• 亿邦公众号
    • APP下载

央企资本下场 这家“物理 AI”黑马又融了数亿元

作者:陈美 文章来源:投中网

2026-06-26 10:00

作者:陈美 文章来源:投中网

2026-06-26 10:00

邦小白快读

EN
全文速览

本文核心信息是物理AI黑马企业深度机智完成数亿元新融资,透露出当前物理AI/具身智能赛道的最新发展情况,核心干货如下

1.融资层面:本轮融资距离上一轮仅1个月,由央企国寿旗下的国寿长三角科创基金领投,十余家市场化机构、产业资本跟投,险资、产业资本、市场化机构集体入场,体现了资本对该赛道和深度机智技术的高度认可。

2.技术层面:深度机智走“人类第一视角”技术路线,已经完成Zero-Shot零样本泛化实操验证,从根源缓解了行业真机数据采集难、成本高的痛点,还在相关国际赛事中取得榜首成绩,技术实力获得验证。

3.商业化层面:目前已经在中高校科创教育、科研基础设施两大赛道落地,拿到多笔订单,产品还实现出海,未来将继续打造原生物理智能大模型,落地更多工业、康养等真实场景。

本文透露出当前具身智能、物理AI赛道的产业与消费趋势,也给AI相关品牌提供了融资、场景落地的参考经验,核心干货如下

1.行业趋势层面:当前资本对具身智能的投资逻辑已经发生转变,从比拼人形机器人硬件样机,转向看重AI大脑能力和底层数据能力,物理AI是接下来的高增长赛道,养老、工业、教育都是核心落地场景。

2.产业协同层面:央企资本这类长线投资方非常认可“技术+场景匹配”的模式,深度机智凭借能适配国寿养老场景的数采设备,获得央企资本领投,这种跨界协同模式值得品牌参考,有线下场景资源的品牌可以对接技术方获得资本支持。

3.产品布局层面:目前市场更认可能解决真实痛点的落地产品,纯概念、纯展示的产品已经不被资本和市场看好,品牌布局AI相关业务需要优先落地真实场景验证价值。

本文给AI相关赛道的卖家提供了明确的增长机会和可参考的发展经验,核心干货如下

1.机会层面:当前具身智能赛道资本加速涌入,底层数据采集、AI科创教育装备、科研基础设施都是需求增长明确的赛道,尤其是能解决真机数据采集难痛点的轻量化产品,市场缺口较大,卖家可以提前布局相关方向。

2.融资与合作层面:当前产业资本、险资都非常看好物理AI和具体场景结合的项目,有技术硬实力的卖家,可以对接线下场景方、产业资本,凭借场景适配能力获得融资支持,也可以像深度机智一样和影石创新这类硬件品牌达成战略合作,共建数据生态,拓展业务边界。

3.发展参考与风险提示:深度机智选择差异化技术路线,直击行业核心痛点,快速实现商业化落地的路径值得中小卖家参考,同时赛道竞争看重核心技术,没有硬技术壁垒的玩家很容易被淘汰,需要提前构建竞争力。

本文给工厂推进智能化转型、对接新的商业机会提供了不少启发,核心干货如下

1.智能化改造机会:当前产业资本对具身智能落地工厂有明确需求,深度机智的轻量化数采设备,可以完整记录一线员工的操作行为,在人类成熟经验和机器执行之间搭建桥梁,帮助工厂沉淀专属的物理操作技能库,非常适合工厂做智能化改造。

2.数字化转型启示:工厂推进智能化不需要一开始就全线更换高端硬件,可以先通过采集成熟工人的操作数据训练机器人,依托人类经验降低机器人训练成本,缩短适配周期,降低转型的门槛和风险。

3.产业合作机会:目前物理AI企业需要大量工厂场景做技术验证和数据沉淀,有丰富一线操作岗位的工厂,可以对接这类AI企业开展合作,既可以推进自身的智能化升级,也能获得产业协同带来的额外收益,还能提前布局未来的智能化生产。

