Sharpa
触觉优先的具身智能机器人公司
Sharpa 融资充足,聚焦触觉机器人;一份未注明日期的行业画像称其尚未产生收入,现场可靠性、安全与单机经济性仍无定论。
覆盖范围与披露说明
公开证据集中在产品里程碑、融资及试点落地;详细收入、利润率、安全与合规等运营数据尚未披露。
封面要素
公司概况
Sharpa 是一家总部位于新加坡的具身智能机器人公司,由 Hesai 联合创始人于 2024 年底创立。公司研发触觉操作硬件与人形机器人,以 SharpaWave 灵巧手和 North 机器人平台为主要公开产品,目前正将相关系统投入真实商业餐饮场景中开展测试验证。
- 创始人
- Li Yifan (David Li), Xiang Shaoqing, Sun Kai
- 创立地点
- Singapore
- 总部
- Singapore
- 产品
- 面向真实场景中接触密集型任务的 SharpaWave 触觉灵巧手、CraftNet 操作模型和 North 人形机器人。
- 客户
- 餐饮服务、机器人实验室及其他真实场景中的机器人操作任务。
- 商业模式
- 硬件产品销售叠加部署与服务安排;定价与经常性收费条款尚未披露。
- 阶段
- Growth-stage private
- 融资情况
- MarketScreener 报道称其于 2026 年 8 月完成一轮 $669M 的可转换优先股融资,投后估值达 $3.3B。其他数据库记录了 $670M 累计披露金额或 $784M 融资总额;各渠道的轮次口径与发生时间尚未拉通对齐。
执行摘要
主要优势
- 头部战略机构与风投机构资金支持充沛
- 触觉操作与人形机器人产品栈定位鲜明
- 已在真实客户场景开展商业试点
主要风险
- 公开证据尚不足以证明长期现场可靠性或安全表现
- 收入、利润率、烧钱速度与现金跑道仍未披露
- 供应商与部署集中度带来执行依赖风险
未决问题
- 经审计的收入、利润率与烧钱速度披露
- 跨数周的现场可靠性与人工介入数据
- 实际部署的安全、合规与隐私证据
- Dairy Queen 试点的合同条款与经常性经济模型
- 经核对的累计融资与单轮融资历史
- 明确高管角色与运营职责的公司法定备案文件
目录
01公司概况
1.1 企业定位与全球运营版图
Sharpa 成立于 2024 年底,是一家专注于通用机器人与触觉操作系统的具身人工智能公司。公开资料显示,其总部设在新加坡,商务运营位于 Mountain View,主要制造与研发基地设在上海 [CO001]。MarketScreener 报道称,公司于 2026 年 8 月完成 $669 million 可转换优先股融资,投后估值为 $3.3 billion(RMB 22 billion)[CO002]。一份未注明日期的行业资料称 Sharpa 尚未产生收入,但其实际确认收入及试点部署的经济效益均未获第三方独立证实 [CO003]。人员规模指标从 LinkedIn 关联的 74 人到媒体报道的超 100 人不等;两者统计口径不同,并非经核实的人员区间 [CO004]。公司仍在各运营地招聘 AI 工程与商业化岗位 [CO005]。
| 指标 | 数值 | 日期 | 可信度 | 尽调盲区 |
|---|---|---|---|---|
| 估值 | $3,300M (RMB 22B) | 2026-08-28 | 高 | 一级市场轮次条款、清算优先权及老股转让成交量未经证实 |
| 报道的 8 月融资金额 | $669M 轮次(MarketScreener 报道) | 2026-08-28 | 高 | 非各方对齐的累计融资总额;其他数据源的融资分类口径存在差异 |
| 收入运行率 | 未注明日期的行业画像称公司尚无收入;当前营收运行率未披露 | 资料查阅于 2026-09-25 | 中 | 尚无经审计的当前收入,也未披露试点合同金额 |
| 客户数 | 1 个商业试点(Dairy Queen) | 2026-08-29 | 中 | 未披露付费客户合同与科研/实验室部署的区分 |
| 员工数 | LinkedIn 关联成员 74 人;据报道人员超过 100 人 | 2026-08-20 | 中 | 统计人群与观察日期不同;非经核实的员工数区间 |
| 办公地点 | Singapore(总部)、Shanghai(研发/制造)、Mountain View(业务运营) | 2026-09-06 | 高 | 设施确切面积与产线产能未经验证 |
封面指标综合自公司披露、LinkedIn 名录及截至 2026 年 9 月的财务报道;无法获取非公开财务审计数据。
[CO001, CO002, CO003, CO004, CO005]1.2 核心管理架构与治理格局
公开报道将 Hesai Technology 联合创始人 Li Yifan(David Li)、Xiang Shaoqing 与 Sun Kai 列为 Sharpa 创始人,并称该项目独立于 Hesai [CO006]。创始团队的正式高管头衔与日常管理职责目前仍存分歧。Seedtable 将 David Li 列为首席执行官(CEO)、Xiang Shaoqing 列为首席技术官(CTO)、Sun Kai 列为高管(Executive),而 Embodied AI Frontier 则称三位创始人均未担任高管职务 [CO007]。所审阅资料既未解释上述分歧缘由,也未证实各创始人在产品管理与工程上的具体分工。创始人履历虽有助于战略制定与融资推进,但其实际管理权限、运营责任及企业关联仍需依赖工商登记文件与管理层核实 [CO006, CO007]。
| 人员 | 职务 | 背景 | 职能分工 | 关键人依赖 |
|---|---|---|---|---|
| Li Yifan (David Li) | 联合创始人;据报道 CEO 头衔存在争议 | Hesai Technology 联合创始人;据报道在 Hesai(NASDAQ: HSAI)担任 CEO。 | 确切的管理、融资与商业化职责未经验证。 | 职责范围尚未确认,无法评估运营依赖度。 |
| Xiang Shaoqing | 联合创始人;据报道 CTO 头衔存在争议 | Hesai Technology 联合创始人;据报道在 Hesai 担任 CTO。 | 对 Sharpa 硬件与执行器工程的直接职责未经验证。 | 职责范围尚未确认,无法评估运营依赖度。 |
| Sun Kai | 联合创始人;正式管理职务未获证实 | Hesai Technology 联合创始人;据报道在 Hesai 担任首席科学家。 | 对 CraftNet 与触觉模型研发的直接职责未获证实。 | 职责范围尚未确认,无法评估运营依赖度。 |
各方信息源均确认三位创始人身份,但在具体高管头衔上存在分歧。名录头衔无法替代官方申报核实,过往背景也不能证明其当前在 Sharpa 承担实际运营职责。
[CO006, CO007]1.3 融资历史与战略投资方银团
Sharpa 虽已获得风险投资机构与科技巨头的支持,但现有融资数据未能形成相互印证的融资历史。Physix Frontier 记录截至 2026 年 8 月公司累计披露融资额为 $670 million,且未明确具体轮次,亦无带日期的种子轮及早期研发阶段明细 [CO014]。MarketScreener 则单独报道称公司于 2026 年 8 月 28 日完成 $669 million 可转换优先股融资,投后估值为 $3.3 billion [CO002]。上述披露数据不可直接相加,亦不能将单轮融资与累计总额混为一谈。据报道,投资方包括 Alibaba、Tencent、Meituan 与 Meituan Dragon Ball、JD.com、HongShan 以及 Luminous Ventures [CO008]。产业投资方的入局或能提供落地场景渠道,但投资方名单本身无法证实治理权限与客户排他性。报道还显示 Hesai 亦是其执行器供应商,该供应关系应与股权归属或控制权分别评估 [CO013]。
| 利益相关方 | 角色 | 控制权或经济重要性 | 尽调核实要点 |
|---|---|---|---|
| Alibaba Group Holding Limited | 企业战略投资方 | 核心科技巨头基石;具备战略级云/AI 算力支持与国内生态分发杠杆。 | 核实董事会席位、商业排他性条款及算力补贴情况。 |
| Meituan / Meituan Dragon Ball | 企业战略投资方 | 头部即时零售与配送平台;商业操作机器人的潜在落地渠道。 | 核实零售/餐饮门店网络的试点准入及商业化权益。 |
| Tencent Holdings Limited | 企业战略投资方 | 头部科技巨头;具备游戏、仿真及企业云渠道分发触达能力。 | 评估联合 AI 研究安排,以及未来股权融资中的投票权与否决权。 |
| JD.com | 企业战略投资方 | 电商与物流龙头;拥有仓储自动化与履约操作场景。 | 考察仓储测试环境准入权限及物流机器人研发时间表。 |
| HongShan (Sequoia China) | 一线风险投资机构 | 领投的财务投资机构;拥有广泛创投网络与公司治理指导能力。 | 审查保护性条款、清算优先权及后续跟投权。 |
| Hesai Technology (Co-founders / Supplier) | 零部件供应商及创始人前序企业 | 关键硬件供应商(执行器、LiDAR);由 Sharpa 创始团队创立但无直接股权关联。 | 审计执行器转让定价、供应协议排他性及知识产权边界。 |
选取 2026 年 8 月融资轮次及零部件供应链中的主要利益相关方;完整的股权结构表与各方具体持股比例仍属非公开信息。
[CO002, CO008, CO013]1.4 产品里程碑、商业试点与落地实态
Sharpa 的产品路线从灵巧手切入,再向整机延伸;公司优先发展接触密集型触觉操作,而非先解决全身移动 [CO009]。2025 年 10 月,旗舰产品 SharpaWave 仿生手实现量产,配有 22 个主动自由度及触觉传感器,荣获 CES 2026 创新奖 [CO009, CO010]。依托该操作平台,Sharpa 推出全尺寸人形机器人 North,在餐饮零售场景启动商业验证 [CO011]。2026 年 8 月 29 日,Sharpa 将 North 部署于 Dairy Queen 上海暴风雪与汉堡旗舰店,与后厨员工协同执行包含 55 个步骤的甜品制作流程 [CO011]。然而,独立分析师严谨的技术审视指出,平均无故障时间(MTBF)、人工干预率以及真实全工班经济可行性等核心运营指标,目前均未获证实,且未经第三方独立验证 [CO012]。此次部署是 Hesai 供应的执行器与 Sharpa 触觉基座模型在真实世界中的一次早期压力测试,而规模化商业可行性仍取决于多工况下的长期可靠性数据 [CO012, CO013]。
梳理从创立截至 2026 年 9 月的公司治理、融资、产品及商业化关键里程碑。
日期反映据报道的交易公告、活动发布与产品推出时间。
[CO006, CO008, CO009, CO010, CO011, CO012]1.5 深度图表
02市场分析
2.1 市场边界与操作自动化范畴
Sharpa 的市场边界由接触密集型操作与具身物理智能(Physical AI)定义,而非传统的固定式工业自动化。传统工业机器人依赖确定性工作站、硬质工装和结构化环境,因此难以适应动态商业空间和精细的人工作业。Sharpa 研发通用人形机器人及拟人末端执行器,目标是在无需改造场地或定制工作站的情况下完成无人干预的整班作业。其目标市场涵盖具身 AI 硬件、高自由度触觉手、人形机器人本体及多模态基础模型授权;明确排除传统固定式单轴机械臂、刚性输送线基础设施以及纯计算机视觉检测系统。[CM001, CM002, CM003, CM004]
| 类别 / 细分赛道 | 纳入范畴的支出 | 排除范畴的支出 | 目标买方 / 付费方 | 战略相关性 |
|---|---|---|---|---|
| 灵巧末端执行器与触觉手 | 仿人 22-DoF 触觉手、传感器阵列、触觉遥操作手套 | 刚性双指平行动作夹爪、磁吸工装 | 具身智能研究人员、人形机器人整机 OEM | 解决复杂抓取操作瓶颈的核心零部件层级 |
| 全身人形服务平台 | 67-DoF 人形机器人整机、移动底盘、集成双臂算力单元 | 固定基座工业机械臂、自动导引运输车(AGV) | 商业服务企业、快餐零售运营方 | 在现有工人工作流程中实现端到端自主替代人力 |
| 物理 AI 基础模型 | 分层 VTLA 模型(System 0/1/2)、隐空间世界模型、仿真数据 | 基于固定规则的局部运动规划、孤立单模态 LLM API | 机器人开发者、企业级方案集成商 | 支持实时触觉闭环抓取的软件智能层 |
| 存量设施改造自动化 | 免改动门店部署套件、本地边缘服务器、集成工作流 | 场地重构、输送线改造、建筑结构性重新工装 | 门店运营总监、自动化工程副总裁 | 无需重资产建筑重构,保留现有商业占地格局 |
市场边界界定:将灵巧触觉操作硬件及基础模型,与传统的固定工位自动化方案明确区分。
[CM001, CM002, CM003, CM004]2.2 TAM、SAM 与证据约束下的市场测算视角
评估精细操作的潜在市场空间,需要从宏观劳动力替代、专用硬件市场及商业服务工作流等多视角进行综合分析。尽管零售、酒店餐饮及轻工制造领域的体力劳动替代构成数万亿美元的宏观市场,但中短期可触达支出受到制造成本、零部件可用性及操作可靠性的制约。短期内,机器人主机厂(OEM)与具身 AI 研发团队的零部件采购构成数亿美元的可服务获得市场(SAM),这一判断得到 Hesai 执行器模组出货突破 10,000 套等早期产业化节点的支撑。快餐连锁门店的商业试点则展现出向运营级门店工作流扩张的早期迹象。[CM005, CM006, CM007, CM008]
| 测算视角 / 层级 | 覆盖区域 | 预估规模 / 金额 | 测算方法 / 依据 | 置信度 | 局限性与现实困境 |
|---|---|---|---|---|---|
| 全球体力劳动 TAM | 全球 | $3.5T+ 人工支出 | 汇总零售、餐饮、酒店及轻工制造领域的重复性劳动工时 | 低 | 极宽泛的理论上限;假定在各类任务中完全达到人类劳动等效水平 |
| 人形机器人与灵巧硬件 SAM | 北美与东亚 | 到 2030 年达 $28B - $45B | 商业服务领域通用人形机器人及触觉灵巧部件的预测出货量 | 中 | 高度受制于物料清单(BOM)成本下降幅度及执行器量产爬坡速度 |
| 快餐与零售轮班自动化 SOM | 重点一线城市枢纽 | 初期可触达 $1.8B - $3.2B | 基于高客流 QSR 食品制作与免工位改造门店轮班运营的自下而上测算 | 中 | 受限于实际后厨占地空间、吞吐节拍周期以及多店试点转化率 |
| 具身智能零部件与科研市场 | 全球高校科研机构与 OEM 聚集区 | 近期达 $450M - $750M | 面向 AI 研究机构与机器人 OEM 销售灵巧触觉手(Wave)及遥操作系统 | 高 | 受限于活跃人形机器人研究项目与研发团队的总数 |
综合自物理 AI 行业研究、执行器供应商交付节点以及早期企业试点参数。
[CM005, CM006, CM007]市场规模测算层级由宽收窄,从理论上的全球重复性体力劳动替代,逐级下收至近期硬件与核心零部件部署。
各层宽度仅表示层级,不代表数值比例。
数值综合了宏观劳动力统计数据、自下而上的机器人零部件出货量以及商业化 QSR 试点覆盖范围。
[CM005, CM006, CM007, CM014]2.3 买方客群细分与商业落地路径
Sharpa 的商业买方格局呈现分化态势,主要划分为近期的技术开发者客户与中长期的运营型企业客户。科研实验室与人形机器人 OEM 通过工程与采购预算采购 22 自由度 Wave 灵巧手,以加速物理智能研究并规避自研灵巧手的长研发周期。相比之下,连锁快餐运营商与零售物流主管等商业企业客户,则依据工班级投资回报率(ROI)、任务成功率及直接人工替代效果来评估整机人形机器人系统。企业客户的预算决策权掌握在门店运营副总裁与首席转型官手中,而非研发部门。商业落地路径正从受控的单店验证测试向多门店加盟网络铺开演进。[CM008, CM009, CM010, CM011]
| 客户客群 | 目标买方 | 核心使用者 | 经济买单方 | 目标工作流 | 预算归属方 | 关键采买触发点 |
|---|---|---|---|---|---|---|
| 具身智能科研实验室 | 课题组负责人(PI)/ 实验室主任 | 机器人与机器学习研发工程师 | 科研基金 / 研发部门 | 触觉数据采集、遥操作、基础模型基准评测 | 研发负责人 | 对兼容 NVIDIA Isaac Lab 的高分辨率 22-DoF 触觉感知有明确需求 |
| 人形机器人整机 OEM | 硬件工程副总裁 / 首席架构师 | 系统集成工程师 | OEM 采购 / 物料清单(BOM)预算 | 将灵巧手硬件集成至第三方人形机器人本体底盘 | 首席技术官(CTO)/ 采购副总裁 | 加速产品上市进程,免去自研专有灵巧手的工程负担 |
| 快餐连锁企业(QSR) | 运营副总裁 / 首席转型官 | 后厨门店员工与店长 | 门店运营支出(OpEx)/ 资本支出(CapEx) | 高频触碰类食品制作(如 55 步 DQ 暴风雪制作流程) | 门店运营负责人 / 加盟商理事会 | 无需改造工位即可跑通整班作业,且单店 ROI 转正 |
| 轻工与电子装配 | 制造自动化总监 | 产线操作工与车间主管 | 工厂自动化资本预算 | 精密剪线、包装及脆弱微小零件插装 | 全球制造运营副总裁 | 以亚毫米级控制精度,确保最后 1% 装配环节的高可靠性 |
买方细分综合自商业企业试验、科研合作及 OEM 硬件集成渠道。
[CM008, CM009, CM010, CM011]2.4 增长驱动力、尽调盲区与推广约束
服务业长期缺工、时薪上涨,加上多模态触觉基础模型快速进步,都在推动精细操作自动化。人形机器人若无需大规模改造就能进入现有商业场所,便有望提升用工流动率较高的服务行业效率。但采用仍受几项硬约束:数周商业部署中的平均无故障时间、高峰期人工介入频率、任务完成速度和日常维护开支,均缺少经独立核实的运营数据。扩产还取决于高扭矩紧凑型执行器的制造良率。企业买方和投资人需要权衡早期商业宣传与长期可靠性数据的缺口。[CM011, CM012, CM013]
2.5 深度图表
03竞争格局
3.1 竞争版图与直接竞品画像
具身智能与物理操作领域的竞争格局包含专用灵巧手制造商、整机人形机器人研发商以及根基深厚的传统自动化供应商。Sharpa 凭借围绕高分辨率触觉精细操作构建的全栈硬件与模型,占据了差异化生态位。在灵巧手细分领域,直接竞品包括 Fourier Intelligence 与 Inspire Robots 等中国专用机器人制造商,两者均面向工业与科研客户供应多自由度末端执行器。此外,Unitree Robotics 等一体化人形机器人研发商构成相邻竞争对手,其完整机器人平台目前优先侧重运动能力与基础操作,而非高精度触觉反馈。依托 2026 年披露的逾 $669 million 融资及 $3.3 billion 估值,Sharpa 推进旗舰 22 自由度 SharpaWave 灵巧手与 67 自由度 North 人形机器人平台的协同研发,后者专为人类服务环境下的商业任务执行而设计。[CP001, CP002, CP004, CP011]
