 受益于中国本土量产和架构宽度,Qualcomm 则受益于跨域汽车关系。Shenji 的内部渠道给了它一个异常有力的分发机制:不必先赢下外部 RFP,就能被设计进 NIO 和 Onvo 车型。这让它更容易证明实车运行、成本节省和部署规模。但外部销售仍需要更艰难的跃迁。整个竞争集合的公开定价大多不透明,这本身也很说明问题:买方把 ADAS 芯片当作长周期平台评估,而不是有货架价格的零部件。少量浮出的公开套件数据,如 Qualcomm 开发硬件,也凸显商业化路径仍然复杂且依赖合作伙伴中介。[CP014, CP015, CP016, CP017, CP019, CP021]

定价 / 打包对比
竞争对手公开定价 / 打包信号包含能力未知项 / 限制影响
Shenji量产芯片无公开标价;价值主张围绕降本和集成NIO 栈下的智能驾驶算力和更广芯片家族未披露外部 OEM 商业包外部难以直接对标交易经济性
NVIDIA DRIVE审阅来源未显示平台定价有公开标准车载算力加更广开发栈OEM 谈判经济条款不公开既有优势更多来自生态,而非公开标价
Mobileye项目经济条款不公开;OEM 定点规模已有披露EyeQ 家族提供 ADAS、DMS/OMS 和自动驾驶栈选项审阅来源没有可直接对比的单价商业牵引用定点体现,不靠挂牌价格
Qualcomm RideLantronix 列出 SA8255P/SA8775P 认证开发平台,量产定价仍靠谈判Ride 系 SoC 和跨域开发的调试 / 测试入口开发硬件不等于整车项目定价证明该市场靠伙伴主导的工程推进来销售
Black Sesame / Horizon 同业组公开材料强调 TOPS、生态和 OEM 项目,而非标价车规 HPC 平台和工具链量产定价仍不透明竞争对比应更看证明、安全和集成,而不是表面单价

汽车 AI 芯片通过定点、验证和平台关系销售,因此几乎全组公开定价可见度都很差。

[CP012, CP021, CP022, CP031, CP037]
FP003: 护城河 / 就绪度 KPI

Shenji 与主要竞争对手的公开验证指标浓缩对照。

KPI 项有意混合出货、路线图和方案信号,因为直接收入或份额数据大多不公开。

[CP006, CP009, CP011, CP018, CP020, CP025]

3.4 护城河耐久度、反向证据与 Shenji 可能失手之处

Shenji 今天的护城河真实存在,但仍有条件。最强的一层是 NIO 生态本身:母公司愿意吸收早期设计风险,车型线能产生活跃部署数据,国产叙事也契合中国提高本土半导体含量的方向。下一层是性能和成本叙事,尤其是 NX9031 具备四颗 Orin-X 级算力且能节省单车成本的说法。但几类风险会很快侵蚀这条护城河。如果外部 OEM 更愿意在 Horizon、Black Sesame、Qualcomm 和 Nvidia 之间双源采购,而不是押注一个与 NIO 绑定的供应商,Shenji 的内部优势可能无法外溢。如果买方认为软件成熟度和安全验证工具链比性能更重要,Nvidia 和 Mobileye 仍会占上风。如果 Shenji 向机器人和智能体推理的扩张主要停留在叙事,它可能被视为一个较窄的内部芯片项目,却按独立平台估值。AI Weekly 的反向评论已经指向这一方向,认为 NIO 自身生态之外,更大平台叙事的证明仍不均衡。[CP012, CP018, CP019, CP020, CP023, CP024]

护城河耐久度 / 竞争风险登记表
护城河主张威胁严重性剩余暴露缓释措施 / 尽调问题
NIO/Onvo 内部部署是耐久启动底盘客户群停留在内部,外部 OEM 没跟进要求证明内部规模能转化为第三方管线
本土自主可控定位吸引中国 OEMHorizon 和 Black Sesame 讲同样本土化故事厘清 Shenji 除 NIO 血缘外的独特优势
高算力和降本主张可拿下高端项目既有厂商靠软件和安全验证胜出,而非 TOPS获取第三方基准测试和验证报告
泛 AI 平台叙事扩大 TAM机器人和智能体推理证明仍薄跟踪 NX9031U/C 能否产出具名外部客户
外部 OEM 拓展可拓宽护城河Leapmotor 和 Geely 洽谈可能无法转成 SOP 交易核查试点、定点和量产里程碑状态

登记表关注竞争优势耐久度,而不是一次性技术性能主张。

[CP019, CP020, CP023, CP024, CP025, CP032]

3.5 后续投资判断的实用结论

对后续财务和估值章节,竞争层面的实用结论很直接。Shenji 不需要在所有地方击败所有对手,才会具备战略价值;它需要赢下一个可防守的细分市场,让源自 NIO 的部署证明、国产供应定位和足够的软件支持,为 OEM 重复采用建立信任。最可能的路径不是全球替代 Nvidia,而是在高端辅助驾驶项目和部分具身智能相邻场景里,成为可信的中国优先高算力替代方案。这个结果已经有意义。较差的路径是 Shenji 主要仍是一个内部成本分摊载体,公开芯片宣称始终没有转化为广泛外部客户基础。今天的公开证据处在两端之间。与竞争对手相比,Shenji 的产品真实性足以让人认真研究,但外部商业证明还不足以声称它拥有耐久的多客户护城河。[CP019, CP020, CP024, CP025, CP032, CP033]

Chapter 04

04财务情况

4.1 收入模式与变现面

公开证据显示,Shenji 目前可能通过两条路径变现,但只有一条被清楚证明。第一条是把芯片部署到 NIO 和 Onvo 车型中,经济价值可能体现为内部转移定价、相对 Nvidia 的 BOM 节省,以及对更高毛利智能驾驶功能的支撑。第二条是通过授权或向第三方供货实现外部变现,这条路径到 2025 年底才变得可见,披露仍很薄。CarNewsChina 报道称,单项 IP 授权可能价值数百万美元,系统级 SoC 授权可能达到数亿美元,但这些数字描述的是潜在交易规模,而不是已确认合同收入。这个区分很关键。即便外部商用芯片收入仍小,Shenji 已经具备经济相关性,因为它影响母公司成本结构和技术控制。但在公开报道把内部经济贡献与外部销售拆开之前,应把这门业务视为一个正在商业化的故事,而不是一家变现透明的外售型半导体公司。[CI001, CI002, CI003, CI004, CI016, CI017]

收入来源表
来源机制单位当前数值 / 状态质量尽调问题
内部整车供货芯片内部部署到 NIO 和 Onvo 整车项目按车辆 / 平台真实存在,并随 2026 年车型发布放量中 — 经济意义明确,但未公开拆分要求提供转移定价和内部收入确认政策
相比 Nvidia 的成本节省用 NX9031 替代 Orin-X,降低 BOM 和供应商依赖每车 RMBYicai / NIO 声称每车约 RMB 10,000 成本优势低到中 — 主张未独立审计对照实际整车 BOM 和软件支持成本验证
对外 IP 授权向第三方提供技术或 IP 授权按授权 / 平台2025 年底开始公开出现低 — 引用了交易规模,收入未披露要求查看已签合同和收入确认处理
对外 OEM 芯片供货潜在直接或伙伴促成的非 NIO 整车项目按 SOP 项目审阅的公开证据中仍只属前瞻确认定点、试点订单和量产时间
相邻 AI 算力具身智能、智能体推理和先进制造芯片按设备 / 平台叙事可见,变现未证实获取首批客户名单和定价架构

Shenji 几乎肯定有多条经济路径,但公开数据还不能把已实现收入、战略性降本和未来商业化选项拆开。

[CI002, CI003, CI004, CI016, CI017, CI021]
定价 / 变现表
定价 / 变现信号标价与实际包含能力来源质量影响
相比 Orin-X 每车 RMB 10,000 成本优势声称已实现的经济效果,不是公开标价NX9031 替换既有智能驾驶算力若完全兑现,意味着内部利润率撬动显著
单笔 IP 授权价值数百万 USD指示性外部变现,不是已确认合同有限 IP / 技术授权显示入门级授权经济性有可信路径
系统级 SoC 授权价值数亿 USD指示性外部变现,不是已确认合同更广芯片或系统技术授权外部项目若转化,商用芯片上行空间可能很大
量产芯片无公开标价Unknown商用芯片定价包未披露投资者无法对标 ASP 或折扣
未披露软件 / 支持附加定价Unknown验证、工具链和支持服务可能显著影响毛利率和回本周期

公开定价信号几乎全来自媒体报道和降本主张推断,而不是公开商业价目表。

[CI004, CI017, CI021, CI022, CI033]
FI001: 收入模型桥

R&D、自用部署、成本节省和外部授权如何转化为收入与毛利价值。

[CI002, CI003, CI004, CI017, CI021, CI031]

4.2 单位经济与成本结构

公开单位经济证据最强的部分在成本和资本,而不是收入质量。多个来源称 NX9031 开发耗费数十亿人民币,Yicai 将数字缩窄到约 RMB 2.3-3.0 billion,William Li 则把这项投入类比为建设约 1,500 座换电站。对单颗芯片项目而言,这是一笔巨大的资本投入,也解释了 NIO 后来为什么寻求外部投资者。最有利的对冲是 Shenji 宣称的成本优势:Yicai 报道称,相比 Nvidia Orin-X,单车可节省约 RMB 10,000;如果这种优势能在规模部署中保持,就能显著支撑母公司的整车经济性。缺失的是这两个事实之间的连接。公开来源没有披露 Shenji 的毛利率、晶圆或封装经济性、良率、折旧处理、内部转移定价模型,或持续软件支持成本。因此,投资者可以说芯片项目可能改善成本结构,但还无法量化累计节省何时能在独立口径上超过累计研发投入。[CI008, CI009, CI010, CI017, CI022, CI023]

单位经济性表
指标数值 / 空值置信度为何重要尽调问题
NX9031 累计研发投入RMB 2.3B 至 3.0B界定降本或收入必须覆盖的沉没成本负担要求提供资本化政策和剩余路线图支出
相比 Nvidia 的单车节省RMB 10,000可能为 NIO 销量提供实质利润率支撑对照实际生产组合和软件成本验证
独立毛利率null判断 Shenji 是商用芯片还是战略基础设施,独立毛利率是估值关键要求按内部与外部销售拆分产品级毛利率
独立收入 / ARRnull做任何倍数法承销都需要该指标要求季度运行率和订单积压
外部客户收入占比null显示 NIO 之外的多元化要求拆分内部与第三方销售
未来芯片资本强度高,公开未量化决定后续融资需求要求未来两次流片预算和验证支出

空值是公开证据缺失处的刻意留白;这个缺口本身就是重大尽调信号。

[CI001, CI008, CI017, CI022, CI023, CI024]
FI002: 单位经济性桥接图

公开证据能把高额沉没研发投入和可能的单车节省连起来,但毛利率怎么算仍未披露。

[CI008, CI016, CI017, CI022]
FI003: 财务估算区间

公开口径披露了研发支出和内部成本影响的区间,外部商业化经济性仍大多未知。

授权区间用粗略圆整数桶表达据报道的交易量级,而不是已披露的签约合同金额。

[CI004, CI008, CI017]

4.3 资本充足性与母公司语境

Shenji 2026 年 2 月融资改变了融资叙事,但没有终结融资问题。官方口径下,RMB 2.257 billion 新现金进入 Shenji 新发行股份,NIO 保持 62.7% 控制权并继续财务并表。到 2026 年 7 月,Shenji 称自成立以来累计融资接近 RMB 3 billion。这显然减轻了 NIO 独自资助芯片路线图的负担。但这并不意味着公司已经获得长期独立融资。车规级芯片需要反复跨节点迁移、安全验证、软件维护、客户支持,并且可能还要支持 M97 或具身智能芯片等新品。母公司背景很重要,因为 NIO 同时在 2026 年追求 non-GAAP 经营盈亏平衡、改善整车毛利,并扩大 NIO、ONVO 和 FIREFLY 的交付。这使 Shenji 即便在自身独立指标可见之前,也具备财务战略意义:它可以作为成本控制资产、估值重估催化剂和资本分担机制。[CI005, CI006, CI007, CI011, CI012, CI013]

资本充足性表
指标当前数值 / 状态置信度为何重要尽调追问
首轮外部融资以现金认购新股 RMB 2.257B拉长资金跑道,也验证外部投资人胃口核实资金用途计划和分期到位时间
成立以来累计融资截至 Jul 2026 接近 RMB 3B说明融资能力不只靠母公司支持核对 Feb 轮融资与其他融资来源或补贴
融资后 NIO 控制权62.7%,继续并表意味着母公司仍能主导战略和报表合并范围明确少数股东保护和关联方政策
母公司盈利目标2026 non-GAAP 经营盈亏平衡目标解释为什么分担芯片资金压力对 NIO 重要评估若母公司未达目标,Shenji 能否持续获得资金
独立债务 / 项目融资未找到具体公开披露是下行情景和流动性分析的关键直接询问抵押借款、采购承诺和晶圆代工预付款
近期数据催化截至报告日,Q2 2026 业绩定于 Sep 1, 2026 发布说明母公司最新财务更新仍待披露业绩发布后检查是否提及 Shenji

2026 年资本充足性明显改善,但没有公开证据表明 Shenji 已能完全自筹资金,或不受母公司战略牵动。

[CI005, CI006, CI007, CI013, CI018, CI019]
FI004: 资本强度 / 现金流图

外部融资缓解压力,但后续芯片路线图和验证周期仍让 Shenji 维持高资本强度。

[CI005, CI006, CI018, CI019, CI024, CI026]

4.4 牵引力代理指标、集中度与缺口

Shenji 的公开牵引力数据在方向上有用,但经济含义不完整。出货和部署数字说明芯片项目真实且在放量:截至 2026 年 2 月披露的 NX9031 累计出货超过 150,000 颗,2026 年 4 月 Onvo L90 上市时超过 200,000 颗,WAIC 2026 时 NX9031X 出货超过 300,000 颗。与此同时,NIO 的交付更新说明这个内部生态仍在增加大量月度整车销量,包括 7 月超过 35,000 辆、第二季度超过 107,000 辆。ONVO 强劲销量和改款旗舰发布也能吸收更多自研芯片。但这些既是增长信号,也是集中度信号。可见订单簿仍压倒性地绑定 NIO 控制的品牌,公开来源仍没有提供 Shenji 独立收入、外部客户收入占比、毛利率或经审计员工数。即便 NIO 经审计的 Form 20-F,也只确认母公司层面陈述,而不是 Shenji 自身账目。对尽调而言,这意味着业务拥有强运营代理指标,但披露质量弱。[CI001, CI014, CI015, CI016, CI027, CI029]

公开财务缺口表
未公开的私有指标影响重要性当前替代指标具体尽调路径
独立确认收入缺少它,估值主要靠叙事支撑芯片出货量和授权报道要求提供经审计或董事会层面的收入桥接表
按产品 / 客户类型拆分的毛利率区分战略性成本中心和有吸引力的商用芯片业务母公司整车毛利率和降本说法要求提供产品毛利率瀑布图
外部客户收入占比看 Shenji 是否在摆脱 NIO 集中度仅有对 Leapmotor / Geely 接触的报道要求提供客户集中度表
员工数和烧钱速度资金跑道和经营杠杆分析都需要它外部融资和母公司业绩要求提供当前组织架构、薪资支出和月度烧钱速度
晶圆代工 / 封装承诺可能带来隐性现金需求和营运资本风险先进制程资本强度报道要求提供晶圆、封装和测试承诺

这些不是小缺口;正是这些具体阻碍,让只靠公开信息难以形成清晰的独立投资论证。

[CI001, CI022, CI023, CI024, CI025, CI027]