本文明确了当前物理AI/具身智能行业的发展趋势、核心痛点,给相关服务商提供了业务发展方向,核心干货如下

1.行业发展趋势:当前具身智能赛道资本加速涌入,赛道核心竞争点已经从硬件本体制造,转向AI大脑开发和底层数据能力建设,数据采集服务、模型训练开发服务、相关基础设施服务的需求正在快速增长,是服务商接下来可以重点布局的方向。

2.核心客户痛点:目前全行业普遍面临几个核心痛点,分别是训练机器人所需的真机数据体量不足、采集成本高昂、场景覆盖单一、数据质量参差不齐,同时现有多类模型彼此割裂,普遍缺乏物理常识,无法满足真实落地需求。

3.解决方案参考:深度机智的“人类第一视角数据采集+以动作为中心建模”方案,从根源缓解了数据采集难题,分体式数采设备的设计大幅提升了长时间采集的效率,服务商开发相关服务可以参考该思路,还可以通过跨界合作的方式,联合不同领域企业共同搭建底层数据生态,拓展业务空间。

本文给AI相关平台透露了当前赛道的需求,也明确了平台布局和风险规避的方向,核心干货如下

1.赛道需求层面:当前物理AI、具身智能赛道的优质初创企业,除了资本支持外,非常看重产业场景资源的对接,有技术实力的初创企业更愿意和能链接技术方、场景方、资本方的平台合作,依托平台资源实现快速落地。

2.招商与布局方向:物理AI是当前资本高度看好的高增长赛道,已经出现了一批获得央企资本、产业资本认可的优质初创企业,平台可以针对性开展招商,提前布局下一个增长风口,还可以围绕具身智能搭建产业协同生态,满足不同参与者的对接需求,提升平台自身的价值。

3.风险规避层面:当前赛道不少玩家只侧重做硬件样机展示,没有核心技术和落地能力,平台引入相关项目时,需要优先考察底层技术实力和商业化落地能力,规避只讲故事没有硬实力的项目风险,同时可以重点支持能解决行业核心痛点的技术型项目。

本文透露出当前具身智能领域的最新产业、技术动向,对相关研究有很高的参考价值,核心干货如下

1.产业新动向:当前资本已经从投资硬件转向押注底层技术,集体看好“人类第一视角”路线的具身通用智能底座,原本偏向长期投资、风格审慎的央企资本也开始下场布局物理AI赛道,赛道融资节奏明显加快,产业资源也在向该方向聚集。

2.技术新动向:行业已经完成验证,确认了“人类学习”技术路线的可行性,也就是依托人类第一视角数据训练机器人,能从根源解决真机数据采集难的行业痛点;技术路线上也打破了VLA或世界模型的二元争论,提出要围绕真实任务整合各类模型的能力,解决实际问题。

3.商业模式新参考:深度机智在技术落地早期,就跑通了中高校科创教育装备+科研基础设施的双赛道商业化模式,已经拿到多笔真实订单,还开启了出海布局,为整个赛道的早期商业化提供了可参考的新方向。

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

This article covers a key update from Deep Ingenuity, a rising star in the physical AI space, which has just completed a multi-hundred-million RMB new financing round, and shares the latest developments in the physical AI/embodied intelligence sector:

1. Financing: This new round comes just one month after the company’s last funding. It is led by the China Life Yangtze River Delta Science and Technology Innovation Fund, an investment vehicle under state-owned insurance giant China Life, with participation from more than ten market-oriented institutions and industrial capital. The simultaneous entry of insurance capital, industrial capital and market investors reflects strong investor confidence in both the sector and Deep Ingenuity’s technology.

2. Technology: Deep Ingenuity pursues a "first-person human perspective" technical approach, and has already completed zero-shot generalization validation in real-world practice. This fundamentally addresses the industry’s long-standing pain points of high costs and difficulties in collecting real-machine data. The company also topped the leaderboard in related international competitions, confirming its technical strength.