| 竞争对手 | 品类 | 规模 / 融资 | 目标客群 | 差异化优势 | 局限性 |
|---|---|---|---|---|---|
| Sharpa | 灵巧手与人形机器人 | $669M 融资轮;$3.3B 估值(2026 年) | 企业服务与零售自动化 | 搭载 DTA 的 22-DoF 触觉手;垂直模型一体化整合 | 处于商业化早期;多日连续无故障运行时间(MTBF)未经验证 |
| Fourier Intelligence | 直接灵巧硬件竞品 | 老牌中国机器人 OEM | 康复与通用机器人科研 | 量产灵巧手产品,具备广泛硬件分发网络 | 与专用传感器相比,触觉阵列密度较低 |
| Inspire Robots | 直接灵巧硬件竞品 | 已获多轮融资的中国机械臂与末端厂商 | 工业灵巧操作与检测 | 深度成本优化的微型直线执行器集成 | 缺乏一体化全身人形基础平台 |
| Unitree Robotics | 相邻人形机器人 OEM | 高出货量四足及人形机器人制造厂商 | 商用安防、消费级与科研 | 硬件规模化量产能力与激进的定价策略 | 多采用结构更简易的末端夹具,而非精细触觉手 |
规模与融资数据基于已披露的 2026 年融资事件及公开媒体报道;各项能力体现主要产品规格参数。
[CP001, CP002, CP004, CP005, CP011]3.2 性能基准与定价体系
评估操作性能揭示了传统平口夹爪与多自由度触觉手之间的本质取舍。传统工业自动化依赖平口夹爪,在结构化工作流中可靠性极高,但对位精度要求极为严苛。轴孔装配实测显示,受反复重定位影响,平口夹爪每次尝试耗时 20 至 70 秒;而 SharpaWave 凭借手内动态姿态调整,仅需 5 至 10 秒即可完成相同任务。此外,在仿真环境中训练的触觉反馈策略在物理硬件上的任务完成率达 80%,而易受视线遮挡影响的纯视觉模型仅为 30%。商业层面上,非上市机器人企业的公开标价均高度保密;灵巧手通常作为高端科研或企业开发者套件出售,而整机人形机器人则通过定制化企业试点协议交付,而非标准化目录定价。[CP003, CP006, CP008, CP010]
| 采买评估标准 | Sharpa Wave / North | Fourier / Inspire | 现状平行动作夹爪 | 对应尽调评估结论 |
|---|---|---|---|---|
| 主动自由度(DoF) | 22 DoF(手部)/ 67 DoF(人形整机) | 单手 6-12 DoF | 1-2 DoF(平行对夹动作) | Sharpa 在手内复杂姿态调整的运动学灵巧度上处于领先地位 |
| 触觉感知分辨率 | 0.02N 灵敏度;动态触觉阵列(DTA) | 基础力传感器 / 指尖压力感知 | 开关式触点检测或扭矩限位 | 触觉反馈带来 80% 的真实环境成功率,而纯视觉方案仅为 30% |
| 操作循环节拍耗时 | 5-10 秒(轴孔装配) | 估计为 15-30 秒 | 单次尝试 20-70 秒 | 灵巧手消除了夹爪所需的低效反复对准动作 |
| 仿真到真实环境迁移(Sim-to-Real) | 经 NVIDIA Isaac Lab 与 ADEPT 验证 | 专有 SDK / ROS 功能包 | 确定性运动规划 | 共享灵巧先验将训练步数从 9B 降至 3B |
据技术发布材料披露,性能指标来源于实机操作试验与仿真基准测试。
[CP002, CP003, CP006, CP010]| 产品品类 | 打包模式 | 所含功能与权益 | 公开定价 / 折扣策略 | 战略尽调含义 |
|---|---|---|---|---|
| 商用灵巧手 | 硬件整机销售 + SDK 授权许可 | 多 DoF 灵巧手、触觉控制器、Isaac Lab 驱动程序 | 非公开商业报价(公开标价未披露) | 相较于简易气动夹爪,单机成本较高 |
| 全身人形整机平台 | 直接企业级部署 / 试点合同 | North 机器人本体、算力套件、视触觉模型 | 定制化企业合同;处于试点阶段定价 | 试点资金投入门槛较高,制约了早期落地规模 |
| 执行器与关节基础设施 | Tier-1 零部件供应合同 | 直驱模块、关节壳体、编码器 | 供应商阶梯批量定价(基于 Hesai 10k+ 规模) | 上游模块出货规模有助于压低远期物料清单(BOM)成本 |
各非公开机器人平台均未公开披露标价;所述结构代表试点部署中观察到的商业合同模式。
[CP005, CP007, CP013]3.3 转换成本、供应链依赖与多家共用
客户转换成本与供应商绑定程度取决于软件生态融合深度与核心零部件供应架构。Sharpa 在战略上将硬件与 NVIDIA 的 Isaac Lab 及 ADEPT 强化学习框架深度集成,开发者得以缩减仿真训练周期,将预训练精细操作策略直接迁移至 SharpaWave 硬件。软件对主流仿真技术栈的兼容固然降低了开发者的采用门槛,但同时也削弱了机器人应用开发者的长期软件转换壁垒。在硬件供应链端,核心执行器模组由 Hesai 供应,其专用机器人执行器产能已突破 10,000 套。然而,Hesai 亦向 Unitree、Galbot 和 Galaxea 等超过 50 家具身智能机器人公司供应激光雷达与执行器部件。多个 OEM 共用这类一级供应商网络,降低了单一人形机器人厂商垄断精密执行器的可能。[CP005, CP006, CP009]
3.4 护城河韧性与商业替代风险
Sharpa 竞争壁垒的持久性取决于其能否保持触觉基座智能的领先优势,而非仅依赖自研机械硬件。随着上游一级零部件供应商的产品线从灵巧手执行器扩展至标准化整机人形关节,机械硬件能力在各竞品机器人厂商之间将不可避免地趋于同质化。因此,Sharpa 的核心防御屏障在于其自研的 CraftNet 视触语动(vision-touch-language-action)模型、动态触觉传感阵列以及高频实时抓握控制。显著的商业化风险依然存在:尽管 Dairy Queen 门店 55 步餐饮制作等试点展示了现实可用性,但验证平均无故障时间、自主运行时间及人工干预率的独立跨日数据集仍未获证实。若无法在现场可靠性与相较于低成本人工或专用固定自动化的单机经济效益上建立优势,长期的商业替代风险仍不可忽视。[CP007, CP012, CP013, CP014]
3.5 深度图表
04财务分析
4.1 营收模式与商业化路径
现有公开证据仅支持得出非常有限的收入模式结论。Multiples 将 Sharpa 的商业模式归类为产品销售,而 The Frontier of Embodied AI 则指出该公司尚未产生收入。因此,2026 年 8 月 North 在上海 Dairy Queen 门店的部署更适合作为商业化进展的证据,而非已确认收入的证明。现有资料均未披露已签合同金额、硬件出货量、标价、实际平均售价(ASP)、软件订阅费、维护费或收入确认政策。外部来源仅将硬件销售归类为收入来源;餐厅部署收入以及软件或服务收入仍属投资假设,在审阅合同或发票前不应计入基准情景。[CI001, CI002, CI003, CI004, CI005]
| 收入流 | 变现机制 | 计价单位 | 当前体量 / 状态 | 收入质量 | 尽调调取项 |
|---|---|---|---|---|---|
| 硬件直销 | 第三方数据库将该业务归类为产品销售 | 按单台机器人计 | 据报道未产生收入;标价与实际成交价均无法获取 | 在披露付费出货量及利润率之前,商业模式未经证实 | 发票、出货量、平均售价(ASP)、折扣、退货以及收入确认政策 |
| 餐饮场景部署 | North 在实际运营的 Dairy Queen 门店执行工作流 | 合同计价单位未披露 | 处于实地部署状态;合同金额与已确认收入无法获取 | 由于试点商业条款与付费主体未披露,具体质量未知 | 客户合同、付费方身份、部署费、经常性费用以及补贴条款 |
| 软件与服务 | 已留存证据中未见独立的对外在售产品 | 不可获取;已从基准情景收入中剔除 | 缺乏具体在售方案、价格及续约证据,无法纳入投资评估 | SKU 目录、维保条款、软件费、搭售率以及续约数据 |
截至 2026-09-25 的局部公开视角;null 表示无明确支持的商业计价单位,且推测的软件或服务收入已从基准情景中剔除。
[CI002, CI003, CI005]| 在售产品 / 方案 | 合同模式 | 公开标价与实际成交价 | 折扣与未知项 | 证据来源 |
|---|---|---|---|---|
| SharpaWave 触觉手 | 产品销售分类;具体合同形式未披露 | 公开标价未披露;实际平均售价(ASP)无法获取 | 批量折扣、退货率、质保及技术支持搭售情况未知 | Multiples 数据库;产品复杂度获 Brett Adcock 侧面印证 |
| North 人形机器人平台 | 合同模式无法获取 | 公开标价与实际成交价均无法获取 | 产权归属、租赁、联合开发及补贴条款未知 | RoboMorrow 部署分析 |
| Dairy Queen 工位自动化 | 实际运营部署;计费模式无法获取 | 商业合同价值与实际确认收入均无法获取 | 付费方、进场安装成本、服务范围及合作伙伴补贴均未知 | RoboMorrow 部署分析 |
所留存的公开来源均未披露公开标价或实际成交价;因此各项合同模式均标为无法获取,而非主观推断。
[CI005, CI009, CI012, CI015]从商业任务执行转化至企业经常性收入与毛利率的传导路径。
| 节点 / 连接 | 节点 / 起点 | 终点 | 说明 |
|---|---|---|---|
| 节点 1 | 餐饮订单需求 [store-demand] | ||
| 节点 2 | North 人形机器人执行 55 步工序 [robot-execution] | ||
| 节点 3 | 实际结算单价 × 付费工作量 [raas-billing] | ||
| 节点 4 | 执行器与零部件主营业务成本(COGS) [hardware-cogs] | ||
| 节点 5 | 维护与硬件磨损支出 [field-service] | ||
| 节点 6 | 毛利润 / 边际贡献 [net-contribution] | ||
| 连接 1 | 餐饮订单需求 [store-demand] | North 人形机器人执行 55 步工序 [robot-execution] | 产生任务量 |
| 连接 2 | North 人形机器人执行 55 步工序 [robot-execution] | 实际结算单价 × 付费工作量 [raas-billing] | 转化为可计费产出 |
| 连接 3 | North 人形机器人执行 55 步工序 [robot-execution] | 执行器与零部件主营业务成本(COGS) [hardware-cogs] | 消耗硬件寿命 |
| 连接 4 | 实际结算单价 × 付费工作量 [raas-billing] | 毛利润 / 边际贡献 [net-contribution] | 确认收入 |
| 连接 5 | 执行器与零部件主营业务成本(COGS) [hardware-cogs] | 毛利润 / 边际贡献 [net-contribution] | 扣除直接成本 |
| 连接 6 | 维护与硬件磨损支出 [field-service] | 毛利润 / 边际贡献 [net-contribution] | 扣除日常运营开支 |
本图仅为概念性投资评估测算推导;留存资料中没有任何公开来源披露 Sharpa 的实际定价、合同模式、成本分摊或毛利率。
[CI002, CI005, CI007, CI015]4.2 成本结构、毛利率与单机经济模型
基于现有公开证据,Sharpa 的成本结构无法被准确量化。Wave 灵巧手拥有 22 个自由度,每个指尖集成了 1,000 多个触觉传感像素,体现出高零部件密度的产品特性;RoboMorrow 报道称供应商 Hesai 为 Sharpa 提供执行器模组。Dairy Queen 的部署在现有门店设备环境下执行包含 55 个步骤的工作流,但现有资料均未披露物料清单(BOM)、执行器采购价、制造良率、安装人工、保修准备金、现场运维成本、能耗或折旧假设。RoboMorrow 亦明确指出当前缺失的核心运营变量:任务成功率、人工干预频次、订单周期耗时、设备正常运行时间、平均无故障时间以及整班作业的经济性。在这些指标经过多门店、跨时段实测,并与客户实际支付价格核对之前,毛利率和客户投资回收期都无从判断。[CI006, CI007, CI008, CI009]
| 指标 | 预估数值 | 置信度 | 核心意义 | 尽调调取项 |
|---|---|---|---|---|
| 硬件物料清单(BOM) | 低 | 决定硬件毛利率门槛 | 零部件级 BOM、供应商报价、良率、质保预留金以及组装人工成本 | |
| Dairy Queen 出餐周期耗时 | 低 | 决定餐厅吞吐量与峰值产能 | 跨多小时客流班次的工时实测研究 | |
| 人工干预频率 | 低 | 直接影响实际节省的人力成本与 ROI | 每 1,000 笔已完成订单的干预日志 | |
| 客户回本周期 | 低 | 大客户加盟商采纳的核心决策因素 | 加盟商总拥有成本(TCO)模型 | |
| 硬件制造毛利率 | 低 | 体现实现盈利所需的规模门槛 | 按出货量阶梯拆分的实际 ASP、COGS、质保、运费、安装与维保成本 | |
| 现金消耗与现金跑道 | 低 | 检验融资资金能否支撑团队跑通商业里程碑 | 现金余额、月度烧钱速度、已承诺资本开支、负债、受限资金及悲观情景下的现金跑道 |
表中所有 null 均代表留存公开证据中缺失该指标,并非数值为零;各行分别明确了评估该指标所需的非公开尽调证据。
[CI006, CI007, CI008, CI009, CI015]4.3 资金充裕度与融资依赖性
Multiples 报道称其历经三轮融资累计吸纳 $784 million 资金,MarketScreener 则记录其于 2026 年 8 月 28 日完成一轮发行可转换优先股的融资,募集 $669 million,投后估值达 $3.3 billion。上述融资规模展现出其获取资本的能力,但无法证实当前的在手现金余额与跑道期(runway):证据中未提供期末现金余额、月度烧钱率、受限资金条款、负债、项目融资、资本承诺或计划资金使用时间表。LinkedIn 显示 Sharpa 规模为 51 至 200 名员工,官方招聘页面则列有 3 个 AI 工程岗位、2 个商业化岗位和 1 个实习生岗位。这些仅反映运营规模信号,不能作为测算资金消耗率的依据。必须审阅管理层财务报表及正式融资文件,方能评估该轮资金能否支撑商业化进程跨越至下一轮关键里程碑。[CI001, CI003, CI004, CI010, CI011, CI012]
4.4 财务研判与投资尽调障碍
当前财务判断受证据缺口制约,不只是因为公司尚无收入。$3.3 billion 的投后估值与 $669 million 融资抬高了市场预期,但现有证据无法证明公司已有收入、实现了怎样的成交价格、获得客户复购或实现正向单机贡献。最有力的商业证据只有一家门店的实际餐饮工作流;最重要的反向证据则是干预率、周期时间、正常运行时间、可靠性和整班经济性仍不明。因此,投资逻辑取决于公司能否在资金耗尽前把技术表现转化为可复制的部署。投资判断应以以下尽调结果为前提:经审计的收入、分批次部署数据、正式客户合同、BOM 与服务成本明细、现金流与烧钱率报告、融资债权债务条款,以及经董事会批准、将下一轮融资触发条件与可量化毛利率及可靠性指标绑定的业务规划。[CI002, CI007, CI008, CI013, CI014]
4.5 深度图表
05产品与技术
5.1 产品定义与物理硬件技术栈
Sharpa 提供软硬件垂直一体化的具身智能平台,旨在解决以人为中心的运营环境下的精细操作难题。在物理层,公司的核心商业化产品为 Sharpa Wave 仿生手,具备 22 个主动自由度,外形按人手尺度设计。每个指尖均集成自研动态触觉阵列(Dynamic Tactile Array),提供超过 1,000 个触觉传感像素与 5 mN 的压力灵敏度,实现高灵敏度的闭环力控调节。Wave 灵巧手于 2025 年 10 月实现量产,并获得 CES 2026 创新奖。在灵巧手之上,Sharpa 推出了 North 全尺寸人形机器人,作为一款拥有 67 个自由度的移动与操作平台,目前正在冰淇淋制作等零售及服务场景中开展商业试点验证。[CE001, CE002, CE009, CE010]
| 模块或资产 | 目标用户 | 状态与成熟度 | 核心差异化 | 尽调缺口 |
|---|---|---|---|---|
| Sharpa Wave Hand | 机器人实验室与人形机器人 OEM 厂商 | 量产(2025 年 10 月) | 配有 22 个主动 DoF,单指尖 Dynamic Tactile Array 触觉阵列超 1000 像素 | 磨损接触循环下的传感器长期耐用性 |
| North Humanoid | 商业与服务业运营商 | 商业试点(2026 年) | 集成精细移动操作能力的 67 DoF 全身平台 | 多班次连续作业下的平均故障间隔时间(MTBF) |
| CraftNet Model | 具身智能开发者与运营商 | 内部部署(2026 年) | 分层 VTLA 架构,配有 100 Hz 的 System 0 触觉控制 | 在校准测试场景之外的真实世界泛化能力 |
| Sharpa Wave SDK | 软件集成工程师 | 开发者版本(2026 年) | 面向定制机器人技术栈与 Isaac Lab 仿真的跨平台 API | API 向后兼容性与文档完备度 |
成熟度状态、差异化特性与架构依赖关系整理自 Sharpa 官方技术披露、开发者代码库及行业分析师研报。
[CE001, CE002, CE009, CE010]5.2 CraftNet VTLA 架构与高频控制系统
为驱动其物理操作硬件,Sharpa 开发了端到端分层视触语动(VTLA)基础模型 CraftNet,该模型依托认知与运动三层分级架构构建。System 2 作为「推理大脑」,以约 1 Hz 的频率运行开源视觉语言模型,负责理解人类语义指令、长程目标任务拆解及空间操作规划。System 1 作为「运动大脑」,以 10 Hz 运行以生成大范围运动轨迹以及触碰前的上半身与手部姿态构型。尤为关键的是,System 0 作为「高频交互大脑」以 100 Hz 运行,实时接入连续触觉与本体感受遥测数据,接管「最后一毫米」接触阶段。当精细物理操作遇到非预期阻力时,System 0 会向 System 1 反向传播状态更新,实时修正物理动作意图。[CE003, CE004, CE005, CE008, CE014]
涵盖从底层硬件触觉阵列到顶层 System 2 认知规划的六层产品架构。
- 认知规划层(System 2)
- 运行频率约 1 Hz,基于开源视觉语言模型(VLM)将复杂人类指令拆解为时序操作步骤。
- 粗粒度动作层(System 1)
- 运行频率约 10 Hz,基于多模态输入生成上半身运动轨迹与接触前的手部姿态。
- 交互控制层(System 0)
- 运行频率约 100 Hz,负责处理触觉反馈,并在物体交互的最后数毫米微距内实时调节微电机输出力。
- 开发者与仿真中间件
- 包括 Sharpa Wave SDK、Isaac Lab 触觉传感器资产以及 Sharpa Pilot 配置软件。
- 物理硬件层
- 涵盖 22-DoF 自由度的 Sharpa Wave 灵巧手与 67-DoF 自由度的 North 全身人形机器人平台。