4.5 财务结论

因此,财务结论是复杂但仍值得进一步尽调。Shenji 并不像上市半导体公司那样财务清晰。它的财务可读性来自 NIO 内部战略资产身份:项目有真实规模、沉没成本高、具备真正的降本潜力,也刚完成外部资本分担。这已经不只是叙事。但证据仍不足以支撑独立经济性判断。外部授权真实到值得纳入考量,但披露不足以建模。母公司盈利改善和销量增长带来有利经营背景,但 Shenji 的精确贡献仍未知。资本充足性已有改善,但如果 Shenji 要成为外售平台,未来芯片和外部客户支持几乎肯定需要超过首轮外部融资的资金。投资者因此应把公开财务证据视为支持性但不完整,并在形成强投资判断之前,要求数据室级别的桥接,说明芯片出货如何转化为确认收入、毛利率和现金需求。[CI018, CI019, CI020, CI021, CI022, CI024]

Chapter 05

05产品与技术

5.1 产品定义与 SKU 图谱

Shenji 的产品已不再只是单颗旗舰自动驾驶芯片。到 2026 年中,公开证据显示,公司围绕使用场景而不只是单一车型项目,排布出一组相关算力产品。装机重心仍是面向 NIO 和 Onvo 车内辅助驾驶的 NX9031 / NX9031X,但 WAIC 2026 把可见产品图谱扩展到面向具身智能和先进制造的 NX9031U、面向智能体推理的 NX9031C,以及作为智能感知芯片的 NX6031。这一点改变了尽调应如何界定业务。Shenji 最适合被理解为一个从 NIO 车载算力基础长出来的领域专用 AI 硅平台,而不是一个一次性的内部料号。不过,公开表面并不均衡。公告、展示材料和架构宣称反复出现,但还没有面向外部客户的公开商用产品目录,也没有标准化零部件规格、价目表、软件文档或支持 SLA。[CE001, CE002, CE003, CE004, CE005, CE006]

产品模块 / 资产矩阵
模块 / 资产主要用户状态 / 成熟度差异化尽调缺口
NX9031 / NX9031X 辅助驾驶 SoCNIO 与 Onvo 智能驾驶栈已量产并部署在已验证的整车栈里提供高算力密度;定位为多 Orin 方案的替代品无公开商用数据表、定价或外部支持 SLA
NX9031U 智能计算芯片具身智能 / 先进制造开发者在 WAIC 2026 公开预览最高 800 TOPS,接口丰富,并针对非车载 AI 场景采用风冷无公开客户名单或量产时间表
NX9031C 智能体推理芯片智能体 / 分布式 AI 工作负载已公开展示将产品族从驾驶延伸到智能体推理未看到公开规格表或部署证据
NX6031 感知芯片车端感知 / 较低端边缘任务在 WAIC 2026 展示将产品线从高端 AD 计算向外扩展公开技术细节很少
M97 联合开发芯片外部 OEM / 更广泛汽车客户据报道已流片;2026 年报道中目标 Q3 发布面向更宽客户群设计,性能等级为 700+ TOPS需确认发布、SDK 和 SOP 客户

Shenji 的产品覆盖面已延伸到多个芯片和平台,但外部产品化证据仍不如内部部署证据成熟。

[CE001, CE003, CE004, CE005, CE006, CE017]
工作流 / 用例表
用户任务当前工作流Shenji 解决方案可衡量收益限制
NIO 旗舰辅助驾驶在高端 EV 内运行感知、规划和控制ET9 及后续车型搭载 NX9031将算力收进自研栈,声称达到 4 倍 Orin 级性能公开证据只在 NIO 生态内最强
Onvo 子品牌铺开将子品牌 ADAS 升级到 NIO World Model 路线L90 和部分 L80 配置搭载 NX9031扩大跨品牌共用硬件 / 软件底座基础配置仍可使用 Nvidia 路线
具身智能开发机器人或制造场景需要本地高算力边缘推理NX9031U 加 Ruidong 平台把车端验证过的算力重新封装给相邻 AI 领域未公开披露量产部署
智能体推理需要面向分布式智能体优化的芯片NX9031C 与分布式智能体平台显示公司尝试在车辆以外变现缺少公开工作负载基准
外部 OEM 采用寻找国产高性能 ADAS 替代方案M97 / 授权动作有望成为成本更低的国产替代,并具自主可控吸引力公开信息尚未确认已签约 SOP 定点

各用例的成熟度证据不同;只有 NIO/Onvo 车端工作流已清楚通过规模验证。

[CE004, CE005, CE006, CE017, CE018, CE020]

5.2 架构与运行流

可见架构以高性能车规级 SoC 为中心,目标是用更集成的算力层替代多芯片辅助驾驶栈。公开描述称 NX9031 采用 5nm 制程,拥有超过 50 billion 个晶体管、32 核 big.LITTLE CPU 架构、最高 8533 Mbps 的 LPDDR5X 支持、546 GB/s 内存带宽、高动态范围 ISP,以及约 1000 TOPS 算力;NIO 将其呈现为可比四颗 Nvidia Orin-X。NIO 也把这颗芯片放进更广的全栈架构里,与智能驾驶芯片、操作系统和智能底盘控制结合。实际运行中,芯片只是垂直整合工作流中的一层:感知、认知、规划、跨域任务调度和整车执行都在其上方或周围。系统视角是优势,因为 NIO 能软硬件一起优化;它也是尽调难点,因为相比公开开发套件、文档或更广泛集成商工具链的竞争生态,Shenji 的公开技术表面更不开放。[CE008, CE009, CE010, CE011, CE012, CE013]

技术 / 运营架构表
层级 / 组件作用依赖风险
车规级 SoC辅助驾驶中央计算先进制程代工和封装链条未公开点名供应或出口政策依赖仍不透明
CPU / NPU / GPU 异构计算运行感知、规划和推理组合编译器 / 运行时 / 工具链未公开外部开发者支持不清晰
内存子系统(LPDDR5X,546 GB/s)支撑 transformer 和感知工作负载高带宽内存和组件供应成本或组件约束可能影响放量
ISP 与传感器输入处理摄像头和 LiDAR 输入传感器栈与域控制器集成公开传感器融合细节仍不完整
车载 OS / 跨域调度将芯片连接到底盘、驾驶和其他域NIO 操作系统和全栈软件团队第三方可移植性可能低于开放生态

架构叙事最强的形态是垂直整合的 NIO 栈,还不是广泛文档化的外部平台。

[CE008, CE009, CE010, CE011, CE012, CE013]
FE001: 产品架构图

Shenji 的公开架构把芯片、平台软件和整车系统集成分层打通,而不是拿出一套独立商用 SDK 栈。

[CE001, CE002, CE003, CE004, CE005, CE011]
FE002: 客户工作流 / 运行流程

产品工作流在 NIO 垂直整合的整车栈内运转,从传感器输入一路到规划与控制。

[CE008, CE009, CE011, CE012, CE014, CE017]

5.3 部署、集成与路线图

部署证明是 Shenji 最清晰的产品优势。这颗芯片不只是完成流片;它已经进入交付车辆和已宣布车型项目。ET9 是最早的旗舰展示,到 2026 年,NIO 官方和二级报道都称 NX9031 系列已部署或正在导入 ES9、ES8、ET 系列改款,以及包括 L90、L80 在内的 ONVO 车型。到 WAIC 2026,Shenji 称 NX9031X 在所有 NIO 和 Onvo 车型上的累计出货超过 300,000 颗。这种规模化路测场景比基准测试幻灯片更有意义,因为它意味着芯片已进入真实 ADAS 工作流、热管理和软件发布验证。路线图也在变宽。Shenji 预览 NX9031U,发布 Ruidong 和分布式智能体平台,并据报道正与 Axera 开发面向更广外部客户的 M97 芯片。开放问题在于,路线图会变成真正的多客户平台,还是主要停留在以 NIO 为中心、带选择性外部变现的技术树。[CE004, CE017, CE018, CE019, CE020, CE021]

路线图 / 发布 / 开发阶段表
日期 / 阶段功能 / 里程碑状态含义来源
Dec 2024NIO Day 让 NX9031 进入公众视野已完成芯片成为 NIO 技术栈的旗舰级关键证据NIO 官方
Apr 2025上海车展将自研芯片列为三项核心智能 EV 组件之一已完成显示其在 NIO 内部的战略核心地位NIO 官方
Apr-Jun 2026Onvo L90/L80 和 ES9 相关铺开采用 Shenji 路线已上市 / 铺开中将硬件底座从旗舰轿车扩展出去CnEVPost / News18A / CarNewsChina
Jul 2026NX9031U、NX9031C、NX6031 和 Ruidong 在 WAIC 2026 展示展示已完成从驾驶 SoC 扩展到更广泛的 AI 平台CnEVPost / News18A
Q3 2026 目标面向更广客户发布 M97 芯片据报道的路线图可能是首个更清晰的商用芯片推进动作CnEVPost / Yahoo Finance

路线图宽度正在显现,但外部客户准备度仍落后于内部部署成熟度。

[CE015, CE017, CE018, CE020, CE021, CE025]
FE004: 产品成熟度 / 能力图

公开证明最强的是已部署的辅助驾驶算力,最弱的是外部工具链和非汽车商业化。

表中高 / 中 / 低反映公开信息层面的证据深度,而不是内部工程评分卡。

[CE003, CE004, CE005, CE006, CE017, CE021]

5.4 差异化与依赖面

Shenji 的差异化在车载硅最常卡住的地方最强:量产证明、与真实整车栈集成,以及经济动机。公开来源一贯把 NX9031 描述为 NIO 降低对 Nvidia 依赖、同时实现高算力密度和更低单车成本的路径。平台叙事也在改善,因为 Shenji 正把车载算力重新包装进具身智能和智能体推理场景。依赖项仍然很多。公开披露没有说明芯片家族背后的晶圆代工厂、封装或长期供货协议。一些报道称,NIO 在首颗芯片报道时没有被列入黑名单,因此仍可接触海外制造伙伴;这条信息有参考价值,但不是耐久保证。与 Nvidia、Qualcomm 和 Horizon 相比,Shenji 公开披露的开发者工具和生态文档也少得多。换句话说,产品在 NIO 栈内看起来很强,但关于可迁移性和第三方赋能的公开证据明显更薄。[CE012, CE018, CE027, CE028, CE029, CE030]

FE003: 关键依赖图

Shenji 的技术成败仍取决于制造、软件集成以及受政策影响的供应输入。

[CE012, CE023, CE028, CE029, CE030, CE034]

5.5 信任、质量与合规面

信任和质量已有部分证据,但文档并不完整。最具体的公开质量信号是车规级制造表述、量产部署,以及至少一篇报道称 NX9031 符合 ASIL-D 风险和安全要求。NIO 更广义的官方材料也强调全栈研发、测试中心、高性能感知单元,以及通用平台和自研芯片平台之间同步的软件发布。这些信号支持一个判断:Shenji 嵌在严肃的系统工程组织里。但这些不能替代产品级文档。本报告审阅的公开材料没有披露详细认证材料、失效率、良率、现场退货统计、公开安全论证文件、网络安全证明,或 Shenji 作为独立供应商面向外部集成商的支持政策。这意味着产品成熟度足以支撑母公司部署,但外部买家若要把 Shenji 当作多车型半导体供应商依赖,仍需要一座连接安全验证、软件维护和长期支持义务的尽调桥。[CE013, CE019, CE023, CE024, CE028, CE032]

信任 / 质量 / 合规表
控制 / 质量信号状态范围缺口
车规级 5nm 量产说法公开反复提及芯片制造定位未披露公开验证材料
量产部署可在 NIO / Onvo 发布和出货说法中看到真实道路运行暴露无公开返修率或可靠性数据
ASIL-D 合规提及Evertiq 在产品报道中提及功能安全信号未看到直接证书或公开安全案例
测试中心和全栈研发说法NIO 官方创新材料提到 8 个测试中心和 12 个技术域母公司工程能力面并非 Shenji 专属质量 KPI 披露
自研 / 通用平台同步发布NIO 称 June 2026 版本可跨芯片平台运行软件维护成熟度未披露 Shenji 独立支持或补丁政策

信任证据存在,但公开证明仍达不到第三方半导体买方在尽调中通常会要求的水平。

[CE013, CE019, CE023, CE024, CE028, CE032]
Chapter 06

06客户情况

6.1 客户基础与分层

Shenji 的客户基础应按控制权和变现方式分层,而不只按终端市场分。第一层、也是被压倒性证明的客户,是 NIO 自有生态:使用 Shenji 芯片作为智能驾驶栈一部分的 NIO 品牌旗舰和改款车型。第二层是 Onvo,它仍受 NIO 控制,但像一个不同的内部渠道,因为它把装机基础扩展到更偏性价比的子品牌。第三层是潜在外部车企客户,例如 Leapmotor 和 Geely;公开证据显示 Shenji 正积极推介并做平台准备,但还没有确认的量产定点。第四层是相邻的非汽车 AI 客户,面向具身智能或智能体推理;这些客户属于路线图叙事的一部分,但目前缺少具名量产用户。这个分层很重要,因为业务已经跨过内部部署的证明门槛,却还没跨过能实质降低客户集中度风险的多元化门槛。[CU001, CU002, CU003, CU004, CU005, CU006]

客户分群表
分群买方 / 用户 / 付款方用例规模收入 / 战略价值缺口
NIO 旗舰品牌NIO 产品和工程团队 / 车辆终端用户 / NIO 母公司高端智能驾驶计算已验证的最大内部客户基础在旗舰层面验证芯片,并支撑垂直整合无按车型拆分的公开收入
Onvo 子品牌Onvo 产品团队 / 家庭 SUV 终端用户 / NIO 控制渠道将智能驾驶栈铺向更大众市场车型2026 年快速扩张的内部渠道扩大装机基础和规模经济无公开选装率或按配置装车披露
潜在外部 OEM外部车企工程和采购团队国产 ADAS 计算替代方案积极接触,但未确认 SOP 定点降低集中度的关键路径需要已签合同和量产时间
外部授权交易对手芯片厂商或平台伙伴技术授权 / IP 变现公开报道至少有一项外部授权关系首个非内部变现信号交易对手名称和续约条款未披露
相邻 AI / 机器人客户机器人、制造或智能体平台开发者具身智能和智能体推理路线图可见,但尚无客户验证长期 TAM 扩展潜力未看到具名量产客户

分群应把内部渠道与第三方客户分开,因为两类采用质量差异很大。

[CU001, CU002, CU004, CU005, CU006, CU015]
FU001: 客户旅程图

Shenji 当前客户旅程先从 NIO 内部开始,延伸到 Onvo,之后才尝试转化外部 OEM 和相邻 AI 用户。

[CU001, CU011, CU013, CU015, CU016, CU020]

6.2 采用轨迹与量产证明

采用轨迹真实存在,但最好通过部署面而不是独立客户数披露来衡量。官方和行业来源显示,截至 2026 年 2 月,NX9031 累计出货超过 150,000 颗;到 2026 年 4 月 Onvo L90 交付更新时超过 200,000 颗;到 WAIC 2026,NX9031X 在 NIO 和 Onvo 车型上的出货超过 300,000 颗。NIO 的交付更新进一步说明这些里程碑为何重要:2026 年母公司生态仍在增加大量月度整车销量,包括 7 月超过 35,000 辆、第二季度超过 107,000 辆。Onvo L90 的发布路径尤其有用,因为它带来量产、展厅和试驾证据,而不只是管理层叙述。官方 Onvo L90 页面也在产品叙事中确认 Shenji 是标准智能驾驶硬件。这些来源合在一起,强有力地证明 Shenji 已经越过原型阶段,进入规模化、面向客户的部署,即便客户集合仍很窄。[CU008, CU009, CU010, CU011, CU012, CU013]

客户增长 / 采用轨迹表
指标数值日期来源置信度含义缺失分母
NX9031 累计出货量150,000+2026-02-26NIO / CnEVPost到融资日,该项目已达到有意义的上路规模未披露按车型拆分的活跃装车基数
NX9031 累计出货量200,000+2026-04-19Onvo 交付报道铺进 Onvo 显著扩大了实地覆盖未拆分 NIO 与 Onvo
NIO 与 Onvo 的 NX9031X 出货量300,000+2026-07-17WAIC 2026 信息确认两个品牌都有量产证据无活跃使用或里程分母
NIO July 2026 交付量35,934 辆2026-08-01NIO IR截至报告日,内部需求引擎仍大July 交付中各配置有多少使用 Shenji 未公开
NIO Q2 2026 交付量107,658 辆2026-07-01NIO IR季度销量支撑芯片持续吸收无按车型拆分的准确芯片搭载比例
Onvo L90 展厅 / 试驾铺开Apr 21 发布前已到达全国门店2026-04-19Onvo / CnEVPost说明面向客户的部署流程已产品化未披露试驾数量