3. Commercialization: Deep Ingenuity has already secured commercial traction and multiple orders in two core segments: university and K12 AI science education, and research infrastructure. Its products have also entered international markets. Moving forward, the company plans to continue developing a native physical AI large model, and expand into more real-world scenarios including industrial applications and elderly care.

This article outlines the latest industrial and consumer trends in the embodied intelligence and physical AI sector, and offers actionable insights for AI-related brands on financing and commercial deployment:

1. Industry Trends: Investor logic in the embodied intelligence space has shifted. Instead of competing to back humanoid robot hardware prototypes, investors now prioritize AI brain capabilities and underlying data infrastructure. Physical AI has emerged as a high-growth segment, with core application scenarios including elderly care, industrial production and education.

2. Industrial Collaboration: Long-term investors such as state-owned central enterprise capital strongly favor the "technology-scenario matching" model. Deep Ingenuity secured the leading investment from China Life in part because its data collection equipment can be adapted for China Life’s elderly care scenarios. This cross-industry collaboration model is a valuable reference for brands: brands with offline scenario resources can partner with technology players to attract capital support.

3. Product Strategy: The market now prioritizes commercialized products that solve real pain points. Purely conceptual, demonstration-only products no longer win favor from investors or customers. Brands looking to enter AI-related businesses should prioritize validating product value in real-world scenarios first.

This article outlines clear growth opportunities and actionable insights for sellers in AI-related sectors:

1. Growth Opportunities: Capital is flowing into the embodied intelligence sector at an accelerating pace, with demand growing strongly for underlying data collection, AI science education equipment, and research infrastructure. In particular, there is a large unmet market need for lightweight products that solve the pain point of difficult real-machine data collection. Sellers can position themselves early in these segments.

2. Financing and Partnerships: Industrial capital and insurance capital are actively looking for physical AI projects that integrate with real-world scenarios. Sellers with solid core technology can partner with offline scenario owners and industrial capital to secure financing by leveraging their scenario adaptation capabilities. Like Deep Ingenuity, they can also form strategic partnerships with established hardware brands such as Insta360 to co-build data ecosystems and expand business boundaries.

3. Development Reference and Risk Warning: Deep Ingenuity’s path—pursuing a differentiated technical route that directly addresses core industry pain points and achieving fast commercialization—provides a strong reference for small and medium-sized sellers. At the same time, competition in this sector relies heavily on core technology; players without solid technical barriers are at high risk of being eliminated, so building competitive moats should be a priority.

This article offers valuable insights for factories looking to advance intelligent transformation and tap into new business opportunities:

1. Opportunities for Intelligent Upgrading: Industrial capital has clear demand for embedding embodied intelligence into factory operations. Deep Ingenuity’s lightweight data collection equipment can fully record the operating behaviors of front-line workers, building a bridge between mature human experience and machine execution, and helping factories build their own proprietary libraries of physical operation skills. This is an ideal solution for factory intelligent transformation.

2. Insights for Digital Transformation: Factories do not need to fully replace all their hardware with high-end equipment at the start of transformation. Instead, they can first collect operating data from experienced workers to train robots, leveraging human experience to cut robot training costs, shorten adaptation cycles, and lower the threshold and risk of transformation.

3. Industrial Collaboration Opportunities: Physical AI companies currently need access to a large number of factory scenarios for technical validation and data accumulation. Factories with abundant front-line operating positions can partner with these AI companies to advance their own intelligent upgrading, earn additional revenue from industrial collaboration, and get a head start on future intelligent production.

This article clarifies the latest development trends and core pain points of the physical AI/embodied intelligence industry, and outlines clear business development directions for relevant service providers:

1. Industry Development Trends: Capital is accelerating into the embodied intelligence sector, and the core of competition has shifted from hardware manufacturing to AI brain development and underlying data capability building. Demand for data collection services, model training and development services, and related infrastructure services is growing rapidly, making these key areas for service providers to focus on.

2. Core Client Pain Points: The industry as a whole currently faces several core pain points: insufficient volume of real-machine data for robot training, high collection costs, limited scenario coverage, uneven data quality, and fragmented existing models that generally lack physical common sense, failing to meet real-world deployment requirements.