- 传感子系统
- 动态触觉阵列(Dynamic Tactile Array),每个指尖具备超过 1000 个触觉像素与 5 mN 的压力灵敏度。
运行频率指标与架构分层定义均源自 Sharpa CraftNet 白皮书披露信息。
[CE003, CE004, CE005, CE010, CE014]5.3 开发者生态、集成方案与仿真工具
Sharpa 为开发者和机器人集成商提供了开放的软件生态,搭建起实验室仿真与物理机器人部署之间的桥梁。开源的 Sharpa Wave SDK 提供原生 C++ 和 Python 运行时库、设备驱动及通信头文件,便于将 Wave 硬件接入既有的 ROS 及定制机器人控制栈。作为 SDK 的补充,Sharpa Pilot 提供了用于设备配置、运行状态监测及固件部署的管理应用程序。模型训练与仿真环节,Sharpa 在 NVIDIA Isaac Lab 中发布了运动学 URDF、USD 及 XML 仿真资产,并配套提供强化学习环境。这一套基础设施使开发者能够在将精细操作策略部署至实体机器人之前,通过传感手套仿真接触动力学并重定向人类遥控操作轨迹。[CE006, CE007, CE011]
| 用户任务 | 当前工作流 | Sharpa 解决方案 | 可量化收益 | 技术局限 |
|---|---|---|---|---|
| 精细物体操作 | 纯视觉轨迹规划,错误率高 | CraftNet System 0 提供 100 Hz 动态触觉反馈 | 亚毫米级闭环抓取微调 | 依赖传感器校准与高速总线 |
| 商业餐饮备餐 | 人工备餐或刚性自动化工装 | North 人形机器人部署双 Wave 灵巧手执行灵巧任务 | 自动化执行冰淇淋打注等多步骤任务 | 试点规模部署,依赖人工监管兜底 |
| 机器人技术栈集成 | 私有定制驱动与人工动作重定向 | Sharpa Wave SDK 与 Isaac Lab 触觉传感器资产 | ROS 与 Isaac 中的直接 sim-to-real 管线 | 适配第三方硬件需定制转接件 |
| 遥操作数据采集 | 仅记录动作轨迹,缺少触觉交互动态数据 | CraftNet 真实世界遥操作,带主动触觉调节 | 加速灵巧操作的后训练收敛 | 遥操作硬件扩容门槛高,且依赖操作员技能 |
工作流梳理与性能优势基于 CraftNet 技术白皮书及机器人实操演示分析。
[CE006, CE007, CE011]5.4 运行依赖、可靠性与工程风险
尽管触觉优先的自主架构展现出可观的技术创新,但 Sharpa 面临着值得严格尽调的重大技术、运营及供应链依赖。高分辨率触觉阵列依赖专用弹性体薄膜与微传感器阵列,在多工班连续工业作业环境下天生容易受到物理摩擦、剪切疲劳及热漂移的影响。System 0 传感器的老化退化或校准漂移会直接削弱精细动作的力度调节,造成掉落物品或接触压力过大的作业风险。此外,尽管公司在 AI 与商用机器人领域积极招募工程专业人才,但规模化实地部署仍需要稳健的持续遥测、现场运维体系以及确定性安全机制,以防止协作服务环境中发生人员伤害。[CE012, CE013]
| 分层或组件 | 架构职责 | 核心依赖 | 运营或技术风险 |
|---|---|---|---|
| System 2(推理脑) | 1 Hz 高层任务拆解与环境推理 | 开源视觉语言模型(VLM)底座 | 边缘长尾任务规划中的推理延迟与模型幻觉 |
| System 1(运动脑) | 10 Hz 基础模型,生成粗轨迹与接触前手部姿态 | 高质量多模态操作数据集 | 接触物体前的粗轨迹规划误差 |
| System 0(交互脑) | 100 Hz 高频触觉反馈与精细运动控制 | 低延迟总线与动态触觉阵列传感器 | 高机械负载下的传感器漂移与信号衰减 |
| Dynamic Tactile Array (DTA) | 指尖触觉传感,具备 1000+ 像素与 5 mN 灵敏度 | 定制弹性体薄膜与微传感器制造 | 磨损老化导致指尖需频繁更换 |
运行架构组件与风险分类源自 CraftNet 系统公开披露及具身智能硬件要求。
[CE003, CE004, CE005, CE013]5.5 深度图表
06客户分析
6.1 客户群细分与买方画像
Sharpa 的早期商业化战略面向四大独立买方与用户客群:商业零售与生活服务运营商、智慧城市与市政基础设施机构、高校及科研院所机器人实验室,以及 AI 与开发者平台伙伴。在零售与餐饮领域,加盟商及门店运营负责人寻求自动化物理操作方案,以期在不改动标准吧台设备的前提下缓解结构性服务用工荒。在公共基础设施领域,JTC Corporation 与海事机构等政府主体评估用于公共园区维护与集装箱港口作业的自主移动机器人车队。高校实验室构成了立即可转化的评估客群,通过采购硬件测试新型触觉灵巧强化学习框架。最后,具身智能开发者与仿真平台厂商将 Wave 灵巧手作为基座模型研发与 Sim-to-Real 迁移的参考基准。 [CU001, CU002, CU003, CU004, CU005, CU006, CU007]
| 客群细分 | 买方 / 使用方 / 付费方 | 用例 | 规模 / 部署情况 | 战略价值 | 尽调缺口 |
|---|---|---|---|---|---|
| 商业服务与零售 | 门店运营总监 / 加盟商 | 实时餐饮制作(冰淇淋出餐) | DQ 中国上海门店试点部署 | 面向消费者的类人操作具备极高公众关注度 | 合同定价、单店经济模型与拓店铺设速度未经证实 |
| 智慧城市与公共基础设施 | 政府机构(JTC、A*Star)/ 公共交通机构 | 多场景园区服务与港口集装箱自动化装卸 | PDD 园区级试点车队与港口转锁试验 | 标杆示范场景,并获政府部门支持 | 依赖补贴型试验场,商业化市场可行性存疑 |
| 高校与科研实验室 | 实验室负责人(PI)/ 机器人学科教研人员(NTU、MIT) | 遥操作研究、齿轮啮合与强化学习(RL)灵巧操作 | 多家高校实验室评估 Wave 灵巧手与 DexTeleop-0 | 验证触觉传感算法,占领从业者心智 | 高校授权折扣大,软件收入获取能力有限 |
| 机器人生态与 AI 开发者 | 具身智能研究团队(NVIDIA、Google、MANUS) | 基础模型训练、sim-to-real 迁移、动作捕捉 | 测试样机已接入 Isaac Lab 与 MANUS 手套 | 树立 Sharpa Wave 作为具身智能模型标杆硬件的地位 | 仅证实生态集成,未见多台采购的持续复购订单 |
客群分类基于截至 2026 年 Q3 的公开商业试点、政府机构协议及高校实验室论文。
[CU001, CU002, CU003, CU004, CU005, CU006, CU007]6.2 采纳路径、性能基准与部署规模
Sharpa 的落地推进路径在 2026 年完成了从孤立的机械手硬件评估向一体化多机器人园区试点的跨越。公司于 2026 年初推出自主研发的 North 人形机器人,迅速在多元运营场景中获得高曝光度部署。在高校与工业合作伙伴环境中开展的基准测试表明,该方案较传统平口夹爪具备显著的作业优势。在与 NVIDIA 就 ADEPT 框架开展的联合研究中,基于 Sharpa Wave 的触觉反馈策略在接触密集型装配测试中成功率达 80%,高于纯视觉基线的 30%;轴孔插入耗时也从平行夹爪的 20 至 70 秒缩短至 5 至 10 秒。这些表现印证了 22 个主动自由度的机械优势,不过商用车队的使用率指标仍高度保密。 [CU002, CU005, CU008, CU009, CU010]
| 指标 | 数值 | 日期 | 来源 | 置信度 | 战略启示 | 缺失分母 |
|---|---|---|---|---|---|---|
| PDD 车队部署 | 覆盖零售与餐饮的多机器人车队 | 2026-04-28 | The Business Times | 高 | 首个区域级多环境试验场 | 实际部署机器人总台数与有效运行时长 |
| 齿轮啮合成功率 | 遥操作成功率 97% | 2026-07-14 | NTU / Sharpa Customer Stories | 高 | 展现出面向工业装配的极高触觉灵巧度 | 总试验周期次数与操作员技能差异 |
| 触觉策略成功率 | 80%(8/10 次试验),纯视觉仅 30% | 2026-08-18 | NVIDIA Robotics / LinkedIn | 中 | 体现触觉模态在接触密集型操作中的优势 | 对未知物理几何结构的泛化能力 |
| 轴孔插装速度 | 5-10 秒(平行夹爪需 20-70 秒) | 2026-08-18 | NVIDIA ADEPT Research | 中 | 通过手内重定位展现出显著的吞吐量提升 | 连续作业下的硬件长期磨损情况 |
采纳轨迹反映的是受控环境下的试点测试与科研评估;实际投产的客户车队总数目前尚未见报道。
[CU002, CU005, CU008, CU009]展现从数字仿真与遥操作起步,逐步迈向现场试点和企业级机队规模化扩展的采纳路径。
示意学术研究、开发者及企业客户接触场景中的典型递进演变过程。
[CU007, CU010, CU012, CU014]6.3 实名客户实证与试点部署
Sharpa 已在企业级零售、市政基础设施和海运物流等领域建立了实名试点项目。Dairy Queen 中国与 Sharpa 合作,在上海一家零售门店部署 North 人形机器人,与后厨员工协同实时制作软冰淇淋。在新加坡,JTC Corporation 与 Sharpa 签约,在占地 50 公顷的 Punggol Digital District 各处的零售、餐饮及物流环境中部署 North 机器人车队,并由园区内的企业创新实验室提供支持。在重工业物流领域,Sharpa 与 A*Star 资讯通信研究院合作开展集装箱自动化装卸项目,聚焦集装箱货运码头中高风险的转锁安装与拆卸作业。与此同时,包括 Nanyang Technological University 和 MIT 在内的高校已验证了利用 Wave 硬件进行复杂工业齿轮啮合以及零样本工具操作。 [CU001, CU002, CU003, CU004, CU005, CU006, CU007]
| 客户 / 合作伙伴 | 客群细分 | 部署情况 / 用例 | 量产 vs 试点 | 成效 | 局限 |
|---|---|---|---|---|---|
| Dairy Queen (DQ) China | 商业零售 / 餐饮 | 在上海真实门店使用 North 人形机器人制作冰淇淋 | 试点 | 使用标准厨房设备成功实现自主备餐服务 | 仅为单店演示;单台回本周期未披露 |
| JTC Corporation | 公共基础设施 / 产业园区 | 在 PDD 开展覆盖零售、餐饮及物流的区域级多机器人部署 | 试点 | 接入 PDD Open Digital Platform 与现场创新实验室 | 部署进度取决于 2026 年园区完工时间 |
| A*Star 资讯通信研究院 | 政府科研 / 海运海事 | 联合研发用于集装箱转锁插拆的港口机器人 | 试点 | 验证港口作业中重载具身智能操作的可行性 | 商用前测试,需通过广泛的安全认证 |
| Nanyang Technological University (NTU) | 学术科研 | 将 DexTeleop-0 遥操作框架应用于精密机械齿轮啮合 | 试点 / 科研 | 复杂接触密集型装配任务成功率达 97% | 属于实验室试验台,而非自动化工业厂房产线 |
| MANUS Meta | 技术合作伙伴 | 使用传感手套在 NVIDIA Isaac Lab 中遥操作 22-DoF Sharpa 灵巧手 | 试点 / 科研 | 力反馈手套与触觉灵巧手无缝集成 | 属于开发者工具评估,并非企业终端客户采纳 |
所有实名客户均属于公开评估试点、联合研发合作或高校部署;商业化经常性收入合同仍未获得证实。
[CU001, CU002, CU003, CU004, CU005, CU007]6.4 留存动态、商业转化与客户集中度风险
尽管试点推进势头迅猛,Sharpa 仍处于未产生商业收入的阶段,面临严峻的客户集中度与商业持续性挑战。现有的企业客户部署主要体现为试点示范与受补贴的公共基础设施项目,而非经常性企业软件或硬件订阅合同。由于预期硬件将走向商品化,加之将通用人形机器人接入老旧棕地仓库管理系统的摩擦尚未消除,物流与工业企业高管对于签订多年期合同表现出犹豫。此外,Sharpa 的商业曝光高度集中于新加坡政府支持的生态和单店品牌展示;其长期净收入留存率(NRR)、硬件维护的单机经济性及跨门店扩张速度均尚未得到验证。 [CU011, CU012, CU013, CU014]
6.5 深度图表
07风险评估
7.1 核心风险画像
Sharpa 的风险评级为「高」:公司正把尚处早期的全栈机器人平台从演示推向顾客实际所在的工作场所,却尚未披露评估安全、可靠性和服务经济性所需的常规运营数据。独立报道指出,其人工干预率、订单周期时间、正常运行时间、平均无故障时间和整班经济性仍属未知。高密度触觉硬件也带来耐用性挑战;加之高达 $3.3 billion 的投后估值与零收入现状,执行层面的任何闪失都将带来实质性的财务冲击。因此,风险热图将现场部署可靠性与人身安全置于融资风险之上,网络安全与供应商依赖紧随其后。[CR001, CR003, CR006, CR009, CR010, CR016]
基于证据权重的优先级评估表明:实际部署中的可靠性、安全性与网络安全风险居于首位,紧迫度相对较低的再融资与技术探索风险次之。
序数分值属于分析师基于所引证据与公开证据缺口作出的定性研判,并非实测概率。
[CR001, CR003, CR005, CR006, CR009, CR010, CR011, CR016]7.2 监管、法律与安全风险敞口
现有公开证据无法证实公司已取得机械或作业场所安全认证、食品加工许可、隐私合规管控、事故记录、诉讼情况,亦未见保险与产品责任保障框架。这反映出证据缺口,虽非违规实证,但对投资决策至关重要,因为 North 被描述为在真实顾客、食材、动力关节、摄像头及联网 AI 系统环绕的环境下作业。在投资落实前,法律顾问与独立安全工程师应梳理各运营司法管辖区的法规要求,调取安全论证报告与事故日志,并核实紧急制动、安全状态保持、卫生、隐私及版本更新管控措施是否已在 DQ 实际部署中经过充分验证。[CR002, CR004, CR005, CR015, CR016]
| 风险领域 | 管辖区 / 场景 | 公开现状 | 发生概率 | 严重程度 | 缓解措施成熟度 | 残留风险敞口 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| 机械与工作场所安全 | 上海餐饮网点及未来拟拓展市场 | 所分配来源中未见认证或安全论证材料 | 中 | 严重 | 未见公开展示 | 高 | 调取适用认证、危险源分析、事故记录以及独立验收测试结果 |
| 食品卫生与产品责任 | DQ 中国商业化工作流 | 公司声称已实现面向真实顾客的备餐;卫生许可与责任划分未见披露 | 中 | 高 | 未见公开展示 | 高 | 审查卫生许可、消毒验证、污染控制、保险及客户合同赔偿条款 |
| 网络安全、隐私与软件更新 | 联网机器人及面向消费者的作业场所 | 所分配证据中未见针对 Sharpa 的安全架构或事故记录 | 中 | 严重 | 仅讨论了通用平台层管控措施 | 高 | 委托第三方进行渗透测试;核实签名更新、日志记录、数据流向、访问控制、环境隔离及安全回滚机制 |
| 跨境供应链与技术管制 | 新加坡公司,采用跨国软硬件合作伙伴 | 据报道其采用了跨境参考技术栈;合同约定与出口管制合规防线未见披露 | 中 | 高 | 未见合作伙伴多元化证据 | 中高 | 排查零部件产地、制裁与出口管制敞口、替代供应商、代码托管权利及终止合作交接支持 |
严重程度排序基于所分配来源的分析师评估;缺失仅代表已审阅材料中未见证据,并不代表已证实违规。
[CR002, CR004, CR005, CR015, CR016]7.3 运营、质量与安全风险
North 号称执行包含 50 至 60 个步骤的工作流,这为误差累积提供了多处隐患,而目前尚未发现公开的多日运行可靠性数据集。Sharpa 触觉优先的架构固然有助于改善操作,但指尖磨损、校准漂移、清洁维护、抓取失误与执行器故障均可能将技术优势转化为繁重的运维负担。此外,公司依赖软硬件混合的技术栈,一旦发生网络安全漏洞便可能演化为物理安全事故。最低限度的风险对冲包应包含:带日期的实地性能数据集、零部件寿命测试、故障树与危险源分析、签名更新与网络隔离控制、版本回滚机制、备件保障承诺,以及在峰值工况下经独立见证的验收测试。[CR002, CR003, CR005, CR006, CR011, CR014]
| 失效模式 | 发生概率 | 严重程度 | 证据信号 | 缓解措施成熟度 | 残留风险敞口 | 未决缺口 / 触发条件 |
|---|---|---|---|---|---|---|
| 长流程工作流失效或隐形人工干预 | 高 | 严重 | 公司声称具备 50-60 步的长流程,但独立的可靠性指标仍未披露 | 早期实地部署 | 高 | 要求调取数周维度的自主完成率、人工干预、节拍时间、正常运行时间及安全停机数据 |
| 在员工或顾客附近发生危险动作或物体抓放失误 | 中 | 严重 | 人形餐饮后厨作业将具身智能置于真实生产环境 | 未发现公开的安全论证材料 | 高 | 一旦发生任何重复的安全重大故障即叫停部署,直至独立完成根因闭环 |
| 触觉手磨损、零点漂移或清洗损伤 | 高 | 高 | 独立分析指出磨损、校准、抗冲击及可维护性属于尚未解决的商业化问题 | 未发现寿命测试数据 | 高 | 调取工作周期寿命、防尘防水与清洗耐受度、失效分布、更换耗时及保修条款 |
| 网络入侵引发危险动作或数据泄露 | 中 | 严重 | 软硬件集成架构扩大了攻击面,将软件安全风险转嫁为物理现实安全 | 仅讨论了平台级防护;Sharpa 具体落地情况未知 | 高 | 要求开展架构审查,提供红队演练成果、签名更新证据、离线模式及应急响应演练记录 |
| 执行器或核心零部件发生批次性质量异常 | 中 | 高 | 据报道有供应商出货超 10,000 个模块,但未披露客户分配、实地故障率或标称寿命数据 | 据报道供应商产线已投产 | 中高 | 引入第二供方并强化来料检验、全流程追溯、备件保障与实地退返阈值 |
发生概率与残留风险敞口仅具方向性参考意义,因所分配材料中未包含经审计的实地数据集、零部件寿命表或事故登记册。
[CR002, CR003, CR005, CR006, CR011, CR014]7.4 伙伴合作、融资及执行依赖
Sharpa 的商业实证高度依赖公司单方描述的 DQ 餐饮工作流,而关键零部件与平台公信力则受制于 Hesai、NVIDIA 和 Unitree 等合作方。Hesai 虽公布了产能规模,但并未披露供给 Sharpa 的配额、采购定价、现场故障率或作业循环寿命,供应链持续性与质量风险敞口依然未决。报道的融资规模提供了充裕的资金实力,但优先股证券、数十亿美元的高估值、未见公开收入以及未披露的烧钱率或毛利率,带来了融资模型层面的风险。因此,投资尽调应着重检验备选供应商储备、合同权益条款、客户集中度、资金跑道期、清算优先权以及维保已部署机器人所需的运维预算。[CR008, CR009, CR010, CR011, CR012, CR015]