轨迹数据证明部署动能,但不能证明收入质量或外部客户多元化。

[CU008, CU009, CU010, CU012, CU013, CU014]
具名客户证据表
客户分群部署 / 用例量产还是试点结果限制
NIO ET9内部旗舰车型NX9031 智驾栈首次部署到旗舰轿车量产证明 Shenji 已进入顶级车型,不是实验室演示未公开芯片级利用率或满意度指标
NIO ES9 / ES8 换代周期内部旗舰 SUV 系列在改款高端车型中扩大部署量产 / 推广显示多条产品线都押注该架构具体配置搭载率和销量未披露
Onvo L90NIO 控制的子品牌智驾硬件标配,全国展车 / 试驾车铺开,官方产品页已集成量产 / 上市Shenji 进入第二品牌并落到零售端的最强公开证据仍在 NIO 控制渠道内
Onvo L80NIO 控制的子品牌部分 LiDAR 配置搭载 Shenji,并配合 NIO World Model上市路径 / 部分量产证据显示 Shenji 能与多配置策略并存基础配置仍走 Nvidia 路线
外部授权交易对手(未具名)第三方NX9031 技术授权 / 许可有商业化证据,但未具名外部变现的首个信号交易对手及续约经济性未披露
与 Leapmotor / Geely 的讨论潜在外部 OEMM97 供应与合作拓展管线阶段,尚未量产显示面向非 NIO 买家的外部 GTM 真实存在未公开确认已签 SOP 或采购订单

该具名证据表把量产部署与管线、授权证据拆开,避免夸大客户多元化。

[CU011, CU013, CU015, CU016, CU017, CU018]
FU002: 采用 / 部署漏斗

公开客户证据从广泛车辆销量收窄到规模小得多的具名第三方证明。

车辆交付和芯片出货阶段并不是严格转化漏斗;这里的排列是为了展示更宽的内部证明,以及更窄的第三方客户证明。

[CU009, CU010, CU012, CU013, CU016]
FU003: 客户证明矩阵

量产证明在 NIO 控制渠道最强,第三方收入持续性最弱。

矩阵单元格描述公开证据质量,而不是内部 CRM 分数。

[CU002, CU015, CU016, CU018, CU020, CU026]

6.3 外部商业化与扩张

外部商业化已经可见,但仍处早期。最强的公开证明是,NIO 在 2025 年底开始对外进行 Shenji NX9031 技术授权,把至少一部分芯片投入从内部成本中心转成商业产品。到 2026 年 3 月,多家媒体称 Shenji 和 Axera 正向包括 Leapmotor、Geely 在内的车企推介 M97 芯片,并预计由合资公司负责出货。这有意义,因为它显示的是上市路径意图,而不只是开放式战略选择权。与此同时,公开记录仍未达到真正的客户多元化里程碑。本报告审阅的来源没有提供已签署的量产定点、量产预测、外部积压订单,或使用 Shenji 芯片规模化运营的具名 robotaxi 客户。WAIC 2026 把可触达客户叙事扩展到机器人和智能体工作负载,但那里的公开客户证明仍是管线,而不是采用。因此,投资者应把外部客户活动视为有希望的漏斗顶部进展,而不是成熟扩张收入。[CU006, CU015, CU016, CU017, CU018, CU019]

6.4 留存、耐久度与证据质量

留存能见度很弱,尽管部署能见度很强。由于 Shenji 主要嵌在 NIO 控制的渠道里,合同期限、续约机制、NRR、GRR 和客户满意度在公开表面上大多不可见。内部耐久度可以从改款车型线的持续导入中推断,但这不等于透明的留存数据。Onvo L90 的首批用户,以及 Shenji 持续向 NIO 和 Onvo 扩展,表明平台承诺存在;官方产品页面也暗示 Shenji 没有被当作实验性选项。但本报告审阅的来源没有披露芯片级渗透率、按配置版本划分的搭载率、外部授权续约条款,或第三方买家的现场满意度指标。对尽调而言,关键区分在于内部架构承诺的持续性,和来自外部客户的商业留存证据。Shenji 在前者上得分高,在后者上披露很差。[CU011, CU013, CU026, CU027, CU028, CU029]

留存 / 复用 / 满意度表
指标数值 / 空值细分置信度尽调追问
净收入留存率null外部客户要求按客户类型提供队列收入桥
总收入留存率null外部客户要求提供年度续约和流失数据
合同期限null外部授权 / OEM 供货要求提供条款清单或匿名已签协议
配置层级搭载率nullNIO / Onvo 车型项目要求按车型和配置提供搭载比例
客户满意度 / 缺陷返修率nullNIO / Onvo 终端用户和 OEM 客户要求提供现场质量看板和服务工单

公开来源显示 Shenji 持续铺开,但没有披露客户经济性的耐久指标。

[CU026, CU027, CU028, CU029, CU032]

6.5 集中度、扩张循环与采购摩擦

客户集中度仍是 Shenji 最大的客户质量弱点。几乎所有可见量产证明都绑定 NIO 控制的品牌,这带来两类相互牵连的风险。第一,Shenji 的需求前景高度依赖 NIO 自身产品节奏、交付执行,以及继续支持自研芯片路线的意愿。第二,外部 OEM 可能不愿依赖一个由竞争车企控制的供应商,尤其在公开文档、支持政策和治理保护都很薄的情况下。因此,扩张循环清楚但脆弱:先在 NIO 内部证明价值,再延伸到 Onvo,用量产可信度接触外部 OEM,之后才追逐更广义的具身智能客户。这条链上任何一环断裂——外部转化慢、NIO 产品疲软,或买方对内供式治理不舒服——都会让 Shenji 卡在集中度很高的准内供模式里。这不否定已经取得的牵引力,但说明客户故事仍更关于验证和集中,而不是广泛客户耐久度。[CU002, CU006, CU016, CU020, CU021, CU030]

扩张与集中度风险表
扩张驱动因素集中度风险影响尽调路径
NIO 旗舰车型换代依赖母公司产品节奏结合 NIO 车型发布节奏测算未来需求
Onvo 推广第二个内部渠道仍不等于第三方多元化衡量 NIO 品牌与 Onvo 的收入占比
外部 OEM 拓展车企可能抗拒依赖竞争对手控制的供应商要求提供治理条款和客户证明
授权变现交易对手未具名,耐久性难判断审查已签合同和续约结构
具身 AI 扩张路线图可能吸引关注,但短期没有量产需求要求提供具名试点、转化漏斗和董事会材料中的管线

从 NIO 验证走向广泛商用采用,仍是客户质量争论的核心。

[CU002, CU016, CU018, CU019, CU020, CU021]
FU004: 扩张 / 集中度图

扩张循环取决于 Shenji 能否把内部证明转化为外部信任,同时不失去在 NIO 内部的推进势头。

[CU011, CU013, CU016, CU020, CU021, CU033]
Chapter 07

07风险

7.1 按严重程度排序的风险框架

对 Shenji 的最高风险解读,并不是芯片跑不起来。公开证据已经显示真实装车、可观出货和持续投入。更严肃的问题是,Shenji 能否在政策敏感的供应链、高度集中的客户基础和有限的公开商用供应商披露之间,继续规模化交付先进车载硅。对投资者来说,这些风险传导很快:供应商或出口摩擦可能影响路线图节奏,客户集中会放大任何 NIO 放缓,薄弱的第三方文档会拖慢外部 OEM 转化。重要的是,风险栈彼此相连。一家公司可以有强内部验证,同时仍让第三方因治理、支持义务或美国政策悬而未决而在商业上犹豫。因此,Shenji 更不像一个二元技术赌注,更像一层层执行和依赖叠起来的故事;即便证明点在增加,风险负担仍高。[CR001, CR003, CR006, CR013, CR017, CR019]

FR001: 风险热力图

Shenji 最高的剩余风险集中在客户集中度、受出口政策影响的供应,以及商用就绪缺口。

序数热力图来自公开证据,而不是内部 ERM 评分。

[CR006, CR017, CR019, CR022, CR029, CR030]
FR002: 风险传导图

多项风险会传导到同一组终点:路线图放慢、外部转化变弱、毛利信心降低,以及估值下行。

[CR006, CR017, CR019, CR029, CR030, CR032]

7.2 监管与法律风险

到 2026 年,监管和法律风险已经从背景担忧变成现实议题。最明确的催化剂是美国国防部 2026 年 6 月发布 Section 1260H “Chinese military companies” 清单,其中包括 NIO。NIO 随即回应称该清单不是制裁清单,不会影响业务,并保留采取法律行动的权利。两点可以同时成立。该清单未必直接禁止证券交易或现有运营,但它仍会给任何嵌在 NIO 战略栈里的业务带来声誉、采购和尽调阻力。与此同时,BIS 及配套法律指引在 2026 年再次确认,针对中国关联实体的先进计算出口规则仍在运行。Shenji 的具体代工厂和制造链没有公开披露,因此投资者无法清楚标出哪些管制今天适用,哪些只会在未来制度收紧时变得重要。正确结论不是 Shenji 今天被堵住了,而是它的法律和监管暴露很实质,且公开表面只显示了一部分。[CR001, CR002, CR003, CR004, CR005, CR006]

监管 / 法律风险登记表
规则 / 案件管辖区状态可能性严重性缓释措施剩余敞口尽调路径
DoD 1260H 中国军工企业名单纳入 NIO美国截至 Jun 2026 仍生效NIO 称该名单不是制裁名单,并将挑战该认定采购和尽调污名仍可能影响合作伙伴或融资叙事跟踪后续限制、客户反应和法律挑战进展
BIS 先进计算出口许可要求美国 / 跨境已生效,并在 2026 指引中进一步明确目前没有证据显示 Shenji 受阻;部分场景存在逐案许可路径后续收紧或母公司关联可能限制先进制程供给或工具链访问将实际晶圆厂 / EDA / 封装链条映射到 ECCN 与 D:5 敞口
面向商用销售的 IP / 许可安排未披露中国 / 全球公开信息不清低-中NIO 有内部部署证据,且至少有一条授权路径外部 OEM 可能要求比公开证据更强的赔偿承诺、质保条款或 IP 保障要求提供标准客户合同包和 IP 风险备忘录
针对中国相关 EV / AI 企业的跨境政策升级美国 / 盟友现实中的宏观风险中-高多元投资者基础和国内需求有缓冲作用可能缩窄客户池,或让海外合作更复杂在更紧政策条件下压力测试商用扩张策略

严重性排序看的是风险如何传导到产品路线图、客户信任和融资叙事,而不只是法律形式。

[CR001, CR002, CR003, CR004, CR005, CR006]

7.3 运营、质量与供应风险

真实部署部分缓释了运营和质量风险,但披露仍不足。正面看,Shenji 不只是一颗已流片芯片;它已经装进交付车辆,并经历了旗舰和子品牌车型的量产导入。这降低了纯技术不可行的概率。但几项运营风险仍未解决。公开来源没有披露代工关系、封装安排、良率、退货率或长期零部件承诺。2026 年 4 月 Onvo L90 报道提醒我们,内存和高性能零部件价格能让整车经济性波动约 RMB 10,000,这意味着即便技术成功,芯片也可能因 BOM 或供应波动在放量时遇阻。公开安全证据也不完整。一个来源提到 ASIL-D 合规,但报告表面没有公开证书、安全论证或失效率仪表盘。对内部部署而言这也许可接受;对外部 OEM 采用而言,这是实打实的尽调阻塞点。[CR011, CR012, CR013, CR014, CR015, CR016]

运营 / 质量 / 安全风险登记表
失效模式可能性严重性缓释成熟度剩余敞口未解决缺口
先进制程供给或封装瓶颈低-中晶圆厂和封装交易对手未披露
存储 / 元器件涨价推高平台成本中-高公开证据只是零散片段,不是完整成本路线图
面向外部买家的公开安全 / 可靠性文档不足未公开安全案例、故障率看板或支持 SLA
软硬件集成移植到 NIO 栈之外时失效低-中可移植性证据远弱于内部部署证据
安全 / 支持流程不透明未看到公开的 Shenji 独立网络安全或补丁支持政策

真实部署降低了产品彻底失败的二元风险,但商用供应商准备度仍缺乏充分文档。

[CR011, CR012, CR013, CR014, CR015, CR016]

7.4 依赖、财务与商业风险

商业和财务风险主要由集中度和资本强度主导。需求基础仍压倒性地在 NIO 和 Onvo 内部,这意味着 Shenji 依赖母公司车型节奏、母公司资本配置,以及母公司继续优先押注自研芯片的意愿。客户问题因此与财务问题无法分开。NIO 在 2026 年以 non-GAAP 经营盈亏平衡为目标,同时推出多个品牌和车型,所以 Shenji 路线图仍暴露在更广的母公司运营压力下。2026 年 2 月融资轮减轻了即时资本负担,但没有消除未来流片、验证成本或外部客户支持需求继续要钱的风险。商用化风险进一步叠加。外部授权存在,M97 接触也真实,但本报告审阅的来源没有确认第三方 SOP 定点。如果 Shenji 无法转化外部客户,其经济性就仍会高度依赖单一公司家族和仍不透明的内部转移定价模型。[CR017, CR018, CR019, CR020, CR021, CR022]

合作伙伴 / 依赖风险登记表
依赖交易对手角色集中度失败情景严重性缓释措施剩余敞口
母公司需求集中NIO / Onvo主要客户生态极高NIO 销量疲软或战略转向降低芯片消化量当前已跨品牌铺开仍集中在同一企业集团内
外部 OEM 转化Leapmotor / Geely / 其他潜在客户商用扩张路径洽谈未转化为 SOP 定点授权和 M97 拓展形成初始管线仍无公开量产客户
合资商业化路径Axera / Chongqing 合资公司M97 出货和伙伴销售路径协调不顺或路线图错位推迟外部供货中-高已有联合开发和公开推介活动合资公司经济性和义务未披露
上游先进计算供给未披露的晶圆厂 / 存储 / 封装供应商制造落地支撑地缘政治或产能冲击推迟供应国内定位和当前运营实际供应商版图不透明
政策敏感市场准入与美国相关的交易对手 / 全球合作伙伴潜在海外或受监管买家CMC 或出口管制顾虑抬高采购摩擦中-高NIO 称当前名单并非基于制裁认知拖累可能持续

依赖按其传导到收入质量、路线图节奏或估值信心的速度排序。

[CR006, CR017, CR018, CR022, CR023, CR029]
人才 / 执行风险登记表
角色 / 职能依赖或缺口可能性严重性缓释措施尽调路径
芯片架构与验证人才专门半导体知识集中在一家相对年轻的拆分公司内部有 NIO 全栈工程基础和此前团队规模支撑要求提供组织架构图、关键人留任条款和继任计划
商业化领导力虽起源于内部体系,仍需要卖给外部 OEMM97 拓展和授权显示已有进展询问外部 GTM、支持和定价权由谁负责
安全 / 质量负责人需要把内部验证转成商用级保证真实部署形成运营反馈闭环要求提供具名质量负责人、审计流程和故障复盘节奏
治理与少数股东利益一致性外部融资后 NIO 仍保留控制权中-高外部投资者和 ESOP 带来一定制衡审查董事会权利、关联交易政策和客户冲突保护
范围管理同时扩到智驾、具身 AI 和智能体推理,可能拉扯执行中-高平台复用可能有帮助要求提供路线图优先级和资源分配框架