3. Solution Reference: Deep Ingenuity’s "first-person human perspective data collection + action-centric modeling" approach fundamentally alleviates data collection challenges, and its split data collection device design greatly improves the efficiency of long-duration collection. Service providers developing related offerings can reference this approach, and can also build underlying data ecosystems through cross-industry collaboration with enterprises from different fields to expand their business scope.

This article outlines the current sector-specific needs for AI-related platforms, and clarifies directions for platform布局 and risk mitigation:

1. Sector Needs: Beyond capital support, high-quality startups in the physical AI and embodied intelligence space now prioritize access to industrial scenario resources. Technology-strong startups prefer to partner with platforms that can connect technology providers, scenario owners and investors, to achieve fast commercial deployment through platform resources.

2. Investment and招商 Direction: Physical AI is a high-growth sector strongly favored by capital today, and a number of high-quality startups have already won recognition from central enterprise capital and industrial capital. Platforms can target their招商 efforts here to position early for the next growth wave, and can also build an industrial collaborative ecosystem around embodied intelligence to meet the matching needs of different participants, boosting the platform’s own value.

3. Risk Mitigation: Many players in the current sector only focus on hardware prototype demonstration, with no core technology or commercial deployment capabilities. When onboarding related projects, platforms should prioritize evaluating underlying technical strength and commercialization capabilities to avoid the risk of backing story-driven projects with no real substance, and should prioritize supporting technology-driven projects that solve core industry pain points.

This article shares the latest industrial and technological developments in the embodied intelligence field, offering high reference value for related research:

1. New Industrial Trends: Investors have shifted their focus from hardware investment to betting on underlying technology, and are collectively bullish on the "first-person human perspective" route for a general embodied intelligence base. Even traditionally cautious, long-term-oriented central enterprise capital has started to enter the physical AI space, driving faster financing cycles and concentrating industrial resources in this direction.

2. New Technology Developments: The industry has now validated the feasibility of the "human-like learning" technical route—training robots based on first-person human perspective data fundamentally solves the core industry pain point of difficult real-machine data collection. This approach also breaks the binary debate between VLA and world models, proposing that capabilities from various models should be integrated around real tasks to solve practical problems.

3. New Business Model Reference: Early in its technology commercialization, Deep Ingenuity has already built a viable dual-track business model focused on university and secondary school AI science education equipment and research infrastructure, secured multiple real commercial orders, and launched international expansion. This provides a valuable new reference for early-stage commercialization across the entire sector.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

距离上一轮融资,仅过去1个月。

当“物理AI”被推向聚光灯下后,资本正加速涌入。

今日,投中网获悉,“物理AI”黑马——深度机智再获数亿元融资,本轮融资由国寿长三角科创基金领投,老股东普华资本、诚通科创基金持续加码,蓝湖资本、博彦科技、磐谷创投、朝晖资本、财鑫资本、道禾长期投资、易高资本、明德资本等十余家市场化机构、产业资本同步入局。

值得注意的是,这距离上一轮融资,仅过去1个月。但相比融资节奏,市场更关注的,或许是本轮投资方的“阵容”与“站位”:险资、产业资本、市场化基金均不约而同坐到了同一张牌桌上,并达成共识。

险资、产业资本、市场化机构齐出手

就本轮投资方来说,国寿长三角科创基金的出手颇具标志性意义。

首先作为领投方,该基金隶属于中国人寿旗下,背靠副部级中央金融企业国寿集团,规模超50亿元,而该集团向来以长线、审慎的投资风格著称,对投资标的的技术硬实力与长期价值有极高的要求。

此番领投,显示出央企资本对深度机智技术实力的看好,以及场景协同前景的认可。

在交流中,深度机智创始人陈凯告诉投中网:“自去年下半年开始,我们就与他们保持深度交流;国寿长三角科创基金高度认可我们在‘人类第一视角’技术路线上的前瞻布局,以及被验证了的全栈技术和商业化落地能力。”