| 依赖项 | 合作方 / 角色 | 集中度信号 | 失效情景 | 严重程度 | 缓解措施 | 残留风险敞口 |
|---|---|---|---|---|---|---|
| 标杆商业化验证 | DQ 中国 / 部署场地 | 公开商业实证高度集中于已披露的单一工作流 | 试点效果不及预期、暂停或未能转化为持续复购部署 | 严重 | 设立多门店履约里程碑,锁定客户背书标杆与优先续约权 | 高 |
| 执行器组件 | Hesai / 供应给 Sharpa 的模块 | 供应商虽披露出货量,但未见针对 Sharpa 的具体配额或现场质量数据 | 良率、耐用度、定价或交付问题将制约生产与售后服务 | 高 | 对关键模块实施双轨采购、维持安全库存,并在合同中锁定质量与供货连续性条款 | 中高 |
| 参考平台与生态 | NVIDIA 与 Unitree / 算力、软件及本体生态 | 产品公信力与系统集成依赖上述外部平台 | 技术路线图、安全性、地缘政治或可用性变化将迫使产品重新设计 | 高 | 保持接口可移植性、储备经充分验证的备选方案,并保留变更控制权 | 中高 |
| 资本与估值支撑 | 大型战略投资者阵营 / 优先股融资 | 据报道完成一轮 $669M 融资,投后估值 $3.3B | 优先股清算负担过重或未达里程碑,将使下一轮融资面临惩罚性条款 | 高 | 审查股权结构表与优先权条款;将资金支出与付费使用率、可靠性及毛利率里程碑挂钩 | 高 |
本登记表仅涵盖已有证据支持的依赖项;合同条款、客户收入集中度以及供应商份额均未公开。
[CR008, CR009, CR010, CR011, CR012, CR015]7.5 缓解措施、监测指标与投资逻辑破坏触发项
投资路径应由里程碑把关,而不是靠叙事推动。持续跟踪自主完成率、每单人工干预次数、安全停机频率、周期时间、正常运行时间、部件更换率、单台服务成本、安全事故、客户续约,以及现金消耗是否符合计划。若安全关键故障反复出现且根因尚未查明,若运行依赖未披露的持续人工控制,或公司无法提供辖区合规与网络安全证据,就应暂停部署。若长期现场可靠性达不到客户对人力替代和吞吐量的要求、关键供应商无法替换,或资本需求上升却没有相应的付费使用量,投资逻辑就会失效。[CR003, CR004, CR005, CR006, CR011, CR016]
7.6 深度图表
08估值分析
8.1 投资论点与反向论点
Sharpa 的核心投资论点立足于其开创性的触觉优先物理 AI 架构,以量产的 22 自由度 SharpaWave 灵巧手与自研 CraftNet 基础模型为支柱。不同于纯视觉机器人平台,Sharpa 的高分辨率触觉传感器具备毫牛级压力灵敏度,使其无需改造场地,也能在非结构化商业环境中完成接触密集型操作。Alibaba、Meituan、Tencent 及 JD.com 等一线产业资本的加持,为其开拓物流与零售分发渠道铺平了道路。相反,反向论点强调 Sharpa 尚未产生商业化收入,且身处资本密集、投资回收期长的硬件赛道。Dairy Queen 的落地展现了出色的演示效果,但长期维护成本、实地干预率以及客户规模化付费意愿均未经验证。[CV001, CV004, CV005, CV006]
| 视角 | 核心支柱 | 核心论点 | 何种情况会改变观点 |
|---|---|---|---|
| 投资逻辑 | 触觉硬件护城河 | 具备 22-DoF 自由度与毫牛级灵敏度的 SharpaWave 灵巧手,能够完成复杂的接触式操作 | 第三方验证失败,或 OEM 厂商迅速将触觉传感技术通用商品化 |
| 投资逻辑 | 战略生态协同 | 获得 Alibaba、Tencent、Meituan 与 JD.com 的支持,带来直接的商业落地渠道 | 战略合作方转而扶持内部竞争性机器人项目 |
| 反向逻辑 | 未产生商业收入风险 | 尽管估值达到 $3.3B,公司披露的商业收入仍为零 | Dairy Queen 及零售试点成功转化为长期商业化经常性合同 |
| 反向逻辑 | 现场可靠性差距 | 55 步零售工作流缺少公开的 MTBF 及人工干预率数据透明度 | 经审计的多月实际运营数据证明自主任务完成率 >99% |
对比触觉手技术领先的商业上行空间,与硬件工程落地及客户集中度风险。
[CV004, CV005, CV006, CV007]8.2 融资背景与对标参照
2026 年 8 月,Sharpa 披露累计融资超 RMB 4.5 billion(约合 $668M 至 $669M),完成了一轮可转换优先股融资,投后估值达 RMB 22 billion($3.3 billion)。本轮交易吸引了 HongShan、Qiming Venture Partners 以及 Meituan DragonBall 等头部机构与产业投资方。在 $3.3 billion 投后估值下,Sharpa 的一级市场估值据报已超越其管理团队创立的激光雷达上市公司 Hesai Technology。相较于 Figure AI 等全球具身智能同行及 LinkerBot 等独立零部件专家,Sharpa 凭借中国产业资本背书和硬件壁垒预期享有溢价倍数,但公开金融市场的可比公司交易估值倍数显著低于此类水平。[CV001, CV002, CV003, CV007, CV008]
| 情景 | 2028 年目标指标 | 隐含估值 | 关键假设与下行触发条件 | 概率信号 |
|---|---|---|---|---|
| 乐观情景 | $150M+ ARR / 部署 5,000+ 只灵巧手 | $5.0B - $6.5B | 在零售和工业领域被 OEM 厂商广泛采用;成功切入家庭场景 | 25% |
| 基准情景 | $40M - $80M ARR / 部署 1,500 只灵巧手 | $2.5B - $3.5B | 在快餐(QSR)与物流领域实现稳定商业化应用;面临温和的利润率承压 | 50% |
| 悲观情景 | <$15M ARR / 试点项目遭取消 | $0.8B - $1.5B | 硬件故障率高企、零售推广缓慢、遭遇折价重组融资 | 25% |
估值倍数参考机器人 B 轮融资与 Hesai 公开市场可比标的;发生概率反映公司仍处在无收入的落地阶段。
[CV008, CV009, CV010, CV011]8.3 估值情景推演与下行情景触发项
考虑到早期机器人商业化落地具有二元对立的特征,情景分析显示出潜在结果的巨大离散度。在乐观情景(Bull case)下,Sharpa 将灵巧手推广至各 OEM 合作伙伴,深化在快餐连锁的零售渗透,并在 2028 年前实现商业收入,支撑以 $5.0B 估值通过 IPO 退出。在基准情景(Base case)下,Sharpa 在高价值细分任务中实现小范围商业化,但面临硬件毛利挤压和部署周期拉长,估值维持在 $3.0B 平盘。在悲观情景(Bear case)下,技术可靠性遭遇瓶颈、客户试点流失或面临低成本零部件厂商竞争,将引致下调轮(down-round)风险或重组清算,估值跌破 $1.2B。核心下行触发因素包括门店运营期间居高不下的人工干预率以及商业收入转化停滞。[CV009, CV010, CV011, CV012]
串联资本规模、触觉硬件验证、商业试点推进以及无收入阶段估值重压,最终推导出投资立场的决策链路。
| 节点 / 连接 | 节点 / 起点 | 终点 | 说明 |
|---|---|---|---|
| 节点 1 | 资本规模(融资 $669M,投后估值 $3.3B) [financing-scale] | ||
| 节点 2 | 硬件护城河(22-DoF 自由度 SharpaWave 灵巧手量产) [hardware-moat] | ||
| 节点 3 | 商业试点(55 步 Dairy Queen 制作工序) [commercial-pilots] | ||
| 节点 4 | 零收入估值重压(截至目前已披露商业收入为零) [revenue-overhang] | ||
| 节点 5 | 尽调综合研判(技术实力强 vs 定价估值风险高) [synthesis] | ||
| 节点 6 | 最终立场:继续研究 / 估值偏高 [decision-outcome] | ||
| 连接 1 | 资本规模(融资 $669M,投后估值 $3.3B) [financing-scale] | 尽调综合研判(技术实力强 vs 定价估值风险高) [synthesis] | |
| 连接 2 | 硬件护城河(22-DoF 自由度 SharpaWave 灵巧手量产) [hardware-moat] | 尽调综合研判(技术实力强 vs 定价估值风险高) [synthesis] | |
| 连接 3 | 商业试点(55 步 Dairy Queen 制作工序) [commercial-pilots] | 尽调综合研判(技术实力强 vs 定价估值风险高) [synthesis] | |
| 连接 4 | 零收入估值重压(截至目前已披露商业收入为零) [revenue-overhang] | 尽调综合研判(技术实力强 vs 定价估值风险高) [synthesis] | |
| 连接 5 | 尽调综合研判(技术实力强 vs 定价估值风险高) [synthesis] | 最终立场:继续研究 / 估值偏高 [decision-outcome] |
决策逻辑路径展示的是定性把关要件,而非定量评分权重。
[CV012, CV013, CV014]8.4 投资建议与尽调诉求
基于现有公开证据,我们给予 Sharpa「继续研究」的投资建议,评定为中等确信度、高风险评级,估值立场偏高。$3.3 billion 的估值已预先计入了在产生商业收入或确立单机经济效益之前的完美执行预期。评估老股转让或后续扩张轮次的机构投资人在出资前,必须索取严格的技术与运营审计数据。重点尽调要求包括:核验上海 Dairy Queen 部署期间的任务完成率与平均无故障时间(MTBF)、审计 SharpaWave 灵巧手的 BOM 成本演变轨迹,以及核查 2026 年 8 月可转换优先股融资的清算优先权条款。[CV001, CV005, CV013, CV014]
8.5 深度图表
免责声明
本报告仅基于公开渠道获取的信息,不构成任何投资建议。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Sharpa is a private embodied AI robotics company headquartered in Singapore with R&D operations in Shanghai and Mountain View, founded in late 2024. | 中 | SO001, SO002 |
| CO002 | Sharpa raised $669 million in convertible preferred equity on August 28, 2026, reaching a post-money valuation of $3.3 billion (RMB 22 billion). | 高 | SO001, SO008 |
| CO003 | An undated Embodied AI Frontier profile describes Sharpa as not yet generating revenue; it does not establish an audited current revenue run-rate or the billing status of pilot deployments. | 中 | SO006 |
| CO004 | Public staffing indicators include 74 LinkedIn-associated members and press reporting of over 100 personnel; these differently defined observations do not establish a verified current headcount range. | 中 | SO002, SO007 |
| CO005 | Sharpa actively recruits for open roles across AI engineering, commercial operations, and general business execution on its official career portal. | 高 | SO005, SO007 |
| CO006 | Sharpa was co-founded by Hesai Technology co-founders Li Yifan (David Li), Xiang Shaoqing, and Sun Kai, operating without binding equity ties to Hesai. | 中 | SO001, SO002 |
| CO007 | Independent reports present conflicting views on executive governance, with Seedtable identifying David Li as CEO while Embodied AI Frontier notes co-founders hold no formal executive roles. | 中 | SO001, SO006 |
| CO008 | Sharpa's investor syndicate includes strategic tech giants Alibaba, Tencent, Meituan, and JD.com alongside financial investors HongShan and Luminous Ventures. | 高 | SO004, SO008 |
| CO009 | Sharpa commenced mass production of its flagship SharpaWave 22-DoF dexterous tactile robotic hand in October 2025, primarily serving research laboratories. | 中 | SO002, SO004 |
| CO010 | SharpaWave received the CES 2026 Innovation Award for advancing tactile intelligence and human-scale 22-DoF dexterous manipulation. | 中 | SO003, SO004 |
| CO011 | Sharpa deployed its North humanoid robot at a Shanghai Dairy Queen store on August 29, 2026, executing a 55-step Blizzard preparation routine in customer-facing operations. | 中 | SO006, SO009 |
| CO012 | Independent analysis highlights that Sharpa's commercial kitchen deployment lacks audited metrics on intervention frequency, task success rates, and shift-level MTBF. | 中 | SO009 |
| CO013 | Sharpa relies heavily on Hesai Technology as an actuation module supplier for its dexterous hands and full-body joints, linking supply-chain ramp to Hesai's production line. | 中 | SO009 |
| CO014 | Physix Frontier reports $670 million in cumulative funding disclosure as of August 2026, explicitly without specifying a round or a dated seed and development-stage breakdown. | 中 | SO003 |
| CM001 | Sharpa develops full-stack Physical AI solutions spanning anthropomorphic tactile hands, full-body humanoid robots, and embodied foundation models. | 高 | SM001, SM005 |
| CM002 | Sharpa's Wave robotic hand features 22 active degrees of freedom and high-resolution tactile sensing for contact-rich manipulation tasks. | 高 | SM001, SM004 |
| CM003 | Sharpa positions its North humanoid robot to automate shifts in retail, hospitality, food service, and manufacturing without requiring facility retrofits. | 高 | SM001, SM002 |
| CM004 | Sharpa focuses its manipulation architecture on solving the last 1% of task completion unassisted to avoid human intervention. | 中 | SM002 |
| CM005 | The addressable market for physical automation spans multi-trillion-dollar global labor spend across repetitive service and industrial workflows. | 中 | SM001, SM002 |
| CM006 | Sharpa reached a valuation of RMB 22 billion following financing rounds backed by strategic internet giants including Alibaba, Meituan, Tencent, and JD.com. | 中 | SM003, SM008 |
| CM007 | Robotics component supplier Hesai delivered more than 10,000 robotic actuation modules by Q2 2026 as humanoid supply chains began series production. | 中 | SM006 |