执行风险偏高,因为 Shenji 同时要做产品规模化和商业模式转型。

[CR018, CR020, CR022, CR023, CR024, CR025]
FR003: 依赖图

Shenji 同时依赖监管方、母公司渠道、上游供应,以及合作伙伴协助的商业化。

[CR004, CR006, CR017, CR023, CR029, CR033]

7.5 缓释因素、监测项与否决标准

Shenji 并非没有缓释因素。外部投资者已经验证这项资产可以融资,NIO 的全栈组织提供内部部署杠杆,出货里程碑也说明技术已经跨过有意义的真实世界门槛。但缓释成熟度并不均衡。很多缓释因素对内部自用最强,对外售扩张较弱。例如,全栈集成有助于 NIO 车内产品表现,但不能单独解决竞争 OEM 对治理的不适。同样,大出货量证明规模,但不能证明外部收入耐久,也不能免疫未来出口限制。有纪律的投资者因此应监测明确否决标准:没有具名第三方量产客户;美国围绕中国关联先进计算的政策姿态恶化;安全或支持文档不足以支撑外部 OEM;以及 NIO 自身经营优先级挤压 Shenji 路线图。公司的上行空间取决于能否把证明转化为可迁移性;否决标准基本都跟踪这一转化失败。[CR016, CR021, CR024, CR025, CR028, CR033]

缓释与止损标准表
风险可监控触发点阈值 / 事件行动含义
外部商业化失败无具名第三方量产客户到下一个重大路线图周期仍没有具名 SOP 客户下调商用平台投资逻辑;仅把 Shenji 视为内部战略资产
政策升级与 NIO / 先进计算相关的新制裁、出口收紧或供应商限制对供应链、工具链或客户采购形成直接限制上调风险评级,并重新评估供应韧性
安全 / 支持缺口OEM 尽调否决 Shenji 文档或支持承诺出现试点失败证据,或缺少安全包暂停投资判断,直到验证包完成审阅
母公司集中NIO 交付或盈利未达预期,显著削弱内部需求支持交付持续疲软或母公司现金承压按集中单一客户供应商测算下行情景
执行过度拉伸路线图扩张快于已验证客户转化多条新芯片线缺乏外部定点或支持基础设施下调估值倍数,并要求路线图收敛

止损标准强调可观察事件:这些事件会打破从内部验证走向多元商用可信度的转型。

[CR004, CR006, CR018, CR020, CR029, CR032]
Chapter 08

08估值

8.1 投资逻辑与价格

核心估值问题不是 Shenji 是否是真资产——它是。公开资料支持这样一幅图景:这是一家已融资、受 NIO 控制的拆分公司,产品已进入量产部署, 芯片路线图在拉宽,也有足够外部投资需求给出独立估值。更难的问题是,这个估值今天还能不能给新投资人留下有吸引力的上行空间。一边, Shenji 拥有多数年轻汽车芯片故事缺少的三件事:真实出货、足够大的母公司生态来消化产品,以及向具身智能和智能体推理扩展平台的可信路径。 另一边,当前公开估值已经吸收了大部分叙事,却没有给投资人留下独立收入、毛利率、按收入划分的客户集中度、外部合同耐久性等基本投资测算输入。 合理姿态因此不是盲目兴奋,而是盯住价格的持续观察。[CV001, CV002, CV003, CV004, CV005, CV006]

建议摘要表
建议置信度风险评级估值立场决策含义
观察合理至偏满持续跟踪该资产,但在按当前估值或更高估值下注前,必须在数据室看到经济性和外部客户证据

该建议对价格敏感:只靠叙事,Shenji 吸引力有限;若披露更充分或进入价格更低,吸引力会明显提高。

[CV018, CV019, CV020, CV021]
投资逻辑 / 反向逻辑表
论点什么会改变判断
真实汽车部署、强母公司生态和扩展后的芯片路线图支撑战略价值具名外部 OEM 定点和商用指标会提高信心
当前官方估值可信,因为外部投资者已在 February 2026 验证这项资产若出现融资过度、外部转化不佳或政策收紧的证据,估值会被削弱
Shenji 可能走出 NIO 车辆,成为更广的 AI 芯片平台若无法在自有场景之外变现,平台可选性会下降
公开披露仍太弱,还撑不起直接投资结论若有独立收入、毛利率和集中度数据,立场可以更积极

乐观情景主要看真实验证能否转成商业经济性;反向逻辑是,这一转化可能停住。

[CV004, CV006, CV007, CV008, CV009, CV010]
FV001: 建议逻辑

真实证明支持继续关注,但信息不透明和客户集中度限制了对当前估值的信心。

[CV001, CV002, CV004, CV006, CV007, CV008]

8.2 当前估值背景与证据充分性

最好的估值锚不是媒体最乐观的简写,而是 2026 年 2 月官方交易。NIO 和 Shenji 披露了 RMB 2.257 billion 的新增现金投资,以及融资后 NIO 持股 62.7%、外部投资人 27.3%、10% 留作激励结构。CnEVPost 和 AInvest 都明确从这个结构推导出约 RMB 8.27 billion 的隐含投后估值。 Yicai 和其他二级报道把业务四舍五入到近 RMB 10 billion,或约 $1.5 billion,方向一致但精度较低。这个差异重要,因为它决定入场纪律。 约 RMB 8.27 billion 的估值看上去还能作为战略资产估值自洽;如果按更高的取整数字买入,投资人就更重地押注平台选择权和未来外部商业化。 问题在于,公开证据尚未把出货、战略重要性和独立外销芯片经济性串起来。因此,投资人有一个可信估值锚,但还没有足够证据支持激进的倍数扩张。[CV001, CV002, CV003, CV004, CV005, CV006]

8.3 乐观、基准与悲观情景区间

在这里,用情景法比假装精确的收入倍数更诚实。悲观情景是假设 Shenji 主要仍服务母公司,无法转化具名外部 OEM,并继续承受政策或供应悬压; 这时业务应低于当前私有市场估值,因为战略价值仍在,但外销选择权会收缩。基准情景是假设 Shenji 继续扩大内部部署,增加有限授权或试点收入, 并维持当前政策条件;在这个结果下,当前官方估值大致合理。乐观情景是假设 Shenji 拿下具名外部汽车客户,证明自己具备外销级支持和安全文档能力, 并显示更宽 AI 平台叙事能在 NIO 之外变现。这种情景可支撑高于当前估值的实质重估,但前提是证据先改善。简言之,当前价格最容易被解释为宽区间的中点, 而不是显然便宜的入场点。[CV014, CV015, CV016, CV017, CV024, CV025]

乐观 / 基准 / 悲观情景表
情景假设估值 / 回报逻辑关键风险概率信号
悲观Shenji 仍以自用为主,没有具名第三方量产客户,政策压力持续估计 RMB 5.5B-7.0B;较官方 ~RMB 8.27B 估值大约 -34% 至 -15%集中度、出口政策敏感性、商业化转化失败若外部 GTM 卡住或 NIO 优先级转向,情景成立
基准内部部署继续累积,授权推进缓慢,没有重大政策冲击估计 RMB 8.0B-9.5B;较官方估值大约 -3% 至 +15%披露仍不透明,多元化仍有限最符合当前公开证据
乐观拿下具名外部 OEM,配齐商业级支持包,并在汽车之外拓展 AI 变现估计 RMB 11.0B-13.0B;较官方估值大约 +33% 至 +57%执行复杂度和竞争仍会左右结果需要证据变强,不能只重复叙事

这些情景估计锚定当前私有估值和证据质量,不是 DCF 输出。

[CV002, CV014, CV015, CV016, CV017, CV024]
投资逻辑破裂与叫停触发表
触发项阈值对投资逻辑的传导行动含义
没有具名第三方量产客户到下一轮重大路线图周期仍未出现商业化可选性的乐观情景大幅削弱把 Shenji 视为战略性自用资产;下调上行情景假设
政策或供应收紧先进制程供应、设备工具或买方采购遭遇新的直接限制资本、时间表和客户转化风险上升上调风险评级;下调估值区间
经济性尽调薄弱管理层无法提供独立收入、毛利率或集中度桥接投资逻辑失去财务支撑证据改善前维持观察 / 放弃
母公司优先级冲突NIO 经营压力或产品节奏明显压低 Shenji 路线图优先级内部验证累积和外部转化潜力都会下降下调基准情景,并拉长兑现时间
支持 / 安全包不足外部 OEM 不接受文档或支持承诺商业化转型被卡住不把外部客户情景写入估值

这些触发项用来检验 Shenji 能否从战略验证升级为可投资的商业经济性。

[CV010, CV024, CV029, CV034, CV035, CV036]
FV002: 估值敏感性

少数未知项对 Shenji 估值的影响,超过继续重复叙事能带来的信息增量。

影响条仅作序数排序,说明哪些因素最能从公开证据层面改变投资判断。

[CV006, CV007, CV010, CV024, CV029, CV033]
FV003: 估值 / 回报区间

官方定价接近宽情景区间的中点,而不是明显遇冷或明显狂热的水平。

区间是基于当前定价和证据质量锚定的情景估计;不是根据已披露收入或现金流数据推导出的结果,因为这些输入并未公开。

[CV002, CV015, CV016, CV017]

8.4 可比公司、退出就绪度与最终判断

可比分析的主要作用,是说明 Shenji 不是什么。它还不是 Mobileye、Nvidia 或 Qualcomm 这样按上市公司方式披露的芯片公司;后者定期提供审计或上市公司文件, 生态能见度也深得多。它也不只是 NIO 的一个运营部门,因为外部投资人已经给它单独定价。正确结论是:Shenji 应被视为一家战略型私有半导体平台, 有真实证据,但外销成熟度不完整。这对应的是观察,而不是投资或放弃。放弃会忽视真实资产质量;直接投资则会夸大当前关于经济性和多元化的证据。 退出就绪度也有限:在 Shenji 披露更清晰的独立指标,或拿下具名外部量产客户之前,没有强公开依据可以假设估值快速上调、IPO 准备就绪或老股流动性容易实现。 下一步信心大概率来自尽调,不来自讲故事。[CV007, CV008, CV009, CV018, CV019, CV020]

可比估值表
可比对象指标倍数 / 估值 / 状态参考意义局限
Shenji Feb 2026 轮融资私募投后估值~RMB 8.27B 隐含官方估值主要估值锚未披露独立损益
Mobileye公开 ADAS/AV 同业,有 20-F 披露面上市同业,财报经审计最接近汽车智能驾驶栈敞口的上市类别业务组合不同,披露质量高得多
Nvidia DRIVE 平台 / Nvidia全球 AV 计算龙头,有 10-K 披露超大市值、多元化上市基准,不是私有芯片可比公司体现全栈车载计算生态的战略价值规模和业务组合不可比
Qualcomm 汽车业务 / Ride多元化上市半导体基准,有 10-K 披露用来校准汽车计算可信度和生态深度的公开基准适合比较生态和商业平台化能力汽车只是 Qualcomm 的一个业务板块
NIO 母公司 SOTP 视角母公司战略资产价值公开市场可能已通过 NIO 部分计入 Shenji,但切分不干净NIO 仍控制 62.7% 并并表,因此值得纳入母公司估值无法单独拆出 Shenji 经济性

这张表更适合做三角校准和披露对照,不适合推导精确市场倍数。

[CV001, CV002, CV022, CV023, CV030, CV031]
最终尽调问题表
主题缺失证据重要性负责人或尽调路径
独立收入和毛利率季度收入桥接,自用与外部销售分拆毛利率只有这些数据,才能把 Shenji 当作业务估值,而不是战略叙事在 NDA 下索取 CFO 或董事会包报告
客户集中度和合同按 NIO、Onvo、授权和外部 OEM 拆分的收入 / 单位份额决定当前估值是否已具备多元化支撑,还是仍停留在自用资产索取客户集中度明细表和合同摘要
商业支持就绪度安全包、验证档案、支持 SLA、质保和赔偿条款外部 OEM 转化靠这些材料支撑索取客户尽调包
供应链韧性晶圆代工、封装、内存和出口管制敞口图政策敏感型硬件业务可能败在依赖链上索取运营和法律风险备忘录
治理和投资者保护董事会权利、关联方政策、少数股东保护、冲突管理公平评估 NIO 控制的供应商,必须先看这些安排索取治理文件和附函摘要

如果不完成这些追问,新投资者大多只是在承保可选性,而不是透明经济性。

[CV007, CV010, CV012, CV033, CV038, CV039]
FV004: 投资 KPI

一页 IC 式读数凸显 Shenji 为什么值得关注,但还不足以轻松点头。

[CV002, CV004, CV006, CV007, CV008, CV010]

免责声明

本报告是基于截至 2026-08-29 公开信息生成的 AI 辅助尽调摘要,不构成投资建议。Shenji 是一家私营、近期分拆的半导体公司,独立披露有限,因此关键的经济性、供应链、治理和客户合同细节仍无法获得,或只能从公开来源间接推断。