技术之外,打动这一央企资本的,还有场景端的适配“空间”。众所周知,国寿长三角科创基金背靠国寿集团,后者手握大量养老、康养线下场景,而深度机智自研的人体工学第一视角数采设备,能够记录专业护理人员在服务过程中的完整动作与交互过程。这些动作数据经脱敏处理后,能沉淀出海量的物理交互细节,成为养老机器人 “大脑” 训练不可或缺的数据燃料。

从某种程度上来说,这种产业协同让央企资本这类长线机构,看到了未来推动服务机器人规模化走进养老院的可行路径,并愿意为此买单。

同样的逻辑,也吸引了博彦科技的出手。陈凯告诉投中网,与保险资本类似,产业资本也对具身智能进入工厂有着明确需求。“我们的数采设备,可以轻量化方式落地工厂,完整记录下一线员工的操作行为,在‘人类经验’与‘机器执行’之间搭建起一座桥梁,沉淀出独特的物理操作技能库。”

这份独特的技能,正演变为行业共识:随着各路资本的出手,对深度机智的押注,已不再是单纯的“财务行为”,而是一场对“人类第一视角”具身通用智能底座的集体投票。

能力突破:Zero-Shot泛化验证“人类学习”路线

如果说,过去两年,具身智能赛道的融资逻辑还比较单一:谁能做出更灵活的人形机器人样机,谁就能拿钱。但如今,在这场竞赛中,资本的目光已从“本体”迁移至“大脑”,从“硬件展示”转向“数据底层能力”。

截至目前,一个残酷的事实摆在所有具身智能“玩家”的前面:训练所需真机数据体量不足、数据采集成本高昂,场景的覆盖也比较单一,数据质量更是参差不齐……但深度机智在过去几个月内完成的一系列验证,却刚好击中了行业痛点“最集中”的位置。

陈凯告诉投中网,今年3月深度机智发布的PhysBrain 1.0基座模型,已实现机器人的自主纠错与跨场景泛化。近期,团队再获关键性突破:依托人类第一视角数据,完成Zero-Shot零样本泛化实操验证。

在魔方入袋的这一实验中,深度机智仅用人类收纳、抓取动作的“第一人类视角”数据做训练策略,就可让机器人在无需输入任何真机训练数据的情况下,自主完成抓取、搬运、收纳整套连贯操作。

这一技术成果,验证了“人类学习”路线从人类数据、模型训练向真实机器人执行迁移的可行性;也证明了,“人类第一视角”数据蕴含的物理交互常识,可以直接迁移到机器人身上,从而从根源上缓解行业“真机数据采集难”的瓶颈。

在深度机智内部,也将这些能力拆解为在成立不到一年时间之内,对“人类学习”路线底层能力的集中落地:首先是2025年12月建成数十万小时级人类第一视角高质量数据集DeepAct;之后,今年2月全自由度拟人体灵巧操作机器人Prime场景落地;接近着,3月发布PhysBrain 1.0具身通用智能基座模型,并对2025年10月确立的三大技术原则——Human-Centric Data(以人类数据为起点)、Aciton-Centric Modeling(以动作为中心建模)Robot for AI(机器人为AI而生),完成全部验证。

注:深度机智由北京中关村学院与中关村人工智能研究院孵化,近期作为产学研企业受到央视《朝闻天下》与《新闻联播》的关注与报道。

与此同时,深度机智在世界模型与数据引擎方向也取得进展。5月15日,深度机智在WorldArena Track 2 Data Engine赛道以88.5分断崖夺冠。此前在Track 1中,深度机智Z-WM也以64.96分超越前榜首WorldScape v0.2,并在后续榜单中保持前列。

这一系列的成绩,并不意味着深度机智选择了单一的“世界模型路线”。相反,它进一步验证了公司Action-Centric Modeling(以动作为中心建模)的技术判断:具身智能不应被简单归为VLA或世界模型之争,而是要围绕真实任务,把多模态大模型、视觉基础模型、世界模型等能力组织起来,共同服务于动作理解、动作预测和动作执行。