| CM008 | Sharpa segments its market into embodied AI researchers, robot manufacturers integrating Wave, and commercial enterprises deploying North. | 高 | SM001, SM005 |
| CM009 | Teleoperation integrations using MANUS data gloves and NVIDIA Isaac Lab target embodied AI research customers collecting demonstration datasets. | 高 | SM001, SM009 |
| CM010 | Sharpa deployed its North humanoid robot at a Dairy Queen store in Shanghai to execute a 55-step Blizzard preparation workflow. | 中 | SM005, SM006 |
| CM011 | Commercial enterprise adoption of humanoid labor in food service requires handling deformable objects and maintaining task state through visual and tactile feedback. | 中 | SM001, SM006 |
| CM012 | Independent datasets verifying task success rates, human intervention frequencies, cycle times, and MTBF for Sharpa's commercial deployments have not been published. | 中 | SM006 |
| CM013 | Commercial store operations evaluate humanoid automation on shift-level unit economics and labor offset rather than demonstration capability. | 中 | SM006 |
| CM014 | Initial commercial deployment commitments in service robotics are concentrated in single-store flagship trials rather than multi-store production fleets. | 中 | SM005, SM006 |
| CP001 | Sharpa competes directly against Chinese dexterous hand manufacturers Fourier and Inspire Robots in the multi-DoF robotic manipulation segment. | 中 | SP003 |
| CP002 | Sharpa positions the SharpaWave hand as a technical benchmark with 22 active degrees of freedom and a proprietary Dynamic Tactile Array. | 中 | SP003, SP006 |
| CP003 | Parallel-jaw grippers represent the status-quo alternative in automated manipulation, requiring 20 to 70 seconds per peg-insertion trial compared to 5 to 10 seconds for the Sharpa Wave. | 中 | SP002 |
| CP004 | Sharpa has raised substantial venture funding including a disclosed $669 million financing round in August 2026 at a $3,300 million valuation. | 高 | SP007, SP005 |
| CP005 | Hesai supplies robotic actuation modules to Sharpa while simultaneously serving over 50 embodied AI robotics customers including peers Unitree, Galbot, and Galaxea. | 中 | SP004 |
| CP006 | Hardware partner integrations such as MANUS teleoperation gloves in NVIDIA Isaac Lab allow developers to build manipulation policies across compatible robotic platforms. | 高 | SP008, SP010 |
| CP007 | Independent validation of Sharpa's commercial deployments remains unproven regarding full-shift MTBF, intervention frequency, and task economics. | 中 | SP004 |
| CP008 | Sharpa actively hires across engineering, commercial, and operational functions to scale its hardware and software operations. | 中 | SP009 |
| CP009 | Competitors such as Unitree and general-purpose robotics OEMs maintain multi-vendor component procurement strategies to avoid single-supplier lock-in. | 中 | SP004 |
| CP010 | Tactile simulation policies on hardware achieve an 80% success rate compared to 30% for vision-only policies in precision contact tasks. | 中 | SP002 |
| CP011 | Sharpa's North humanoid targets commercial workflow tasks including retail food preparation, testing end-to-end multi-step manipulation in human environments. | 中 | SP003, SP004 |
| CP012 | Strategic positioning in high-degree-of-freedom manipulation requires balancing expensive tactile sensor arrays against cheaper incumbent parallel grippers. | 中 | SP002, SP003 |
| CP013 | Component suppliers expanding from dexterous hand actuators toward full-body humanoid joints may commoditize mechanical subsystems across competing robot builders. | 中 | SP004 |
| CP014 | Sharpa has issued official press communications highlighting its product innovations and technical recognition within international robotics forums. | 高 | SP001, SP006 |
| CI001 | Sharpa was valued at $3.3 billion following an August 2026 financing round that raised $669 million in convertible preferred stock. | 高 | SI001, SI002 |
| CI002 | Sharpa has not yet generated commercial revenue and remains an early-stage pre-revenue robotics developer. | 中 | SI004 |
| CI003 | Multiples reports $784 million in cumulative capital across three rounds through August 2026. Its $667 million latest-round entry differs from MarketScreener's $669 million round; funding scope, timing, and currency-conversion basis have not been independently reconciled. | 中 | SI001, SI002 |
| CI004 | Sharpa's investor syndicate includes Alibaba Group, Meituan, Tencent, JD.com, and Qiming Venture Partners. | 高 | SI001, SI002, SI007 |
| CI005 | Sharpa launched a commercial kitchen deployment on August 29, 2026 at a Dairy Queen store in Shanghai to prepare Blizzard products. | 中 | SI003 |
| CI006 | The Dairy Queen preparation workflow requires a 55-step manipulation sequence executing in an un-retrofitted human kitchen environment. | 中 | SI003 |
| CI007 | Task success rates, human intervention frequencies, and cycle times for Sharpa's restaurant operations have not been publicly disclosed. | 中 | SI003 |
| CI008 | The long-term shift economics, hardware uptime, and mean time between failures for Sharpa's food service robots remain unproven. | 中 | SI003 |
| CI009 | Sharpa relies on Hesai as an external supplier of robotic actuation modules for its dexterous manipulation hardware. | 中 | SI003 |
| CI010 | Sharpa employs between 51 and 200 staff members headquartered in Singapore. | 中 | SI006 |
| CI011 | Sharpa's careers page lists three AI engineering openings, two commercial openings, and one internship. | 中 | SI005 |
| CI012 | Sharpa's products include the 22-degree-of-freedom SharpaWave tactile hand and the North humanoid platform. | 中 | SI007, SI008 |
| CI013 | Sharpa was evaluated against public robotics peers including UBtech Robotics, Symbotic, and Doosan Robotics. | 中 | SI001 |
| CI014 | Sharpa's reported valuation depends on future commercialization because its retained independent profile reports no generated revenue. | 中 | SI001, SI004 |
| CI015 | SharpaWave has 22 degrees of freedom and more than 1,000 tactile-sensing pixels per fingertip, indicating a component-dense hardware design. | 中 | SI009 |
| CE001 | Sharpa develops core embodied AI technologies centered on high-performance tactile dexterous hands, full-body humanoids, and multimodal foundation models. | 高 | SE001, SE004 |
| CE002 | The Sharpa Wave robotic hand features 22 active degrees of freedom and high-density tactile sensors on each fingertip. | 中 | SE005, SE006, SE009 |
| CE003 | CraftNet is structured as a hierarchical VTLA architecture incorporating a 1 Hz System 2 reasoning brain, a 10 Hz System 1 motion brain, and a 100 Hz System 0 interaction brain. | 中 | SE004 |
| CE004 | System 0 operates at approximately 100 Hz to process real-time tactile feedback and modulate fine-motor force during the last millimeter of object contact. | 中 | SE004 |
| CE005 | System 2 operates at approximately 1 Hz as an open-source vision-language model responsible for task decomposition, environmental perception, and long-horizon planning. | 中 | SE004 |
| CE006 | Sharpa provides the open-source Sharpa Wave SDK to supply libraries, headers, and runtime components for integrating Wave hardware into custom robotics software stacks. | 中 | SE001, SE002 |
| CE007 | Sharpa maintains public software repositories for device control, including Sharpa Pilot for configuration and firmware updates, and WM-Craftnet for in-hand manipulation research. | 高 | SE001, SE004 |
| CE008 | Sharpa emphasizes a tactile-first manipulation paradigm to overcome the limitations of conventional vision-only models that lack high-frequency force feedback. | 中 | SE004, SE006 |
| CE009 | The Sharpa Wave dexterous hand entered mass production in October 2025 and won a CES 2026 Innovation Award. | 中 | SE006, SE009 |
| CE010 | Sharpa's North full-body humanoid robot incorporates 67 degrees of freedom and has been tested in commercial trials including food service operations. | 中 | SE006 |
| CE011 | CraftNet utilizes real-world teleoperation data and simulation loops in NVIDIA Isaac Lab to refine simulated trajectories with realistic post-contact forces. | 中 | SE001, SE004 |
| CE012 | Sharpa is actively hiring engineering talent across artificial intelligence and commercial operations to scale its robotics development. | 中 | SE008 |
| CE013 | Fine manipulation using high-density tactile arrays faces operational risks from sensor drift, physical wear, and complex real-time calibration requirements. | 中 | SE004, SE006 |