证据索引

结论
编号陈述可信度来源
CO001 Anhui Shenji Technology Co., Ltd. was registered in Hefei on June 17, 2025 with initial registered capital of RMB 10 million. SO004
CO002 Public reporting says Shenji's registered business scope includes integrated-circuit chip design and sales and that its address is aligned with NIO China's headquarters in Hefei. SO004
CO003 NIO officially describes GeniTech Co., Ltd. (Shenji) as the subsidiary primarily responsible for its intelligent-driving-chip business. SO001, SO002
CO004 Bai Jian, NIO's head of chips and intelligent hardware, is the publicly named legal representative of Shenji. SO004, SO022
CO005 NIO announced successful tape-out of the NX9031 in July 2024 and later tied that chip to its flagship ET9 launch narrative. SO006, SO021
CO006 NIO describes the NX9031 as the world's first automotive-grade 5nm advanced smart-driving chip. SO005, SO019
CO007 Independent and semi-official reporting describe NX9031 as having more than 50 billion transistors and a 32-core big.LITTLE CPU architecture. SO018, SO021
CO008 Reviewed sources describe NX9031 as supporting LPDDR5x memory up to 8533 Mbps, 546 GB/s of bandwidth, and image-processing latency below 5 milliseconds. SO003, SO021
CO009 NIO says one NX9031 delivers about 1000 TOPS or real-world compute roughly equivalent to four Nvidia Orin-X chips. SO003, SO021
CO010 Public sources tie NX9031 vehicle deployment to the ET9 and other NIO models, showing that the chip moved from R&D to commercial vehicle use. SO004, SO016
CO011 NIO said cumulative shipments of the Shenji NX9031 had exceeded 150,000 units by February 2026. SO001, SO002
CO012 By WAIC 2026, Shenji said cumulative shipments of the NX9031X across NIO and Onvo had exceeded 300,000 units. SO009, SO023
CO013 The NX9031U is a 5nm automotive-grade intelligent-computing chip positioned for embodied intelligence, intelligent computing, and advanced manufacturing with up to 800 TOPS of equivalent compute. SO009, SO010
CO014 WAIC 2026 coverage says Shenji also displayed the NX9031C and an NX6031 intelligent-sensing chip alongside NX9031X and NX9031U. SO009, SO022
CO015 On February 26, 2026, NIO and Shenji announced definitive agreements for RMB 2.257 billion of cash investment into newly issued Shenji shares. SO001, SO002
CO016 After completion of the investment transaction, a NIO subsidiary would retain 62.7% of Shenji, outside investors would hold 27.3%, and share-incentive-plan administrators would hold 10.0%. SO001, SO003
CO017 Using the official financing terms, Shenji's post-money valuation works out to about RMB 8.27 billion. SO001, SO002
CO018 Yicai characterized the same financing round as valuing Shenji at nearly RMB 10 billion, or about US$1.5 billion. SO003
CO019 Secondary coverage consistently names Hefei state-backed capital, IDG Capital, and China Fortune-Tech Capital among Shenji's first outside investors. SO002, SO003, SO020
CO020 NIO said the fresh capital would support development and promotion of high-end chips and strengthen its long-term strategy in autonomous driving and embodied intelligence. SO001, SO003
CO021 At WAIC 2026, Shenji said it had completed nearly RMB 3 billion in financing since being established in June 2025. SO009, SO022
CO022 Late-2025 coverage reported that NX9031 technology had entered external licensing, changing the chip program from pure internal cost center toward potential outside revenue. SO015, SO028
CO023 Chip-industry sources cited by CarNewsChina said a single NX9031 IP authorization could be worth several million US dollars while a system-level SoC authorization could reach hundreds of millions. SO015
CO024 In March 2026, CnEVPost reported that Shenji and Axera were pitching a new M97 smart-driving chip to automakers including Leapmotor and Geely. SO011
CO025 The reported M97 chip was described as having more than 700 TOPS and being positioned against Horizon Robotics' Journey J6P, with a third-quarter 2026 release target. SO011
CO026 Onvo confirmed in April 2026 that its updated L90 SUV would use the Shenji NX9031 together with the latest Nio World Model. SO012, SO008
CO027 CnEVPost reported later in April 2026 that high-end trims of the Onvo L80 would also use Shenji NX9031. SO013
CO028 CarNewsChina reported additional deployment momentum through the ET5 refresh and ES9 launch, indicating that Shenji was spreading across multiple NIO vehicle lines. SO014, SO017
CO029 Morgan Stanley framed GeniTech as an increasingly visible AI-chip-platform call option that could help move NIO beyond the market's cash-burning-EV narrative. SO014, SO029
CO030 AInvest argued the funding round is tactically positive but could still become a valuation trap if Shenji cannot turn strategic narrative into standalone profitability. SO025
CO031 Procurement Magazine framed Shenji as part of NIO's chip-insourcing strategy to reduce dependence on external suppliers such as Nvidia. SO024
CO032 William Li previously said NX9031 R&D cost billions of yuan, roughly comparable to building 1,500 battery-swap stations. SO002, SO009
CO033 Yicai reported that NX9031 R&D cost was roughly RMB 2.3 billion to RMB 3 billion. SO003
CO034 Yicai reported that NX9031 provides about RMB 10,000 of per-vehicle cost advantage versus Nvidia Orin-X. SO003
CO035 NIO reported 326,028 full-year vehicle deliveries in 2025, giving Shenji a large internal deployment base relative to most startup chip programs. SO026
CO036 NIO's March 2026 guidance targeted full-year non-GAAP operating-profit break-even, reinforcing why externalizing part of the chip program's capital burden mattered strategically. SO026, SO027
CO037 By mid-2026, public commentary increasingly described Shenji as evolving from captive automotive silicon into a broader AI silicon platform. SO014, SO028
CO038 Shenji's public domain coverage at WAIC 2026 was autonomous driving, embodied intelligence, and agent inference. SO009, SO022
CO039 AI Weekly said Shenji's deployment proof is strongest inside NIO's own automotive ecosystem and that evidence for the broader platform pitch remains uneven. SO023
CO040 The 10 percent share-incentive pool indicates Shenji is being structured as a standalone talent-retention vehicle rather than only as an internal cost center. SO001, SO025
CO041 Reviewed public sources do not disclose Shenji's standalone revenue, ARR, external customer count, audited headcount, or gross margin. SO001, SO002, SO025
CO042 Shenji made its first independent public appearance at WAIC 2026 rather than appearing only as an internal NIO technology project. SO009, SO022
CO043 NIO's official event materials in 2025 and 2026 repeatedly elevated Shenji/NX9031 as part of the parent company's full-stack technology message. SO007, SO008
CM001 Shenji's current market belongs to autonomous-driving and ADAS compute rather than the entire automotive semiconductor stack. SM011, SM020
CM002 A Shenji-relevant market boundary includes high-compute ADAS and autonomous-driving SoCs, domain controllers, and sensor-fusion compute, but excludes infotainment-only SoCs, body controllers, and most non-AI vehicle silicon. SM001, SM011
CM003 Future Markets Inc says Level 2+ systems typically require 30-100 TOPS, Level 3 around 100-250 TOPS, and Level 4 about 250-1,000+ TOPS of AI inference capability. SM003, SM007
CM004 Public market sources describe software-defined vehicles as shifting compute from fragmented embedded controllers toward centralized, AI-heavy vehicle architectures. SM001, SM007
CM005 Future Market Insights expects China to hold the largest country share of automotive SoC demand as local EV makers expand domestic chip use. SM001
CM006 Future Market Insights estimates the global automotive SoC market at USD 22.8 billion in 2026. SM001
CM007 Future Market Insights projects the broad automotive SoC market to reach about USD 49.0 billion by 2036. SM001
CM008 DIResearch's public summary estimates the global autonomous-driving-SoC market at about USD 33.0 billion in 2026. SM002
CM009 DIResearch projects the global autonomous-driving-SoC market to reach about USD 50.1 billion by 2033 at a 6.14 percent CAGR from 2026. SM002
CM010 Semiconductor Insight estimates the autonomous-driving-SoC market at USD 6.8 billion in 2026 and USD 15.9 billion by 2034. SM004
CM011 Semiconductor Insight estimates the ADAS SoC market at USD 9.2 billion in 2026 and USD 22.5 billion by 2034. SM005
CM012 The spread between USD 6.8 billion, USD 9.2 billion, USD 22.8 billion, and USD 33.0 billion shows that public automotive-AI TAM depends heavily on scope definition. SM001, SM002, SM004, SM005
CM013 Shenji's realistic SAM today is narrower than the broad automotive SoC market and is closest to high-compute ADAS and autonomous-driving programs that need 200+ TOPS-class vehicle AI compute. SM001, SM003, SM021
CM014 Shenji's current SOM is primarily its captive installed base within NIO and Onvo plus a small set of publicly reported third-party outreach efforts. SM013, SM014, SM015, SM017
CM015 NIO and Onvo vehicle programs are Shenji's primary near-term buyers. SM012, SM014
CM016 Leapmotor and Geely were the only external OEMs publicly named in reviewed 2026 reporting as prospective Shenji chip customers. SM013
CM017 NX9031U expands Shenji's stated target market into embodied intelligence, intelligent computing, and advanced manufacturing. SM016, SM021
CM018 For automotive AI chips, the economic buyer is usually a vehicle program, ADAS engineering, sourcing, and platform-software decision group rather than the end driver. SM006, SM007
CM019 Public comparison sources treat performance density, functional safety level, power, and OEM integration as core buying criteria in automotive HPC chips. SM006, SM007
CM020 NVIDIA describes DRIVE as an end-to-end autonomous-vehicle platform spanning training, simulation, validation, and in-vehicle computing from L2++ to L4 robotaxi. SM008
CM021 Mobileye describes EyeQ as scalable, automotive-grade, low-power silicon supporting programs from advanced driver assistance to full autonomy. SM009
CM022 Horizon Robotics says Journey is China's first and largest mass-produced automotive computing solution and that it leads urban smart-driving deployments. SM010
CM023 Rising ADAS regulation and safety mandates are cited as major growth drivers for automotive AI SoC demand. SM004, SM005
CM024 Power efficiency and thermal management are critical automotive-chip constraints because high-compute autonomy stacks must operate inside EV energy and cooling limits. SM003, SM007
CM025 Future Markets Inc says Level 4 systems drawing 400-600 watts can reduce EV range by roughly 7-10 percent. SM007
CM026 Automotive AI chips face high capital intensity because they require advanced-node tape-outs, auto-grade validation, and long operational-lifetime requirements. SM006, SM007
CM027 NIO's Onvo L90 rollout described NX9031 as gradually replacing previously used Nvidia products, illustrating the internal adoption pathway away from incumbent silicon. SM014, SM018
CM028 Procurement Magazine framed Shenji as part of NIO's strategy to reduce dependence on external suppliers such as Nvidia. SM018
CM029 The OFweek spinoff report links Shenji to China's broader automotive-semiconductor self-sufficiency push. SM020
CM030 Yicai and semi-official technical reporting frame NX9031 as roughly four-Orin-X-class compute, placing Shenji in the high-compute band of the market rather than in entry-level ADAS. SM019, SM021
CM031 NIO reported 326,028 vehicle deliveries in 2025, while Shenji/NIO public statements disclosed 200,000+ and then 300,000+ chip-shipment milestones in 2026, confirming a meaningful captive installed base. SM012, SM014, SM021
CM032 AI Weekly argued that Shenji's broad AI-platform narrative remains less proven outside NIO's own automotive ecosystem than inside it. SM022
CM033 NIO's full-stack materials market the proprietary smart-driving chip together with sensing, domain integration, and software coordination rather than as a standalone component. SM011, SM023
CM034 The market evidence supports a large addressable category for automotive AI compute, but it does not support valuing Shenji on the broadest all-vehicle SoC category by default. SM001, SM002, SM005
CM035 Shenji's potential third-party value proposition in China combines compute performance, domestic supply-chain sovereignty, and lower reliance on foreign incumbents. SM018, SM019, SM020
CM036 Shenji's market story now includes robotics and agent inference, but public evidence of revenue-bearing demand in those adjacencies is much thinner than vehicle-deployment evidence. SM016, SM021, SM022
CM037 Public evidence is sufficient to show strong market tailwind but insufficient to build a precise third-party SOM model for Shenji. SM013, SM017, SM022
CP001 Shenji's direct competitive set in public automotive AI compute comparisons includes Nvidia DRIVE, Mobileye EyeQ, Horizon Journey, Qualcomm Ride, and Black Sesame Huashan. SP009, SP010, SP014, SP016, SP018, SP020
CP002 A practical substitute for buying Shenji is continuing to build on Nvidia-centric platforms or other incumbent automotive AI chip stacks. SP008, SP009
CP003 NVIDIA describes DRIVE as an end-to-end platform spanning training, simulation, validation, and in-vehicle compute from L2++ to L4 robotaxi. SP009
CP004 Mobileye positions EyeQ as a scalable, automotive-grade, low-power compute family spanning advanced driver assistance through full autonomy. SP010
CP005 Horizon Robotics describes Journey as China's first and largest mass-produced automotive computing solution and says it leads urban smart driving. SP014
CP006 Horizon says the Journey 6 family spans from 10+ TOPS to 560 TOPS and is already tied to mass-production partnerships with more than 10 OEMs and brands. SP015
CP007 Qualcomm positions Snapdragon Ride as an automotive platform inside its broader Snapdragon Digital Chassis strategy rather than as a stand-alone chip sale. SP016
CP008 Black Sesame's Huashan A2000 was publicly described as a 7nm intelligent-driving chip family cleared for global sale after U.S. reviews. SP018
CP009 Mobileye said its EyeQ6H-based Surround ADAS programs now imply future delivery of more than 19 million systems after a major U.S. automaker win. SP011
CP010 A March 2026 Mobileye program puts its Driver Monitoring System on EyeQ6L for a leading U.S. automaker with start of production targeted for 2027 across millions of vehicles. SP012
CP011 Shenji's strongest public competitive proof point is real deployment scale, with 300,000+ NX9031X shipments across NIO and Onvo disclosed by WAIC 2026. SP001
CP012 Yicai reported that NIO describes NX9031 as about four times Orin-X in actual compute and about RMB 10,000 cheaper per vehicle. SP005
CP013 Shenji has expanded its public product family beyond NX9031X to include NX9031U, NX9031C, and NX6031. SP001, SP002
CP014 Nvidia's competitive moat is system-level ecosystem breadth, not only a single chip specification. SP009, SP019
CP015 Mobileye's moat comes from automotive-focused low-power silicon plus very large disclosed OEM program wins. SP010, SP011, SP012
CP016 Horizon's moat is China-local mass-production deployment plus software-hardware co-optimization across a full Journey family. SP014, SP015
CP017 Qualcomm's moat is cross-domain automotive integration rather than uniquely disclosed public ADAS program scale in the reviewed sources. SP016, SP017
CP018 Black Sesame competes as a domestic open-ecosystem challenger with aggressive performance marketing and BYD-linked proof. SP018
CP019 Shenji's main moat today is captive deployment and parent-controlled distribution through NIO and Onvo. SP004, SP006, SP007
CP020 Shenji's clearest weakness versus incumbents is limited public evidence on external OEM wins, software-tooling depth, and standardized commercial packaging. SP003, SP023
CP021 Public pricing transparency is low across this competitor set, so market comparison relies more on platform proof and ecosystem depth than unit price. SP016, SP017, SP020
CP022 Lantronix sells approved Snapdragon Ride-related development platforms for testing and debugging, illustrating that buyer engagement often begins through complex engineering workflows rather than simple SKU procurement. SP017
CP023 Switching away from incumbents is hard because automotive AI chips are bound up with tooling, validation, safety, and platform integration rather than raw compute alone. SP009, SP010, SP020, SP021
CP024 Onvo's switch toward NX9031 and gradual replacement of Nvidia products shows Shenji can win where the parent controls both the vehicle program and software stack. SP004, SP008
CP025 Shenji's reported external OEM outreach to Leapmotor and Geely broadens the competitive story, but it is still prospecting rather than proof of large-scale share capture. SP003
CP026 The competitive landscape fragments into global full-stack incumbents, China-local mass-production specialists, platform integrators, open-ecosystem challengers, and captive vertically integrated programs. SP009, SP014, SP016, SP018, SP020
CP027 Shenji is strongest in NIO-centric premium passenger-vehicle deployments and is much less proven across unrelated external brands. SP001, SP004, SP007