陈凯判断,未来机器人或许不再需要先“自己练一遍”才能干活,而是从人类的操作中直接“汲取经验”。“这些技术能力的突破,为今后直接依靠‘人类数据’降低机器人训练成本、提升泛化能力、缩短本体适配周期,提供了关键支撑。”

商业化加速:多产品线落地验证

技术路线的验证,也带来了商业端的回报。投中网获悉,目前深度机智全线产品均已实现商业落地,多条产品线签下多笔订单,主要集中在两大赛道:

一是中高校科创教育赛道。以Prime Lite为例,这是一款轻量化教育人形机器人。区别于其他教育机器人,Prime Lite不是简单的编程工具,而是一套完整的具身智能开发平台。目前,深度机智已与北京市十一学校共建具身智能示范教室,并与多所头部高中及高校达成合作。

据悉,作为合作落地的具体实践,近期首场具身智能科创工作坊也在北京十一中学举办,近50名师生零距离体验了Prime Lite的组装与调试。

国内布局之外,深度机智的教育方案也落地澳大利亚,新加坡、韩国、哈萨克斯坦等院校正在对接中,加速了中国具身智能教育方案的出海。

陈凯认为,Prime Lite之所以能走进高校,源于它可以让学生从零开始设计、制造机器人并部署模型。“我们希望将前沿技术转化为学生能理解、可动手、持续探索的教育内容,培养未来的‘造物主’。”

当然,更深层次的价值在于,通过教育场景批量培育开发者,为未来产业规模落地储备人才队伍。

二是科研基础设施赛道。当行业普遍意识到高质量第一视角数据的价值时,深度机智的轻量化数据采集方案受到多家头部数据企业关注,他们正批量采购设备用于数据采集。

关于数采设备的设计思路,陈凯透露,团队放弃了传统全头戴式方案,将电池与主控模块置于腰部。这既减轻了佩戴者头部负担,也大幅提升了长时间采集作业的效率。值得一提的是,这一分体式第一视角数据采集设备的设计思路,与美国Scale AI最近发布的分体式数采设备不谋而合,而深度机智同类同质化设备早在去年6月便已完成定型落地。

最新进展是,深度机智已与影石创新达成战略合作,依托影石在全景视觉领域的技术积累,共同加速物理AI底层数据生态的规模化建设。

谈及长远规划,陈凯表示这场竞赛远未结束。“接下来,深度机智的核心目标:一是持续扩大原生人类第一视角数据集规模;二是在模型层面打造统一的‘原生物理智能大模型’,加速拟人机器人进驻工业、实验室、康养等真实场景。”

之所以强调“原生”,源于陈凯发现当下多模态大模型、VLM模型、世界模型虽各司其职,却彼此割裂,普遍缺乏物理常识。深度机智要做的,是摆脱多模型拼接方案,打造统一的原生物理智能大模型,从底层构建物理真实性、多模态协同、通用任务规划三大能力,解决行业痛点。

注:文/陈美,文章来源:投中网(公众号ID:China-Venture),本文为作者独立观点,不代表亿邦动力立场。

【版权提示】亿邦动力倡导尊重与保护知识产权。未经许可,任何人不得复制、转载、或以其他方式使用本网站的内容。如发现本站文章存在版权问题,烦请提供版权疑问、身份证明、版权证明、联系方式等发邮件至run@ebrun.com,我们将及时沟通与处理。

已点赞

+1

{{ praise_text }}

好文章,值得被更多人看到

广告

下一篇

征集报名《品牌数字化服务商资源手册》

《品牌数字化服务商资源手册》展示数字化服务商风采,亿邦全年零售大会通发,触达更多资源,马上加入!

马上入驻

咖啡里没有咖啡豆?可口可乐下场 首款“无豆咖啡”来了

从这个角度来看,可口可乐不是在做咖啡的“替代品”,而是在为咖啡寻找一个“备份方案”。