| CE014 | Tactile state vectors from System 0 back-propagate to System 1 to adjust physical intent when fine-grained execution encounters unexpected contact resistance. | 中 | SE004 |
| CU001 | Sharpa has entered strategic partnerships with JTC, A*Star, and Grab to deploy autonomous robots and co-develop physical AI applications in Singapore. | 高 | SU001, SU007 |
| CU002 | Under its partnership with JTC, Sharpa will deploy a fleet of North humanoid robots across retail, food and beverage, and logistics settings at the Punggol Digital District. | 高 | SU001, SU003 |
| CU003 | Sharpa is collaborating with the A*Star Institute for Infocomm Research on automated container handling functions, including twist-lock coning and deconing at ports. | 中 | SU001 |
| CU004 | Dairy Queen China partnered with Sharpa to deploy North humanoid robots for commercial ice cream preparation in a Shanghai store. | 高 | SU003, SU004, SU008 |
| CU005 | Researchers at Nanyang Technological University achieved a 97 percent success rate on precision gear meshing using the DexTeleop-0 framework and the Sharpa Wave hand. | 高 | SU003, SU004 |
| CU006 | SimToolReal researchers demonstrated zero-shot reinforcement learning for general tool manipulation utilizing the Sharpa Wave 22-DoF dexterous hand. | 高 | SU003, SU004 |
| CU007 | MANUS and NVIDIA demonstrated teleoperation of the 22-DoF Sharpa hand inside NVIDIA Isaac Lab using MANUS motion-capture data gloves. | 中 | SU005, SU006 |
| CU008 | NVIDIA robotics research on the ADEPT framework demonstrated that the Sharpa Wave completed peg-insertion tasks in 5 to 10 seconds compared to 20 to 70 seconds for parallel-jaw grippers. | 中 | SU006 |
| CU009 | During hardware trials, tactile feedback policies on the Sharpa Wave achieved an 8 in 10 success rate compared to 3 in 10 for vision-only policies. | 中 | SU006 |
| CU010 | Sharpa's customer journey progresses from simulation-based reinforcement learning in NVIDIA Isaac Lab to teleoperated data collection and real-world pilot deployments. | 中 | SU005, SU006 |
| CU011 | According to embodied AI market analysis, Sharpa has not yet generated commercial revenue and relies on pilot deployments and venture funding. | 中 | SU008 |
| CU012 | General-purpose robotics adoption in logistics faces friction from physical motion training data bottlenecks and the legacy challenge of fragmented islands of automation. | 中 | SU002 |
| CU013 | Rapid expected commoditization of general-purpose robot hardware creates buyer hesitation and delays long-term enterprise procurement commitments. | 中 | SU002 |
| CU014 | Sharpa's active commercial deployments remain concentrated in subsidized regional testbeds such as Singapore's PDD and single-store retail demonstration pilots. | 中 | SU001, SU008 |
| CR001 | Field reliability and safety are the highest residual risks because Sharpa's live-deployment claims are not accompanied by an independent operating dataset. | 中 | SR001, SR009 |
| CR002 | Sharpa says the DQ Blizzard workflow requires 50 to 60 consecutive manipulation steps. | 中 | SR004 |
| CR003 | Independent reporting says task success, intervention count, cycle time, uptime, MTBF, and full-shift economics remain undisclosed. | 中 | SR009 |
| CR004 | The assigned sources do not establish Sharpa-specific safety certifications, food-handling approvals, litigation status, or enforcement history. | 低 | SR001, SR003, SR004, SR009 |
| CR005 | Independent analysis identifies robot cybersecurity as a physical-safety risk because the platform combines networked compute, sensors, updates, and powered actuation. | 中 | SR001 |
| CR006 | Independent analysis identifies tactile-fingertip wear, impacts, cleaning, misgrasps, and calibration drift as unresolved serviceability risks. | 中 | SR001 |
| CR007 | Sharpa's careers page listed three AI engineering roles, two commercial roles, and one operations internship when fetched. | 中 | SR003 |
| CR008 | Sharpa's LinkedIn page describes the company as founded in 2024 with 51 to 200 employees and displays 74 employee profiles. | 中 | SR004 |
| CR009 | MarketScreener reported a $669 million financing using convertible preferred stock at a $3.3 billion post-money valuation. | 中 | SR008 |
| CR010 | The Frontier of Embodied AI reported that Sharpa had not generated revenue as of its profile. | 中 | SR007 |
| CR011 | RoboMorrow reported that Hesai supplies modules to Sharpa but does not publicly break out customer allocation, pricing, field failures, or lifetime under a defined load. | 中 | SR009 |
| CR012 | Independent profiles place Sharpa's reported valuation near RMB 22 billion or approximately $3.3 billion. | 中 | SR005, SR007, SR008 |
| CR013 | A name-only result for Sharpa resolves to unrelated surname records, showing that entity-disambiguation is necessary when researching adverse events. | 中 | SR002 |
| CR014 | Independent market data describes Sharpa as a tactile-intelligence company combining touch, vision, and AI for dexterous manipulation. | 中 | SR006 |
| CR015 | Sharpa's reliance on a multinational body, hand, compute, software, and actuator ecosystem creates cross-border continuity and change-control exposure. | 中 | SR001, SR009 |
| CR016 | Public mitigation maturity is low because the assigned evidence does not include an audited safety case, security assessment, field-reliability series, or supplier-continuity plan. | 低 | SR003, SR004, SR009 |
| CV001 | Sharpa disclosed over 4.5 billion yuan in cumulative financing at a reported 22 billion yuan (approximately $3.3 billion) post-money valuation. | 高 | SV001, SV002, SV003 |
| CV002 | The August 2026 funding transaction was structured through the issuance of convertible preferred stock to investors including Alibaba, Tencent, Meituan, and JD.com. | 高 | SV001, SV003 |
| CV003 | Sharpa's reported private valuation of $3.3 billion reportedly surpasses the public market capitalization of Hesai Technology, the listed company co-founded by Sharpa's leadership. | 中 | SV004, SV007 |
| CV004 | Sharpa's flagship SharpaWave hand features 22 active degrees of freedom and millinewton-level pressure sensitivity, enabling contact-rich manipulation tasks. | 中 | SV008, SV010 |
| CV005 | Sharpa has not yet generated commercial revenue and maintains an unverified unit economic profile across early customer trials. | 中 | SV004, SV005 |
| CV006 | Sharpa's proprietary CraftNet foundation model combines vision, touch, language, and action to execute multi-step manipulation workflows without facility modifications. | 中 | SV001, SV002 |
| CV007 | Operational trials such as the 55-step Dairy Queen restaurant deployment lack published datasets for task success rates, MTBF, or human intervention frequency. | 中 | SV005 |
| CV008 | Sharpa targets initial commercial penetration in structured enterprise environments before attempting entry into consumer household markets from 2028. | 中 | SV002 |
| CV009 | In a bull case operating trajectory, rapid enterprise and OEM scaling could support a $5.0 billion to $6.5 billion public market valuation by 2028. | 中 | SV001, SV003 |
| CV010 | In a base case operating trajectory, steady adoption in specialized QSR tasks supports an estimated valuation range of $2.5 billion to $3.5 billion. | 中 | SV001, SV007 |
| CV011 | In a bear case operating trajectory, pilot churn and hardware reliability deficits could force structured recapitalizations below $1.5 billion. | 中 | SV004, SV005 |
| CV012 | The combination of $669 million in private funding and zero current revenue establishes a high entry multiple that demands immediate commercial contract execution. | 中 | SV003, SV004 |
| CV013 | Sharpa's active technical recruitment across engineering disciplines reflects ongoing product development rather than mature commercial scaling. | 中 | SV006, SV009 |
| CV014 | An investment recommendation of research-more balances premier tactile hardware engineering against stretched valuation multiples and pre-revenue execution risks. | 中 | SV001, SV005 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Seedtable | Sharpa — Funding, Investors & Team | Seedtable | Sharpa is a Singapore-headquartered AI robotics company that builds anthropomorphic hands, humanoid robots and the embodied AI models that make them productive on real-world tasks. Founded in 2024 by Hesai Technology executives, it runs business operations in Mountain View and a manufacturing and R&D centre in Shanghai. |