CP028 Mobileye and Horizon both disclose broader named production partnerships than Shenji in the reviewed public source set. SP011, SP012, SP015
CP029 NVIDIA says leading automakers and robotaxi fleets are already building on DRIVE Hyperion. SP009
CP030 CarNewsChina reported that Black Sesame already has BYD-linked relevance around the Huashan A2000 program. SP018
CP031 Feature competition in this market centers on compute, functional safety, power, integration breadth, and deployment proof. SP020, SP021
CP032 Shenji and several domestic peers all argue for localized, all-scenario AI compute, which raises the risk that domestic OEMs will multi-source rather than grant any one vendor a monopoly position. SP014, SP018, SP023
CP033 AI Weekly argued that evidence for Shenji's broader three-domain platform pitch remains uneven outside NIO's own ecosystem. SP023
CP034 Shenji's move into robotics and agent inference also broadens its competitor set toward edge-AI platforms from Nvidia and Qualcomm. SP002, SP016, SP025
CP035 Without clearer external customer wins, Shenji risks being valued more like a captive strategic asset than like a durable multi-customer platform. SP003, SP023
CP036 NIO's in-house-chip strategy is itself a form of internal build, meaning Shenji competes not only with third-party vendors but with each OEM's willingness to develop or control more of its own stack. SP006, SP008
CP037 The reviewed public record does not provide clean apples-to-apples production pricing across Shenji and peers, so diligence should benchmark program structure and design-win quality instead. SP016, SP017, SP020
CI001 Public evidence is still insufficient to build a standalone Shenji underwriting model because recognized revenue, gross margin, and captive-versus-external mix remain undisclosed. SI006, SI007, SI015
CI002 Shenji's observable economic pathways are captive vehicle supply inside NIO/Onvo plus external licensing or supply to third parties. SI009, SI013, SI018
CI003 External licensing of NX9031 technology became publicly visible in late 2025, indicating Shenji was moving beyond pure internal cost-center status. SI013, SI018
CI004 CarNewsChina reported that a single IP authorization could be worth several million U.S. dollars and a system-level SoC authorization could reach hundreds of millions. SI013
CI005 NIO and Shenji officially announced RMB 2.257 billion of cash investment into newly issued Shenji shares on February 26, 2026. SI009, SI010
CI006 After the funding transaction, a NIO subsidiary would retain 62.7 percent of Shenji and NIO would continue consolidating its financial results. SI009, SI015
CI007 By WAIC 2026, Shenji said cumulative financing since establishment was close to RMB 3 billion. SI019
CI008 Yicai reported NX9031 R&D cost at roughly RMB 2.3 billion to RMB 3.0 billion. SI011
CI009 Public reporting also quotes William Li saying NX9031 R&D cost billions of yuan, comparable to building about 1,500 battery-swap stations. SI010, SI022
CI010 NIO reported first-quarter 2026 total revenue of RMB 25.53 billion and vehicle sales of RMB 22.78 billion. SI005
CI011 NIO reported first-quarter 2026 vehicle margin of 18.8 percent. SI005
CI012 NIO reported full-year 2025 total revenue of RMB 87.49 billion and 326,028 vehicle deliveries. SI021
CI013 NIO's March 2026 guidance targeted full-year non-GAAP operating-profit break-even in 2026. SI016, SI021
CI014 NIO delivered 37,705 vehicles in May 2026, 107,658 vehicles in the second quarter of 2026, and 35,934 vehicles in July 2026. SI002, SI003, SI004
CI015 The July 2026 delivery mix was 20,008 NIO vehicles, 10,155 ONVO vehicles, and 5,771 FIREFLY vehicles. SI002
CI016 Shenji shipment disclosures moved from 150,000+ cumulative NX9031 units in February 2026 to 300,000+ NX9031X shipments across NIO and Onvo by July 2026. SI010, SI019
CI017 Yicai reported that NX9031 offers about RMB 10,000 of cost advantage per vehicle versus Nvidia Orin-X. SI011
CI018 The outside round reduces NIO's need to finance Shenji entirely with parent-company cash. SI009, SI015
CI019 AInvest described the transaction as a low-cost, non-dilutive move for the parent company because NIO keeps control while the cash still supports the broader business. SI015
CI020 Morgan Stanley-associated commentary framed Shenji as a potential re-rating catalyst and increasingly visible call option for NIO. SI012, SI017
CI021 Public sources do not separate how chip shipments translate into recognized Shenji revenue versus internal transfer pricing or cost savings inside NIO. SI006, SI013, SI018
CI022 Shenji's standalone gross margin and unit margin are not publicly disclosed. SI006, SI007, SI015
CI023 Future chip roadmaps remain capital intensive because automotive-grade silicon requires repeated tape-outs, validation, and software support. SI023, SI024
CI024 Public evidence is insufficient to determine whether the first outside round alone fully funds Shenji through its next major chip cycle. SI007, SI019, SI023
CI025 No reviewed public source disclosed standalone Shenji debt, project finance, or foundry prepayment obligations. SI006, SI007
CI026 As of the run date, NIO had not yet reported second-quarter 2026 results; it had only scheduled the release for September 1, 2026. SI001
CI027 NIO's 2025 Form 20-F confirms audited consolidated parent-company statements are available, but it does not amount to standalone audited Shenji financial statements. SI006, SI007
CI028 NIO said the U.S. Department of Defense listing was not a sanctions list and would not impact the business, but the event still illustrates a policy-risk overhang for capital-market narratives. SI008
CI029 The ONVO L80, ES9, and other refreshed vehicle launches in 2026 expand the internal product base that can consume Shenji chips or related software stacks. SI002, SI004
CI030 NIO said in June 2026 that it became the first to achieve synchronized releases of intelligent-driving systems across general-purpose and in-house chip platforms. SI003
CI031 From a public-disclosure perspective, Shenji still looks financially closer to a strategic captive semiconductor asset than to a diversified merchant chip vendor. SI001, SI013, SI018
CI032 NIO said the Shenji funding would support development and promotion of high-end and competitive chips. SI009
CI033 Commercial economics outside NIO remain unproven because public sources mention licensing and OEM outreach but not recurring external revenue, backlog, or customer count. SI013, SI019, SI025
CI034 Without a standalone P&L, Shenji's current valuation is best interpreted as strategic optionality plus captive operating leverage rather than as a pure revenue-multiple outcome. SI015, SI017, SI018
CI035 Visible order concentration remains extreme because the overwhelming majority of publicly documented deployment sits inside NIO-controlled brands. SI002, SI003, SI019
CI036 Parent-company profitability improvement and vehicle-margin gains may benefit from chip insourcing, but public reporting does not isolate Shenji's exact contribution. SI005, SI014, SI016
CE001 By WAIC 2026, Shenji publicly presented a product family including NX9031X, NX9031U, NX9031C, and the NX6031 sensing chip. SE006, SE017
CE002 NX9031X is Shenji's high-end assisted-driving chip inside the visible product family. SE006, SE025
CE003 NX9031U is positioned for embodied intelligence, intelligent computing, and advanced manufacturing rather than only for passenger-car ADAS. SE007, SE017
CE004 NX9031X had been deployed across all NIO and Onvo models with cumulative shipments above 300,000 by July 2026, according to Shenji's WAIC messaging. SE006, SE017
CE005 NX9031C is described as the product branch for agent inference, extending Shenji toward AI-agent workloads. SE006, SE017, SE025
CE006 NX6031 was shown publicly as an intelligent sensing chip, indicating the lineup extends beyond one high-end driving SoC. SE006, SE017
CE007 The public product surface therefore looks more like a small family of domain-focused AI chips than a single captive part number. SE006, SE007, SE017
CE008 Public technical descriptions of NX9031 repeatedly cite a 5nm automotive-grade process, more than 50 billion transistors, and a 32-core big.LITTLE CPU architecture. SE013, SE014, SE016
CE009 The same public descriptions cite a high-dynamic-range ISP, LPDDR5X memory support up to 8533 Mbps, and about 546 GB/s memory bandwidth. SE013, SE016
CE010 Public coverage says NX9031 can process LiDAR alongside camera-centric workloads, implying a sensor-fusion-oriented design rather than a pure vision-only path. SE013, SE016
CE011 NIO frames smart-driving chips, operating systems, and smart chassis as three core components of its smart-EV system architecture. SE001, SE002
CE012 NIO's official innovation materials describe a full-stack architecture with cross-domain task management and fusion of a top cockpit chip with a proprietary smart-driving chip. SE002, SE005
CE013 Evertiq reported that NX9031 complies with ASIL-D risk and safety requirements. SE013
CE014 NIO said its full-stack R&D capabilities span 12 tech domains and 8 testing centers, supporting the claim that Shenji sits inside a large systems-engineering organization. SE002, SE005
CE015 NIO Day 2024 gave the NX9031 public flagship visibility before broader 2025-2026 deployment. SE004, SE005
CE016 NX9031 is publicly described as delivering about 1000 TOPS, comparable to the aggregate of four Nvidia Orin-X chips. SE014, SE016
CE017 The ET9 was Shenji's flagship initial production vehicle showcase. SE005, SE012
CE018 By June 2026, NIO officials said the Shenji NX9031 had been deployed in ET9, ES9, ES8, and the 2026 ONVO L90, and was rolling out across the ONVO lineup. SE016, SE009
CE019 Public rollout reporting shows selected ONVO L80 trims using Shenji while some base or pure-vision versions continue to use Nvidia Orin X, meaning Shenji deployment is broad but not yet universal across every trim strategy. SE011, SE009
CE020 Secondary reporting tied the ES9 launch and 2026 vehicle refresh cycle to further Shenji deployment, broadening the installed base beyond the original flagship sedan. SE002, SE016
CE021 WAIC 2026 showed Shenji trying to convert an automotive chip program into a broader AI silicon platform through NX9031U, Ruidong, and agent-platform launches. SE006, SE007, SE017
CE022 William Li described Shenji as the only company in China shipping chips across autonomous driving, embodied intelligence, and agent inference domains. SE025, SE017
CE023 NIO said in June 2026 that its latest World Model release achieved synchronized development and release across general-purpose chip platforms and in-house chip platforms. SE015, SE016
CE024 The public roadmap widened further when Shenji previewed NX9031U with up to 800 TOPS and rich interfaces for non-vehicle AI workloads. SE007, SE017
CE025 CnEVPost and Yahoo Finance both reported that Shenji and Axera were pitching a co-developed M97 chip with more than 700 TOPS and a 2026 release target to outside automakers. SE008, SE023
CE026 The M97 roadmap is strategically important because it appears to be designed for a wider customer base than the internally focused NX9031 program. SE008, SE023
CE027 A core product differentiator is that Shenji already has mass-production field proof in real vehicles rather than only lab benchmarks or reference boards. SE006, SE016, SE018
CE028 Another differentiator is economic motivation: NIO-linked reporting said the in-house chip path can reduce per-vehicle component cost by roughly RMB 10,000 relative to the previous Nvidia-heavy architecture. SE016, SE018
CE029 Public sources reviewed do not identify Shenji's foundry, packaging, or long-term manufacturing counterparties. SE013, SE018
CE030 Unlike larger incumbent ecosystems, Shenji does not yet present a public merchant-style catalog with external price sheets, support terms, or standardized public datasheets on the reviewed surface. SE005, SE019, SE024
CE031 Compared with Nvidia and Horizon public materials, Shenji exposes less open ecosystem documentation for third-party integrators. SE019, SE020, SE021, SE024
CE032 Trust evidence is stronger on deployment scale and synchronized software release than on public disclosure of formal validation artifacts. SE004, SE016, SE023
CE033 The roadmap is therefore credible for continued NIO-centered use, but still only partially proven as an external platform roadmap. SE021, SE023, SE025
CE034 External buyers would still need product-specific safety documents, cybersecurity commitments, and support policies before treating Shenji like a standard merchant automotive-silicon vendor. SE013, SE019, SE024
CE035 Shenji's portability risk is higher than its in-house deployment risk because NIO controls the surrounding software, validation, and vehicle architecture for current deployments. SE011, SE012, SE023
CE036 No reviewed public source disclosed field-failure rates, chip yield, return metrics, or public long-term support SLAs for Shenji as a standalone supplier. SE005, SE019, SE024
CE037 The exact diligence bridge for an external customer would need a merchant datasheet, software toolchain documentation, validation package, and manufacturing/support commitments that are absent from the current public surface. SE005, SE019, SE024
CU001 The most clearly proven Shenji customer relationship is still the captive NIO ecosystem, where NIO product teams effectively act as buyer, integrator, and payer. SU001, SU021
CU002 Onvo expands Shenji's installed base, but it should still be treated as a NIO-controlled channel rather than true third-party diversification. SU002, SU005, SU006
CU003 Public evidence reviewed does not show a broad independent customer roster outside NIO-controlled brands. SU012, SU020, SU021
CU004 Prospective external OEMs are best treated as pipeline customers rather than active production customers on the current public record. SU013, SU014, SU015
CU005 WAIC 2026 materials broadened the target customer narrative to embodied-intelligence and agent users, but no named production customers for those categories were disclosed in reviewed sources. SU011, SU019, SU024
CU006 Shenji already has two monetization-quality customer types beyond internal vehicle deployment: external technical licensing counterparties and prospective external OEM supply customers. SU012, SU013, SU015
CU007 The customer verdict therefore depends on how much weight one gives internal deployment scale versus the still-early state of outside customer conversion. SU001, SU011, SU020
CU008 NIO said cumulative NX9031 shipments exceeded 150,000 units by the February 2026 financing announcement. SU001, SU018
CU009 By April 19, 2026, CnEVPost reported cumulative Shenji NX9031 shipments exceeding 200,000 units as the updated Onvo L90 headed to stores. SU005, SU006
CU010 By WAIC 2026, Shenji said NX9031X cumulative shipments across NIO and Onvo models exceeded 300,000 units. SU011, SU019
CU011 The ET9 provides flagship production proof that Shenji reached a real paying end-market inside NIO rather than staying a prototype. SU023, SU012
CU012 NIO delivered 107,658 vehicles in the second quarter of 2026 and 35,934 vehicles in July 2026, showing the captive ecosystem remained a large demand engine. SU009, SU010
CU013 Onvo L90 is the strongest public second-brand customer-proof case because official pages, launch materials, and shipping coverage all tie the vehicle to Shenji smart-driving hardware. SU002, SU003, SU005
CU014 Onvo L80 offers only partial proof because Shenji is used in selected LiDAR trims while other versions retain Nvidia-based smart-driving hardware. SU007, SU006
CU015 Late-2025 coverage said NX9031 had entered external technical licensing, marking Shenji's first publicly visible non-captive commercialization. SU012, SU013
CU016 By March 2026, Shenji and Axera were reported to be pitching the M97 chip to external automakers including Leapmotor and Geely. SU013, SU014, SU015
CU017 The public evidence for Leapmotor and Geely is pipeline-level outreach and collaboration discussion, not confirmed production purchase orders. SU013, SU014, SU015
CU018 Adjacent AI customer expansion remains a top-of-funnel story because products are showcased and positioned for new users, but named production customers are absent from the reviewed public record. SU011, SU019, SU024
CU019 No reviewed source publicly confirmed a third-party SOP design win or large-scale external OEM deployment as of the run date. SU015, SU020, SU024
CU020 Visible customer concentration is extreme because the overwhelming majority of named deployments are inside NIO-controlled brands. SU002, SU011, SU021
CU021 External OEM procurement may be harder because Shenji is controlled by a competing automaker and public governance / support transparency is limited. SU013, SU017, SU021
CU022 The Onvo L90 store-shipping and test-drive rollout shows Shenji embedded in a retail go-to-market process, not only in engineering demos. SU005, SU004
CU023 Official product-page language on the Onvo L90 presents Shenji NX9031 as standard smart-driving hardware rather than as an experimental premium option. SU002, SU003
CU024 The existence of external licensing is meaningful as customer proof, but the unnamed counterparty keeps evidence quality below the standard of a named production customer. SU012, SU013
CU025 The strongest external expansion evidence is therefore not a closed customer win but a combination of licensing, M97 outreach, and the broader platform narrative presented at WAIC. SU011, SU015, SU019
CU026 No reviewed public source disclosed Shenji net revenue retention, gross revenue retention, churn, or renewal rates. SU017, SU020, SU021
CU027 No reviewed public source disclosed external licensing contract length, renewal mechanics, or customer-level revenue concentration. SU012, SU013, SU017
CU028 Continuing rollout across NIO and Onvo suggests internal architectural commitment, but it is not equivalent to transparent third-party retention data. SU005, SU009, SU010
CU029 Public customer-satisfaction evidence is also absent at the chip-vendor level: there are no reviewed disclosures on defect-return rates, SLA performance, or OEM satisfaction. SU017, SU020, SU021
CU030 Shenji's expansion loop currently runs from NIO flagship proof to Onvo scale-up to external OEM outreach; each step is visible, but only the first two are production-proven. SU005, SU013, SU015
CU031 Because external-customer conversion is still early, Shenji's customer story today is better described as strong adoption proof plus unresolved diversification risk. SU011, SU020, SU021
CU032 The first users and launch materials for the Onvo L90 support the view that Shenji is already embedded in a mainstream retail product rather than a small experimental fleet. SU003, SU004, SU005
CU033 The most important diligence blocker is a customer concentration bridge showing revenue or unit share by NIO brand, Onvo, licensing counterparties, and external pilots. SU017, SU021, SU020
CU034 A second diligence blocker is evidence that an external OEM can adopt Shenji without unacceptable governance, support, or competitive-conflict concerns. SU013, SU017, SU021
CU035 From a customer-quality perspective, Shenji has crossed the adoption threshold but not yet the diversification threshold. SU001, SU011, SU021
CR001 The June 2026 Federal Register notice made NIO's inclusion on the U.S. Department of Defense 1260H Chinese military companies list a live public-regulatory fact. SR001, SR003, SR006
CR002 NIO said the CMC list is not a sanctions list and would not impact the business or restrict transacting in its securities. SR010
CR003 Even without sanctions-like effect, the CMC designation can still create procurement, partner, and financing stigma for businesses tightly associated with NIO. SR003, SR006, SR010
CR004 BIS in 2026 reaffirmed that advanced-computing export licensing rules for China-linked entities remain active and case-specific rather than irrelevant. SR002, SR004, SR005
CR005 15 CFR 744.23 maintains end-use controls for advanced-node integrated circuits and related items, keeping advanced compute policy relevant to Shenji's sector. SR008, SR004
CR006 Because Shenji's exact foundry and manufacturing chain are not publicly disclosed, investors cannot cleanly map its exposure to current or future export-control tightening. SR004, SR005, SR011
CR007 The main policy risk is therefore indirect but material: Shenji may not be blocked today, yet it operates in a category where regulatory conditions can change quickly. SR002, SR004, SR005
CR008 The 1260H development and BIS guidance together show that U.S. policy risk is no longer merely hypothetical for NIO-linked advanced-computing assets. SR001, SR002, SR006
CR009 Legal and policy posture could weigh more heavily on overseas or U.S.-linked commercial relationships than on domestic Chinese demand. SR003, SR006, SR029
CR010 NIO said it would proactively engage with DoD and take legal action if necessary, which is a mitigation but also confirms the issue rose to formal legal-management attention. SR010, SR003
CR011 Onvo-related reporting said the updated L90 cost increased by about RMB 10,000 partly because of high-performance memory component price increases, signaling BOM sensitivity. SR012
CR012 Public sources reviewed do not identify Shenji's foundry, packaging, or long-term supply agreements. SR011, SR013
CR013 Vertical integration reduces dependence on Nvidia, but it also shifts more supply-chain, manufacturing, and validation burden onto Shenji and NIO. SR013, SR029
CR014 A public mention of ASIL-D compliance exists in Evertiq coverage, but the reviewed public surface does not include a primary certificate or safety-case package. SR011, SR014
CR015 No reviewed public source disclosed field failure rates, return rates, or chip-yield data for Shenji. SR014, SR027
CR016 Mass-production deployment across NIO and Onvo materially lowers the risk that Shenji is only a lab-stage or non-viable product. SR019, SR023, SR027
CR017 Demand concentration remains extreme because the overwhelming majority of named production proof sits inside NIO and Onvo. SR019, SR024, SR025
CR018 Public evidence does not show a confirmed third-party SOP design win, leaving merchant-commercialization risk materially unresolved. SR021, SR022, SR018
CR019 Public sources still do not disclose Shenji standalone revenue, gross margin, burn, or cash runway, which keeps financial-model risk high. SR015, SR016, SR030
CR020 NIO's stated 2026 non-GAAP operating break-even target creates parent-company pressure that could shape how much capital and patience Shenji receives. SR016, SR024
CR021 The February 2026 outside financing reduced immediate funding pressure but did not eliminate future capital dependency risk for additional chip cycles. SR017, SR015, SR030
CR022 Shenji is simultaneously trying to sustain automotive deployment and widen scope into embodied AI and agent inference, which increases execution complexity. SR019, SR020, SR023
CR023 M97 commercialization adds partnership and coordination risk because Shenji is depending on Axera and a joint venture for external supply expansion. SR021, SR022
CR024 The public executive and governance surface for Shenji remains thinner than its technical narrative, which raises key-person and commercial-accountability questions. SR017, SR030
CR025 Late-2025 licensing coverage described a team of more than 600 professionals across design, verification, and testing, implying both strength and concentration of specialized talent. SR018
CR026 NIO's response to the DoD list indicates the company is prepared to defend itself legally, which slightly mitigates but does not erase the policy-risk overhang. SR010, SR003
CR027 No reviewed public source disclosed a Shenji-specific export-control mitigation plan covering supplier substitution, tooling continuity, or contingency sourcing. SR004, SR005, SR013
CR028 NIO's ESG and governance materials provide some evidence of parent-company compliance systems, but they do not substitute for Shenji-specific merchant-supplier controls. SR014, SR017
CR029 External OEMs may hesitate to buy from a supplier controlled by a competing automaker, especially when public support and governance terms are under-disclosed. SR021, SR022, SR030
CR030 Shenji's performance and cost claims remain only partially independently validated, creating proof-quality risk even though deployment is real. SR011, SR023, SR026
CR031 The chip platform still depends heavily on NIO software, validation, and vehicle architecture, which raises portability risk for external deployments. SR024, SR027, SR029
CR032 If NIO vehicle demand slows or brand strategy shifts, Shenji's internal absorption and learning loop could weaken quickly because concentration is so high. SR024, SR025, SR030
CR033 Tighter export controls or supplier restrictions would likely transmit first into roadmap timing and external OEM confidence, then into margin and valuation. SR004, SR005, SR011
CR034 Because debt obligations, foundry prepayments, and long-term purchase commitments are not publicly disclosed, hidden working-capital or liquidity risks cannot be ruled out. SR015, SR017, SR030
CR035 NIO's multi-brand launch agenda means Shenji shares operating attention with many other strategic priorities at the parent level. SR016, SR024, SR028
CR036 At present, Shenji looks safer as a strategic captive asset than as a fully underwritten merchant semiconductor business. SR018, SR021, SR030
CR037 A practical mitigation is that real shipments and repeated public launches give Shenji operating proof that many early-stage chip ventures never achieve. SR019, SR023, SR027
CR038 A first clear kill trigger would be failure to announce any named third-party production customer by the next major chip roadmap cycle. SR018, SR021, SR022
CR039 A second kill trigger would be any evidence that outside OEM diligence fails because Shenji cannot provide sufficient safety, support, or governance documentation. SR014, SR027, SR029
CR040 A third kill trigger would be direct regulatory or supply-chain tightening that materially disrupts Shenji's advanced-node roadmap or customer conversions. SR002, SR004, SR005
CR041 Real deployment, outside financing, and parent-company engineering depth are meaningful mitigants, but none of them alone resolves concentration or policy risk. SR017, SR019, SR014
CR042 Overall, Shenji should be treated as a high-risk, dependency-heavy platform transition rather than as a standard mid-risk hardware company. SR008, SR017, SR021
CV001 The official February 2026 Shenji financing brought in RMB 2.257 billion of cash while leaving NIO with 62.7 percent control, outside investors with 27.3 percent, and the incentive plan with 10 percent. SV001, SV002
CV002 CnEVPost and AInvest explicitly derived an implied Shenji post-money valuation of about RMB 8.27 billion from the financing structure. SV002, SV004
CV003 Yicai and several secondary summaries rounded Shenji's value to nearly RMB 10 billion or roughly USD 1.5 billion. SV003, SV008
CV004 Morgan Stanley-associated commentary framed Shenji as a re-rating catalyst for NIO and as a strategic AI-silicon option beyond cars. SV005, SV006, SV026
CV005 Outside financing validates that Shenji is viewed by investors as a separable strategic asset rather than only as an internal cost center. SV001, SV005, SV007
CV006 Public evidence still does not disclose Shenji standalone revenue, gross margin, external revenue mix, burn, or cash runway. SV004, SV015, SV024
CV007 Because core merchant-economics inputs are missing, Shenji cannot yet be valued with high confidence like a disclosed public semiconductor company. SV004, SV016, SV017, SV018
CV008 Real operating proof exists: shipments progressed from 150,000+ by February 2026 to 300,000+ across NIO and Onvo by July 2026. SV001, SV012
CV009 NIO's 2026 delivery and financial disclosures show the parent ecosystem remains large enough to keep absorbing Shenji chips and supporting proof generation. SV009, SV010, SV011
CV010 External commercialization is visible but early, with licensing evidence and external OEM outreach but no confirmed third-party production win on the reviewed public record. SV013, SV014, SV027
CV011 The official implied mark around RMB 8.27 billion is a more reliable anchor than the looser “nearly RMB 10 billion” shorthand. SV001, SV002, SV003
CV012 At a rounded near-RMB-10-billion mark, investors would be paying more for optionality than the public evidence currently supports. SV003, SV004, SV015
CV013 Shenji is best valued today as a strategic private platform asset with incomplete economics disclosure, not as a transparently underwritten merchant silicon company. SV004, SV006, SV024
CV014 The bear case assumes Shenji remains mostly captive, misses named external production wins, and faces continuing policy or supply overhang. SV010, SV014, SV030
CV015 Under the bear case, a reasonable estimated range is roughly RMB 5.5 billion to RMB 7.0 billion, below the current official mark. SV002, SV014, SV030
CV016 The base case assumes internal deployment continues, external monetization progresses slowly, and major policy conditions do not worsen. SV009, SV010, SV013
CV017 Under the base case, a reasonable estimated range is roughly RMB 8.0 billion to RMB 9.5 billion, which brackets the current official mark as approximately fair. SV002, SV004, SV016
CV018 The bull case requires named external OEM wins, merchant-grade support and safety readiness, and credible monetization outside the NIO ecosystem. SV014, SV026, SV027
CV019 At the current evidence level, the appropriate recommendation is watch rather than invest or pass. SV006, SV015, SV024
CV020 Confidence in that recommendation is medium because product proof is real, but economics and diversification remain opaque. SV006, SV012, SV024
CV021 Risk rating should remain high and valuation stance should be considered fair-to-full on public evidence. SV015, SV024, SV030
CV022 A pass recommendation would underweight Shenji's real deployment proof, strategic parent role, and demonstrated ability to raise outside capital. SV001, SV005, SV012
CV023 An invest recommendation would overstate the precision of the current underwriting case because the key revenue and margin bridges are still absent. SV004, SV015, SV024
CV024 The bull-case valuation range requires evidence improvement, not merely repetition of the broader AI-platform narrative. SV006, SV013, SV027
CV025 The most credible immediate upside catalyst would be a named external production customer rather than another broad platform presentation. SV014, SV026, SV027
CV026 AInvest noted muted market reaction to the funding news, which suggests the financing event alone did not create an obvious public-market repricing. SV004
CV027 That muted reaction can be read two ways: the event de-risked Shenji somewhat, but investors still wanted harder evidence on downstream economics. SV004, SV015
CV028 The absence of any public down-round language or distressed financing signal is a favorable valuation datapoint, even if it is not sufficient to prove cheapness. SV001, SV005, SV007
CV029 Customer concentration and incomplete external conversion are the two clearest reasons the valuation should not command a full merchant-platform premium yet. SV010, SV014, SV024
CV030 Mobileye, Nvidia, and Qualcomm are useful comparables mainly because they demonstrate what public disclosure and ecosystem transparency look like in this category. SV016, SV017, SV018, SV019
CV031 Those peers are imperfect valuation comps because they are public, diversified, or at a much later disclosure stage than Shenji. SV016, SV017, SV018
CV032 NIO parent-company valuation is a relevant but noisy lens because Shenji is still controlled and consolidated inside NIO. SV001, SV024, SV026
CV033 Until Shenji provides cleaner standalone reporting, a large part of its mark remains a sum-of-the-parts judgment on NIO strategy rather than a pure chip-business multiple. SV001, SV006, SV024
CV034 Shenji is not yet exit-ready for a high-conviction IPO-style valuation because public reporting still falls short of what listed peers provide. SV016, SV017, SV018, SV024
CV035 Secondary liquidity or rapid markup assumptions are also hard to support publicly without evidence of merchant revenue quality or named external customers. SV014, SV015, SV024
CV036 A first thesis-break trigger would be continued failure to announce a named third-party production customer by the next major roadmap cycle. SV014, SV027, SV030
CV037 A second thesis-break trigger would be any policy or supply disruption that directly affects advanced-node roadmap timing or external procurement confidence. SV019, SV022, SV030
CV038 The most important diligence ask is a standalone revenue and gross-margin bridge separating captive from external economics. SV004, SV024, SV025
CV039 A second critical diligence ask is a customer concentration schedule covering NIO, Onvo, licensing counterparties, and external OEM pipeline stages. SV010, SV014, SV024
CV040 A third critical diligence ask is the merchant support package: safety validation, warranty, support SLAs, and governance protections for outside buyers. SV019, SV024, SV030
来源
编号出版方标题引文
SO001 NIO Inc. NIO Inc.’s Subsidiary Enters into Definitive Agreements for RMB2.257 Billion Investment
SO002 CnEVPost Nio confirms its chip unit secures $330 million in 1st-round funding
SO003 Yicai Global Nio’s Smart-Driving Chip Unit Nears USD1.5 Billion Valuation in First Fundraiser That will give them a collective 27.3 percent stake. Shanghai-based Nio will retain a controlling interest of 62.7 percent.
SO004 OFweek NIO Spins Off Chip Unit: China's First 5nm Automotive Semiconductor Company Emerges The newly incorporated company, registered in Hefei on June 17 with 10 million yuan in initial capital, marks a significant milestone.
SO005 NIO NIO Inc. Releases 2025 ESG Report - News
SO006 NIO NIO Day 2024 : The NIO ET9 Officially Launched, and the firefly Brand Unveiled - News
SO007 NIO NIO Showcases Its Three EV Brands at Auto Shanghai 2025, firefly to Enter 16 International Markets this Year - News
SO008 NIO NIO Presents Three Brands at Auto China 2026 - News
SO009 CnEVPost Nio chip unit showcases multiple chips at WAIC 2026 in Shanghai The high-end NX9031X chip has been deployed across all Nio and Onvo models, with cumulative shipments exceeding 300,000 units.