| SO002 | CIO Bulletin | Hesai Founders Launch Sharpa Robotics Startup | Three Hesai Technology co-founders, CEO Li Yifan, CTO Xiang Shaoqing, and Chief Scientist Sun Kai, have started up a second big entrepreneurship called Sharpa, the second company in their entrepreneurial history. The startup was established near the end of late 2024 and is operational without any binding equity relationship with Hesai and has already grown to over 100 employees. |
| SO003 | Physix Frontier | Sharpa · Profile | Cumulative disclosed (round not specified) (details moved to this field). ... Total fundingUSD 670M ... Data cutoff2026-08-30 |
| SO004 | Robotico Market | Sharpa - Humanoid Robotics Company | Market Data & Profile | Sharpa develops high performance robotic hands with advanced tactile sensing. SharpaWave won CES 2026 Innovation Award with 22 DOF at human scale. |
| SO005 | Sharpa | Sharpa Careers | AI Engineering - Experienced Hires / Fresh Graduate Hires ... 3 Open roles ... Commercial ... 2 Open roles |
| SO006 | Embodied AI Frontier | Sharpa: Investors, Funding | Founded in late 2024 by Hesai co-founders Yifan Li, Shaoqing Xiang, and Kai Sun (none hold executive roles; no equity or operational control ties to Hesai). Global HQ in Singapore with offices in Shanghai and Mountain View. |
| SO007 | Sharpa | LinkedIn | Sharpa is an AI robotics company dedicated to developing ultra-high performance robots and core components ... Headquarters Singapore ... Founded 2024 | |
| SO008 | MarketScreener | Sharpa Pte. Ltd. announced that it has received $669 million in funding from a group of investors | Sharpa Pte. Ltd. announced that it has raised $669 million in a round of funding on August 28, 2026. The transaction involved participation from new investors Alibaba Group Holding Limited, HongShan Capital Advisors Limited, JD.com, Inc., Luminous Ventures, Meiuan, Meituan Dragon Ball, Shenzhen Transsion Holdings Co., Ltd., and Tencent Holdings Limited. |
| SO009 | RoboMorrow | Hesai passes 10,000 robot actuators as lidar supplier expands into Physical AI infrastructure | What is still not proven ... We still do not know the task success rate, number of human interventions, average order cycle time, uptime, MTBF or the economics of a full shift. “Fully autonomous”, “zero-retrofit” and “full shift” are primarily Sharpa and deployment-partner claims. |
| SM001 | Sharpa | Sharpa: AI robotics platform for dexterous manipulation | Deploy North to automate shifts in retail, hospitality, food service, and manufacturing, without retrofitting your space and tools. |
| SM002 | Sharpa | About Sharpa | Sharpa is a full-stack AI robotics company building general-purpose robots that complete shifts unassisted in real-world environments. By solving the manipulation bottleneck - the hardest problem in robotics - our robots can work autonomously without retrofitting the environment. |
| SM003 | Physix Frontier | Sharpa · Profile | Sharpa, founded in late 2024 by Hesai’s team, builds AI robots for dexterous manipulation. Products include the SharpaWave tactile hand and North humanoid. Valued at RMB 22 billion, it secured funding from Alibaba, Meituan, Tencent, and JD.com. |
| SM004 | Robotico Market | Sharpa - Humanoid Robotics Company | Market Data & Profile | Sharpa develops high performance robotic hands with advanced tactile sensing. SharpaWave won CES 2026 Innovation Award with 22 DOF at human scale. |
| SM005 | Embodied AI Frontier | Sharpa: Investors, Funding | Sharpa is a Singapore-based full-stack embodied AI company whose competitive edge is high-resolution tactile dexterous hands, building humanoid robots and embodied AI models upward from the hand, aiming for general-purpose robots that complete complex real-world tasks end-to-end. |
| SM006 | RoboMorrow | Hesai passes 10,000 robot actuators as lidar supplier expands into Physical AI infrastructure | We still do not know the task success rate, number of human interventions, average order cycle time, uptime, MTBF or the economics of a full shift. |
| SM007 | Sharpa | LinkedIn | Sharpa is an AI robotics company dedicated to developing ultra-high performance robots and core components, unlocking the limitless possibilities of future general-purpose robotic applications. | |
| SM008 | MarketScreener | Sharpa Pte. Ltd. announced that it has received $669 million in funding from a group of investors | Sharpa Pte. Ltd. announced that it has raised $669 million in a round of funding on August 28, 2026. |
| SM009 | MANUS | Customer Success Stories - Sharpa | Teleoperating a 22-DoF Sharpa Hand Inside NVIDIA Isaac Lab using MANUS Gloves |
| SP001 | PR Newswire | Sharpa News and Press Releases | News from Sharpa |
| SP002 | Sharpa | LinkedIn | When comparing dexterous hands vs parallel grippers on tasks like peg insertion, the parallel-jaw gripper needs 20-70 seconds per trial while the Wave only needs 5-10 seconds. | |
| SP003 | Embodied AI Frontier | Sharpa: Investors, Funding | The Frontier of Embodied AI | It competes with Chinese hand makers such as Fourier and Inspire Robots, differentiating on tactile precision and full-stack in-house capability. |
| SP004 | RoboMorrow | Hesai passes 10,000 robot actuators as lidar supplier expands into Physical AI infrastructure | The company says it now works with more than 50 embodied-AI companies on lidar, naming Unitree, Robbyant, Galbot, Galaxea and Dexmal among them. |
| SP005 | Physix Frontier | Sharpa · Profile - Physix Frontier | Sharpa, founded in late 2024 by Hesai’s team, builds AI robots for dexterous manipulation. |
| SP006 | Robotico | Sharpa - Humanoid Robotics Company | Market Data & Profile | Sharpa develops high performance robotic hands with advanced tactile sensing. SharpaWave won CES 2026 Innovation Award with 22 DOF at human scale. |
| SP007 | MarketScreener | Sharpa Pte. Ltd. announced that it has received $669 million in funding from a group of investors | The company issued convertible preferred stock in the transaction. |
| SP008 | MANUS | Customer Success Stories - Sharpa - MANUS | Teleoperating a 22-DoF Sharpa Hand Inside NVIDIA Isaac Lab using MANUS Gloves March 9, 2026 |
| SP009 | Sharpa | Sharpa Careers | AI Engineering - Experienced Hires / Fresh Graduate Hires |
| SP010 | The Robot Report | Sharpa Archives - The Robot Report | NVIDIA releases new and updated tools for physical AI developers |
| SI001 | Multiples | Sharpa Valuation, Funding & Investors | Multiples | Sharpa is currently valued at $3.3B, based on its undisclosed-stage round in August 2026. |
| SI002 | MarketScreener | Sharpa Pte. Ltd. announced that it has received $669 million in funding from a group of investors | The transaction is raised at a post-money valuation of $3,300 million. |
| SI003 | RoboMorrow | Hesai passes 10,000 robot actuators as lidar supplier expands into Physical AI infrastructure | We still do not know the task success rate, number of human interventions, average order cycle time, uptime, MTBF or the economics of a full shift. |
| SI004 | The Frontier of Embodied AI | Sharpa: Investors, Funding | The company has not yet generated revenue. |
| SI005 | Sharpa | Sharpa Careers | AI Engineering - Experienced Hires / Fresh Graduate Hires |
| SI006 | Sharpa | LinkedIn | Company size 51-200 employees | |
| SI007 | Physix Frontier | Sharpa · Profile | Valued at RMB 22 billion, it secured funding from Alibaba, Meituan, Tencent, and JD.com. |
| SI008 | Robotico Market | Sharpa - Humanoid Robotics Company | Market Data & Profile | Sharpa develops high performance robotic hands with advanced tactile sensing. |
| SI009 | Brett Adcock | Singapore's Sharpa unveiled SharpaWave, a lifelike robotic hand | Singapore's Sharpa unveiled SharpaWave, a lifelike robotic hand —Features 22 DOF to balance for dexterity and strength —Each fingertip has 1,000+ tactile sensing pixels and 5 mN pressure sensitivity |
| SE001 | GitHub | Sharpa - GitHub | Sharpa is an AI robotics company building ultra-high performance robots and core components to unlock the limitless possibilities of future general-purpose robots. |
| SE002 | DeepWiki | sharpa-robotics/sharpa-wave-sdk | DeepWiki | Loading... |
| SE003 | The Robot Report | Sharpa Archives - The Robot Report | NVIDIA releases new and updated tools for physical AI developers |