SO010 CnEVPost Nio chip unit previews NX9031U as AI ambitions move beyond cars
SO011 CnEVPost Nio's chip unit approaches Leapmotor and Geely for potential supply deals, report says
SO012 CnEVPost Nio Onvo to equip updated L90 SUV with in-house developed Shenji NX9031 chip
SO013 CnEVPost Nio Onvo L80 mirrors L90 strategy with in-house developed chip integration
SO014 CnEVPost Nio moves beyond cash-burning EV label on chip unit's AI push, Morgan Stanley says The bank sees the chip business as an increasingly visible call option in addition to the auto business.
SO015 CarNewsChina Nio's self-developed advanced intelligent driving chip enters external licensing market, report says This marks a pivotal moment for Nio, transforming a substantial R&D expenditure into a potential income stream.
SO016 CarNewsChina Production version Nio ET9 flagship sedan with 120 kWh battery exposed before launch
SO017 CarNewsChina Nio ET5 2025 spied: brings in-house 5nm chip with AI
SO018 Evertiq EV maker Nio has developed a 5nm processor for self-driving The firm says its system-on-chip Shenji NX9031 5nm processor packs 50 billion transistors, and is 4x more advanced than Nvidia's Drive Orin X.
SO019 Wccftech Chinese Automotive Firm NIO Unveils Industry's First 5nm "Smart Driving" AI Chip, NX9031 With 50+ Billion Transistors & 32 Big.Little Cores
SO020 MemoryMarket NIO's Chip Subsidiary "Shenji" Secures Over RMB 2.2 Billion in Series A Funding | MemoryMarket
SO021 Internet Info Agency NIO's In-House "Shenji NX9031" Chip Rolls Out Across Multiple Models The chip features over 50 billion transistors and employs a 32-core big.LITTLE CPU architecture.
SO022 Internet Info Agency NIO's Chip Subsidiary Shiji Unveils Multi-Scenario AI Chip Platform and Embodied Intelligence Development Solutions at WAIC
SO023 AI Weekly Nio's Shenji Shows Off Three-Chip AI Lineup at WAIC 2026 The evidence for the broader platform pitch is still uneven outside Nio's own automotive ecosystem.
SO024 Procurement Magazine How NIO's Chip Insourcing Strategy Cuts Supplier Dependence
SO025 AInvest Nio's Chip Raise: A Tactical Funding Win or a Valuation Trap? The deal itself is a positive, de-risking a capital-intensive R&D path. But the flat price action suggests the market sees it as a known quantity or perhaps a minor event in a broader narrative of cash burn and profitability.
SO026 NIO Inc. NIO Inc. Reports Unaudited Fourth Quarter and Full Year 2025 Financial Results
SO027 AInvest NIO Inc.’s 2026 Guidance: Gross Margin Forecasts and Model Launch Timelines Clash
SO028 China Biz Insider NIO GeniTech: From Auto Chip to AI Silicon Platform
SO029 China Biz Insider NIO's In-House Chip Arm Draws RMB 2.3 Billion in External Funding, Signaling a Potential Re-Rating Catalyst
SM001 Future Market Insights Explore the Global Automotive SoC Market — analysis of key trends, regional growth, top players, and a 10-year forecast from 2026 to 2036.
SM002 MarketResearch / DIResearch Global Autonomous Driving SOC Chip Competitive Landscape Professional Research Report 2026
SM003 DataCentre Magazine Next-Generation Automotive Computing Market 2026: Nvidia Leads with Drive Platform, Orin SoC, and Next-Gen Thor Targeting Level 4 ADAS Performance - Global Long-term Forecast to 2036
SM004 Semiconductor Insight Autonomous Driving SoC Market, Trends, Business Strategies 2026-2034
SM005 Semiconductor Insight ADAS (Advanced Driver Assist) SoC Market, Trends, Business Strategies 2026-2034
SM006 viable.works Automotive HPC Chip Comparison: Comprehensive Guide (Q1/2026)
SM007 Future Markets Inc Next-Gen Automotive Computing Market 2026-2036 | ADAS, AI
SM008 NVIDIA NVIDIA: End-to-End Platform for Robotaxis & Autonomous Vehicles
SM009 Mobileye The Evolution of EyeQ
SM010 Horizon Robotics Smart Driving Technology_Smart Driving Solutions_Smart Driving Vehicles
SM011 NIO NIO Full Stack
SM012 NIO Inc. NIO Inc. Reports Unaudited Fourth Quarter and Full Year 2025 Financial Results
SM013 CnEVPost Nio's chip unit approaches Leapmotor and Geely for potential supply deals, report says
SM014 CnEVPost Nio Onvo starts nationwide shipping of 2026 L90 SUV ahead of Apr 21 launch
SM015 CnEVPost Nio Onvo L80 mirrors L90 strategy with in-house developed chip integration
SM016 CnEVPost Nio chip unit previews NX9031U as AI ambitions move beyond cars
SM017 CarNewsChina Nio's self-developed advanced intelligent driving chip enters external licensing market, report says
SM018 Procurement Magazine How NIO's Chip Insourcing Strategy Cuts Supplier Dependence
SM019 Yicai Global Nio’s Smart-Driving Chip Unit Nears USD1.5 Billion Valuation in First Fundraiser
SM020 OFweek NIO Spins Off Chip Unit: China's First 5nm Automotive Semiconductor Company Emerges
SM021 CnEVPost Nio chip unit showcases multiple chips at WAIC 2026 in Shanghai
SM022 AI Weekly Nio's Shenji Shows Off Three-Chip AI Lineup at WAIC 2026
SM023 NIO NIO Inc. Releases 2025 ESG Report - News
SM024 NIO NIO Presents Three Brands at Auto China 2026 - News
SM025 NIO NIO Showcases Its Three EV Brands at Auto Shanghai 2025, firefly to Enter 16 International Markets this Year - News
SP001 CnEVPost Nio chip unit showcases multiple chips at WAIC 2026 in Shanghai
SP002 CnEVPost Nio chip unit previews NX9031U as AI ambitions move beyond cars
SP003 CnEVPost Nio's chip unit approaches Leapmotor and Geely for potential supply deals, report says
SP004 CnEVPost Nio Onvo starts nationwide shipping of 2026 L90 SUV ahead of Apr 21 launch
SP005 Yicai Global Nio’s Smart-Driving Chip Unit Nears USD1.5 Billion Valuation in First Fundraiser
SP006 NIO NIO Full Stack
SP007 NIO Inc. NIO Inc. Reports Unaudited Fourth Quarter and Full Year 2025 Financial Results
SP008 Procurement Magazine How NIO's Chip Insourcing Strategy Cuts Supplier Dependence
SP009 NVIDIA NVIDIA: End-to-End Platform for Robotaxis & Autonomous Vehicles
SP010 Mobileye The Evolution of EyeQ
SP011 Mobileye Mobileye Surround ADAS Adds Second Top 10 Automaker | Mobileye News
SP012 Nasdaq / BusinessWire Mobileye Secures Major DMS Production Program with Leading U.S. Automaker
SP013 Mobileye Mobileye News | Stay Updated on the Latest Innovations
SP014 Horizon Robotics Smart Driving Technology_Smart Driving Solutions_Smart Driving Vehicles
SP015 Horizon Robotics Journey 6 Series_Intelligent Computing_Intelligent Automotive-Horizon Robotics
SP016 Qualcomm Snapdragon Ride
SP017 Lantronix SA8255P and SA8775P Ride SX4.0 Automotive Development Platforms
SP018 CarNewsChina BYD ADAS chip supplier, Black Sesame Technologies' Huashan A2000 passed US government reviews, ready for global sales
SP019 DataCentre Magazine Next-Generation Automotive Computing Market 2026: Nvidia Leads with Drive Platform, Orin SoC, and Next-Gen Thor Targeting Level 4 ADAS Performance - Global Long-term Forecast to 2036
SP020 viable.works Automotive HPC Chip Comparison: Comprehensive Guide (Q1/2026)
SP021 Future Markets Inc Next-Gen Automotive Computing Market 2026-2036 | ADAS, AI
SP022 Semiconductor Insight Autonomous Driving SoC Market, Trends, Business Strategies 2026-2034
SP023 AI Weekly Nio's Shenji Shows Off Three-Chip AI Lineup at WAIC 2026
SP024 OFweek NIO Spins Off Chip Unit: China's First 5nm Automotive Semiconductor Company Emerges
SP025 NeuralCoreTech Edge AI Hardware 2026: On-Device Intelligence, Architecture & Chip Comparison
SI001 NIO Inc. NIO Inc. to Report Unaudited Second Quarter 2026 Financial Results on Tuesday, September 1, 2026
SI002 NIO Inc. NIO Inc. Provides July 2026 Delivery Update
SI003 NIO Inc. NIO Inc. Provides June and Second Quarter 2026 Delivery Update
SI004 NIO Inc. NIO Inc. Provides May 2026 Delivery Update
SI005 NIO Inc. NIO Inc. Reports Unaudited First Quarter 2026 Financial Results
SI006 NIO Inc. NIO Inc. Files Its Annual Report on Form 20-F
SI007 NIO Inc. NIO Inc. Files Its Annual Report on Form 20-F
SI008 NIO Inc. NIO Inc. Responds to the U.S. Department of Defense “Chinese Military Companies” List
SI009 NIO Inc. NIO Inc.’s Subsidiary Enters into Definitive Agreements for RMB2.257 Billion Investment
SI010 CnEVPost Nio confirms its chip unit secures $330 million in 1st-round funding
SI011 Yicai Global Nio’s Smart-Driving Chip Unit Nears USD1.5 Billion Valuation in First Fundraiser
SI012 CnEVPost Nio moves beyond cash-burning EV label on chip unit's AI push, Morgan Stanley says
SI013 CarNewsChina Nio's self-developed advanced intelligent driving chip enters external licensing market, report says
SI014 Procurement Magazine How NIO's Chip Insourcing Strategy Cuts Supplier Dependence
SI015 AInvest Nio's Chip Raise: A Tactical Funding Win or a Valuation Trap?
SI016 AInvest NIO Inc.’s 2026 Guidance: Gross Margin Forecasts and Model Launch Timelines Clash
SI017 China Biz Insider NIO's In-House Chip Arm Draws RMB 2.3 Billion in External Funding, Signaling a Potential Re-Rating Catalyst
SI018 China Biz Insider NIO GeniTech: From Auto Chip to AI Silicon Platform
SI019 CnEVPost Nio chip unit showcases multiple chips at WAIC 2026 in Shanghai
SI020 CnEVPost Nio chip unit previews NX9031U as AI ambitions move beyond cars
SI021 NIO Inc. NIO Inc. Reports Unaudited Fourth Quarter and Full Year 2025 Financial Results
SI022 OFweek NIO Spins Off Chip Unit: China's First 5nm Automotive Semiconductor Company Emerges
SI023 Future Markets Inc Next-Gen Automotive Computing Market 2026-2036 | ADAS, AI
SI024 DataCentre Magazine Next-Generation Automotive Computing Market 2026: Nvidia Leads with Drive Platform, Orin SoC, and Next-Gen Thor Targeting Level 4 ADAS Performance - Global Long-term Forecast to 2036
SI025 AI Weekly Nio's Shenji Shows Off Three-Chip AI Lineup at WAIC 2026
SE001 NIO Auto Shanghai 2025 press release
SE002 NIO Auto China 2026 press release
SE003 NIO Investor Relations Shenji funding announcement
SE004 NIO NIO Day 2024 event recap
SE005 NIO NIO ET9 product page
SE006 CnEVPost Nio chip unit showcases multiple chips at WAIC 2026 in Shanghai
SE007 CnEVPost Nio chip unit previews NX9031U as AI ambitions move beyond cars
SE008 CnEVPost Nio's chip unit approaches Leapmotor and Geely for potential supply deals
SE009 CnEVPost Onvo confirms the 2026 L90 will use Shenji chip
SE010 CnEVPost Onvo L90 to adopt Nio Shenji smart driving chip
SE011 CnEVPost Onvo L80 will offer LiDAR and pure vision variants
SE012 CarNewsChina Shenji NX9031 enters external licensing market
SE013 Evertiq EV maker Nio has developed a 5nm processor for self-driving
SE014 Wccftech NIO showcases 5nm automotive chip NX9031
SE015 NIO Investor Relations NIO provides June and second quarter 2026 delivery update
SE016 News18A NIO's in-house Shenji NX9031 chip rolls out across multiple models
SE017 News18A NIO's chip subsidiary unveils multi-scenario AI chip platform at WAIC
SE018 Procurement Magazine How NIO's chip insourcing strategy cuts supplier dependence
SE019 CnEVPost Shenji topic page
SE020 NVIDIA Autonomous vehicle solutions overview
SE021 Horizon Robotics Horizon smart driving overview
SE022 BusinessWire / ResearchAndMarkets Next-generation automotive computing market 2026-2036
SE023 Yahoo Finance NIO in discussions with Leapmotor and Geely for chip supply
SE024 Lantronix Snapdragon Ride automotive development platform store page
SE025 EV Nio CEO says chip unit is China's only three-domain AI chipmaker
SU001 NIO Investor Relations Shenji funding announcement
SU002 ONVO Onvo L90 official product page
SU003 NIO ONVO L90 officially launched
SU004 NIO ONVO L90 preorder starts
SU005 CnEVPost Onvo starts nationwide shipping of 2026 L90 ahead of Apr 21 launch
SU006 CnEVPost Onvo confirms updated L90 will use Shenji NX9031 chip
SU007 CnEVPost Onvo L80 LiDAR version uses Shenji chip
SU008 News18A Shenji chip rolls out across multiple NIO and Onvo models
SU009 NIO Investor Relations NIO provides July 2026 delivery update
SU010 NIO Investor Relations NIO provides June and second quarter 2026 delivery update
SU011 CnEVPost Nio chip unit showcases multiple chips at WAIC 2026 in Shanghai
SU012 CarNewsChina Shenji NX9031 enters external licensing market
SU013 Yahoo Finance NIO in discussions with Leapmotor and Geely for chip supply
SU014 News18A Shenji chip seeks external partnerships; M97 to supply Leapmotor and Geely
SU015 CnEVPost Nio chip unit approaches Leapmotor and Geely for potential supply deals
SU016 Theia Global Shenji secures over RMB 2.2 billion investment to expand R&D
SU017 Autoworld Journal Nio's chip unit Shenji secures $330 million in external funding
SU018 Yicai Global Shenji nears USD1.5 billion valuation in first fundraiser
SU019 China Biz Insider GeniTech from auto chip to AI silicon platform
SU020 AI Weekly Nio's Shenji shows off three-chip AI lineup at WAIC 2026
SU021 Procurement Magazine How NIO's chip insourcing strategy cuts supplier dependence
SU022 CarNewsChina Nio ES9 started sales in China
SU023 NIO NIO ET9 product page
SU024 EV Nio CEO says chip unit is China's only three-domain AI chipmaker
SU025 NIO Investor Relations NIO provides May 2026 delivery update
SR001 Federal Register Notice of Availability of Designation of Chinese Military Companies
SR002 Bureau of Industry and Security Department of Commerce Revises License Review Policy for Semiconductors Exported to China
SR003 Cleary Gottlieb DoD Adds Major Technology, Automotive, and Biotechnology Firms to 1260H Chinese Military Companies List
SR004 Holland & Knight BIS Publishes Guidance Regarding License Requirements for Advanced Computing Items
SR005 Greenberg Traurig Enforcement Pause Has Limits: BIS Clarifies Ongoing License Requirement
SR006 WilmerHale Pentagon Adds 65 New Entities to the 1260H List
SR007 Bureau of Industry and Security Federal Register Notices index
SR008 eCFR 15 CFR 744.23 advanced-node IC end-use controls
SR009 GovInfo 10 U.S.C. 113 reference PDF
SR010 NIO Investor Relations NIO responds to DoD Chinese military companies list
SR011 Evertiq EV maker Nio has developed a 5nm processor for self-driving
SR012 CnEVPost Onvo L90 to adopt Nio Shenji smart driving chip
SR013 Procurement Magazine How NIO's chip insourcing strategy cuts supplier dependence
SR014 NIO 2025 ESG Report news summary
SR015 AInvest Nio's chip raise: funding win or valuation trap?
SR016 AInvest NIO 2026 guidance: gross margin forecasts and model launch timelines clash
SR017 NIO Investor Relations Shenji funding announcement
SR018 CarNewsChina Shenji NX9031 enters external licensing market
SR019 CnEVPost Nio chip unit showcases multiple chips at WAIC 2026 in Shanghai
SR020 CnEVPost Nio chip unit previews NX9031U as AI ambitions move beyond cars
SR021 CnEVPost Nio chip unit approaches Leapmotor and Geely for potential supply deals
SR022 News18A Shenji chip seeks external partnerships; M97 to supply Leapmotor and Geely
SR023 News18A Shenji chip rolls out across multiple models
SR024 NIO Investor Relations NIO provides June and second quarter 2026 delivery update
SR025 NIO Investor Relations NIO provides July 2026 delivery update
SR026 Wccftech NIO showcases 5nm automotive chip NX9031
SR027 ONVO Onvo L90 official product page
SR028 NIO ONVO L90 officially launched
SR029 Tech in Asia Nio spins off chip unit to cut reliance on Nvidia
SR030 Autoworld Journal Nio's chip unit Shenji secures $330 million in external funding
SV001 NIO Investor Relations Shenji funding announcement
SV002 CnEVPost Nio confirms its chip unit secures first-round funding
SV003 Yicai Global Shenji nears USD1.5 billion valuation in first fundraiser
SV004 AInvest Nio's chip raise: tactical funding win or valuation trap?
SV005 China Biz Insider NIO's in-house chip arm draws RMB 2.3 billion in external funding
SV006 China Biz Insider GeniTech from auto chip to AI silicon platform
SV007 Autoworld Journal Nio's chip unit Shenji secures $330 million in external funding
SV008 Theia Global Shenji secures over RMB 2.2 billion investment to expand R&D
SV009 NIO Investor Relations NIO reports unaudited first quarter 2026 financial results
SV010 NIO Investor Relations NIO provides June and second quarter 2026 delivery update
SV011 NIO Investor Relations NIO provides July 2026 delivery update
SV012 CnEVPost Nio chip unit showcases multiple chips at WAIC 2026 in Shanghai
SV013 CarNewsChina Shenji NX9031 enters external licensing market
SV014 Yahoo Finance NIO in discussions with Leapmotor and Geely for chip supply
SV015 AInvest NIO 2026 guidance: gross margin forecasts and model launch timelines clash
SV016 SEC / Mobileye Mobileye 20-F filing page
SV017 SEC / Nvidia Nvidia 10-K filing page
SV018 SEC / Qualcomm Qualcomm 10-K filing page
SV019 NVIDIA Autonomous vehicle solutions overview
SV020 NVIDIA DRIVE AGX hardware page
SV021 Chip.computer Autonomous driving chips guide
SV022 NIO Investor Relations Q2 2026 results schedule PDF
SV023 NIO Investor Relations July 2026 delivery update PDF
SV024 NIO Investor Relations NIO files annual report on Form 20-F
SV025 NIO Investor Relations NIO annual report PDF
SV026 CnEVPost Nio moves beyond cash-burning EV label on chip unit's AI push, Morgan Stanley says
SV027 News18A Shenji chip seeks external partnerships; M97 to supply Leapmotor and Geely
SV028 NIO Investor Relations NIO provides May 2026 delivery update
SV029 NIO News in July
SV030 Federal Register / BIS Department of Commerce revises license review policy for semiconductors exported to China