| SE004 | Sharpa | a Hierarchical VTLA Model for Fine Manipul | CraftNet is the combination of System 0 and System 1, forming a VTLA (Vision-Tactile-Language-Action) model that translates high-level, multimodal inputs into continuous, fine-grained actions executed by dexterous humanoid robots |
| SE005 | X (Brett Adcock) | Singapore's Sharpa unveiled SharpaWave, a lifelike robotic hand | Singapore's Sharpa unveiled SharpaWave, a lifelike robotic hand —Features 22 DOF to balance for dexterity and strength —Each fingertip has 1,000+ tactile sensing pixels and 5 mN pressure sensitivity |
| SE006 | Embodied AI Frontier | Sharpa: Investors, Funding | The Frontier of Embodied AI | The flagship Wave hand has 22 active DoF with the proprietary Dynamic Tactile Array (DTA), fingertip force above 20N and 0.02N tactile sensitivity; it entered mass production in October 2025 and won a CES 2026 Innovation Award. |
| SE007 | Sharpa | Sharpa Product Documentation | © Sharpa Pte Ltd. All rights reserved. |
| SE008 | Sharpa | Sharpa Careers | AI Engineering - Experienced Hires / Fresh Graduate Hires |
| SE009 | Robotico Market | Sharpa - Humanoid Robotics Company | Market Data & Profile | Sharpa develops high performance robotic hands with advanced tactile sensing. SharpaWave won CES 2026 Innovation Award with 22 DOF at human scale. |
| SU001 | The Business Times | A*Star, JTC, Grab enter partnerships with robotics firm Sharpa to build Singapore’s physical AI sector | Under JTC and Sharpa’s partnership, the AI robotics company will deploy a fleet of its flagship North robots within the Punggol Digital District (PDD) across various settings – such as retail and food and beverage (F&B) – and use cases. |
| SU002 | LogiShift | Sharpa Case Study: Hesai General Robotics Innovation – LogiShift | The biggest hurdle for general-purpose robots is training data. Unlike ChatGPT, which was trained on the internet, robots need physical motion data. |
| SU003 | Sharpa | All Customer Stories | 97% success on gear meshing: how NTU's DexTeleop-0 framework makes dexterous teleoperation viable at scale |
| SU004 | Sharpa | Tap Customer Story | How shared tactile representations unlock zero-shot contact-rich manipulation on Sharpa Wave — even when the encoder never saw your fingertip |
| SU005 | MANUS | Customer Success Stories - Sharpa | Teleoperating a 22-DoF Sharpa Hand Inside NVIDIA Isaac Lab using MANUS Gloves |
| SU006 | Sharpa | LinkedIn | When comparing dexterous hands vs parallel grippers on tasks like peg insertion, the parallel-jaw gripper needs 20-70 seconds per trial while the Wave only needs 5-10 seconds. | |
| SU007 | Sharpa | Sharpa Careers | AI Engineering - Experienced Hires / Fresh Graduate Hires |
| SU008 | Embodied AI Frontier | Sharpa: Investors, Funding | The company has not yet generated revenue. |
| SR001 | Webiano Digital | NVIDIA, Sharpa and Unitree push humanoid research toward a shared standard | Tactile fingertips must survive repeated contact, cleaning, impacts, misgrasps and calibration drift. Industrial grippers often win because they are simple, rugged and easy to maintain. Dexterous hands must prove they are not only capable in demos but serviceable in daily use. |
| SR002 | Ancestry | Sharpa Surname Meaning & Sharpa Family History at Ancestry.com | There are 82 census records available for the last name Sharpa. Like a window into their day-to-day life, Sharpa census records can tell you where and how your ancestors worked, their level of education, veteran status, and more. |
| SR003 | Sharpa | Sharpa Careers | AI Engineering - Experienced Hires / Fresh Graduate Hires 3 Open roles Commercial 2 Open roles Operations Internships 1 Open roles |
| SR004 | Sharpa | LinkedIn | A single Blizzard takes 50 to 60 consecutive manipulation steps. North has to track whether it's added one, two, or three scoops of Oreo crumbs, plan what comes next, and continuously adjust its grip as it moves from a deformable cup to a spoon to a cabinet door handle. | |
| SR005 | Physix Frontier | Sharpa - Profile | Sharpa, founded in late 2024 by Hesai's team, builds AI robots for dexterous manipulation. Products include the SharpaWave tactile hand and North humanoid. Valued at RMB 22 billion, it secured funding from Alibaba, Meituan, Tencent, and JD.com. |
| SR006 | Robotico | Sharpa - Humanoid Robotics Company | Market Data & Profile | Private Sharpa develops high performance robotic hands with advanced tactile sensing. SharpaWave won CES 2026 Innovation Award with 22 DOF at human scale. Founded 2024$238M raised |
| SR007 | The Frontier of Embodied AI | Sharpa: Investors, Funding | In March 2026 the company first disclosed cumulative funding exceeding RMB 4.5 billion and a post-money valuation of roughly RMB 22 billion (~USD 3.27 billion), backed by Alibaba, Meituan, Tencent, JD.com, Sequoia China, and Qiming Venture Partners. Singapore's EDB has called it a unicorn. The company has not yet generated revenue. |
| SR008 | MarketScreener | Sharpa Pte. Ltd. announced that it has received $669 million in funding from a group of investors | Sharpa Pte. Ltd. announced that it has raised $669 million in a round of funding on August 28, 2026. The transaction involved participation from new investors Alibaba Group Holding Limited, HongShan Capital Advisors Limited, JD.com, Inc., Luminous Ventures, Meiuan, Meituan Dragon Ball, Shenzhen Transsion Holdings Co., Ltd., and Tencent Holdings Limited. The company issued convertible preferred stock in the transaction. The transaction is raised at a post-money valuation of $3,300 million. |
| SR009 | RoboMorrow | Hesai passes 10,000 robot actuators as lidar supplier expands into Physical AI infrastructure | We still do not know the task success rate, number of human interventions, average order cycle time, uptime, MTBF or the economics of a full shift. "Fully autonomous", "zero-retrofit" and "full shift" are primarily Sharpa and deployment-partner claims. Nor should autonomy of the Blizzard workflow be generalized to the entire restaurant: restocking, cleaning, maintenance, safety supervision and exception handling may still require people. |
| SV001 | Venture Capital Tracker | Sharpa Funding, Valuation & Investors | On August 28, 2026 secondary English coverage reported more than ¥4.5 billion (~$670 million) in financing at a reported ¥22 billion post-money, with Alibaba, Meituan, Tencent, JD.com, Transsion, HongShan, Qiming, Meituan DragonBall, and Luminous named. |
| SV002 | Phemex News | Sharpa Discloses Over $620 Million in Funding at 22 Billion Yuan Valuation | General-purpose robotics company Sharpa has disclosed more than 4.5 billion yuan in total funding, marking the first time the startup has publicly revealed its financing history. Investors include Alibaba, Meituan, Tencent, JD.com, Sequoia China, Qiming Venture Partners and Meituan Longzhu, with Sharpa valued at 22 billion yuan post-money. |
| SV003 | MarketScreener | Sharpa Pte. Ltd. announced that it has received $669 million in funding from a group of investors | The company issued convertible preferred stock in the transaction. The transaction is raised at a post-money valuation of $3,300 million. |
| SV004 | The Frontier of Embodied AI | Sharpa: Investors, Funding | The Frontier of Embodied AI | The company has not yet generated revenue. |
| SV005 | RoboMorrow | Hesai passes 10,000 robot actuators as lidar supplier expands into Physical AI infrastructure | We still do not know the task success rate, number of human interventions, average order cycle time, uptime, MTBF or the economics of a full shift. |
| SV006 | Sharpa | Sharpa Careers | AI Engineering - Experienced Hires / Fresh Graduate Hires |
| SV007 | Physix Frontier | Sharpa · Profile | Sharpa, founded in late 2024 by Hesai’s team, builds AI robots for dexterous manipulation. Products include the SharpaWave tactile hand and North humanoid. Valued at RMB 22 billion, it secured funding from Alibaba, Meituan, Tencent, and JD.com. |
| SV008 | Robotico Market | Sharpa - Humanoid Robotics Company | Market Data & Profile | Sharpa develops high performance robotic hands with advanced tactile sensing. SharpaWave won CES 2026 Innovation Award with 22 DOF at human scale. |
| SV009 | Sharpa | LinkedIn | View 74 employees at Sharpa | |
| SV010 | Brett Adcock (@adcock_brett on X) | Brett Adcock (@adcock_brett) on X | Features 22 DOF to balance for dexterity and strength |