AI News AI资讯 1d ago Updated 1d ago 更新于 1天前 52

Physical AI Company Secures Hundreds of Millions in Financing from JinkoSolar, SDIC Innovation to Become 'Brain' of Global Energy Infrastructure 36氪首发 | 物理AI公司获晶科能源、国投创新等数亿融资,要做全球能源基础设施“大脑”

DeepCtrls secured hundreds of millions in RMB Series B funding led by JinkoSolar Strategic Investment, SDIC Innovation, and CMB International, with continued support from Sequoia China and Source Code Capital. The company utilizes its proprietary "PhyAI" engine to integrate physical laws with AI, achieving L4/L5 autonomous closed-loop control that solves the "black box" trust issue in industrial applications. PhyAI demonstrates high generalization with less than 3% error rate, trained on real-ti 物理AI公司深度智控(DeepCtrls)完成数亿元B轮融资,由晶科能源战略投资,国投创新、招银国际领投,资金用于研发及海外市场拓展。 公司自研“PhyAI”物理AI引擎,将物理机理与AI耦合,解决工业AI“黑箱”痛点,实现从L2/L3向L4/L5级自主闭环控制的跨越。 基于30万台设备数据迭代,模型泛化误差低于3%,已服务台积电、腾讯等360+头部客户,并率先实现规模化盈利。 战略定位全球能源基础设施“大脑”,横向拓展至算电协同与源网荷储一体化,纵向追求全场景L5级自主智能。 利用国内复杂场景打磨的技术代差优势,加速全球化布局,重点切入东南亚、中东及北美的高端制造与智算中心市场。

72
Hot 热度
78
Quality 质量
75
Impact 影响力

Analysis 深度分析

TL;DR

  • DeepCtrls secured hundreds of millions in RMB Series B funding led by JinkoSolar Strategic Investment, SDIC Innovation, and CMB International, with continued support from Sequoia China and Source Code Capital.
  • The company utilizes its proprietary "PhyAI" engine to integrate physical laws with AI, achieving L4/L5 autonomous closed-loop control that solves the "black box" trust issue in industrial applications.
  • PhyAI demonstrates high generalization with less than 3% error rate, trained on real-time data from over 300,000 devices across complex sectors like semiconductors and data centers.
  • DeepCtrls has achieved scalable profitability and serves over 360 top-tier clients, including TSMC, Tencent, and CATL, while expanding globally into Southeast Asia, the Middle East, and North America.
  • The strategic focus is evolving from device-level control to system-level optimization and global energy infrastructure reshaping, addressing energy bottlenecks in the AI compute era.

Why It Matters

This development highlights a critical shift in AI application from digital reasoning to physical world control, offering a viable solution to the energy efficiency challenges facing massive AI infrastructure deployments. For industry practitioners, it demonstrates how integrating domain-specific physical constraints with machine learning can overcome the reliability and interpretability barriers that have historically stalled industrial AI adoption. Furthermore, the successful commercialization and global expansion of such technology signal that "Physical AI" is moving from theoretical concepts to a mature, investable market segment with immediate ROI potential in energy-intensive industries.

Technical Details

  • PhyAI Engine Architecture: A proprietary physical AI engine that deeply couples physical mechanisms with AI models, enabling the system to understand underlying equipment physics rather than relying solely on data correlation.
  • Autonomy Level: Achieves L4/L5 level autonomous control, moving beyond diagnostic recommendations (L2/L3) to enable self-optimization, autonomous decision-making, and closed-loop control in real-world environments.
  • Performance Metrics: The model achieves a generalization error of less than 3%, approaching theoretical system limits, and is continuously iterated based on real-time operational data from over 300,000 devices.
  • Scalability & Generalization: The technology allows for cross-device, cross-system, and cross-scenario migration without needing to rebuild models from scratch for each new scenario, facilitating rapid deployment.
  • Application Scope: Initially focused on electromechanical energy systems, the technology is applied to high-complexity, high-energy-consumption scenarios such as semiconductor manufacturing, new energy storage, and intelligent computing centers.

Industry Insight

  • Energy Efficiency as a Competitive Moat: As AI compute demands surge, energy consumption becomes a primary bottleneck. Companies that can deploy AI-driven energy optimization with guaranteed reliability will gain significant cost advantages and regulatory compliance benefits in global markets.
  • Trust Through Physics-Informed AI: The industrial sector requires deterministic and interpretable outcomes. The success of Physical AI suggests that future industrial AI solutions must prioritize hybrid modeling (physics + data) over pure data-driven approaches to gain enterprise trust and achieve scale.
  • Global Market Expansion Opportunities: With Western markets still largely relying on rule-based L2 controls, there is a substantial technology gap. DeepCtrls' strategy of leveraging complex domestic use cases to build robust models for export presents a strong pathway for Chinese AI firms to capture high-value international markets in manufacturing and infrastructure.

TL;DR

  • 物理AI公司深度智控(DeepCtrls)完成数亿元B轮融资,由晶科能源战略投资,国投创新、招银国际领投,资金用于研发及海外市场拓展。
  • 公司自研“PhyAI”物理AI引擎,将物理机理与AI耦合,解决工业AI“黑箱”痛点,实现从L2/L3向L4/L5级自主闭环控制的跨越。
  • 基于30万台设备数据迭代,模型泛化误差低于3%,已服务台积电、腾讯等360+头部客户,并率先实现规模化盈利。
  • 战略定位全球能源基础设施“大脑”,横向拓展至算电协同与源网荷储一体化,纵向追求全场景L5级自主智能。
  • 利用国内复杂场景打磨的技术代差优势,加速全球化布局,重点切入东南亚、中东及北美的高端制造与智算中心市场。

为什么值得看

本文揭示了AI从数字世界推理向物理世界控制落地的关键转折,展示了“物理AI”如何解决工业场景信任与可靠性难题。对于关注AI商业化闭环及能源基础设施智能化的从业者而言,深度智控的案例提供了从技术验证到规模化盈利及出海扩张的完整参考路径。

技术解析

  • PhyAI物理AI引擎:核心技术创新在于将机电能源系统的物理机理深度融入AI模型,而非单纯依赖数据驱动。这种“物理+AI”的路径赋予了决策过程可解释性、确定性和可追溯性,解决了传统工业AI“黑箱不可信”的落地障碍。
  • 控制等级跨越:区别于停留在诊断建议阶段的L2/L3级工业AI,深度智控实现了L4/L5级自主寻优、自主决策和闭环控制。AI不仅能理解物理世界,还能在真实环境中进行安全可靠的实体操作与优化。
  • 高泛化性与低误差:模型基于超过30万台设备的实时运行数据持续迭代,泛化误差低于3%,逼近系统理论最优极限。具备跨设备、跨系统、跨场景的迁移复制能力,无需在新场景从零积累数据,极大降低了部署成本。
  • 标准化产品DeepBot:推出标准化产品DeepBot,支持即插即用,从设备端AI控制模块延伸至系统级优化控制,形成了可规模化的产品矩阵。

行业启示

  • “物理AI”成为工业落地新范式:随着生成式AI在数字世界的能力饱和,AI产业的下半场焦点转向物理世界的自主控制。具备物理机理约束的AI模型因其可靠性与可解释性,将成为高端制造、能源管理等关键基础设施的核心技术底座。
  • 中国场景构筑全球竞争壁垒:中国复杂、高能耗且对稳定性要求极高的工业体系(如半导体、数据中心),为物理AI提供了独特的训练与验证环境。这种在极端场景下打磨出的技术能力,构成了难以被海外竞争对手简单复刻的竞争壁垒。
  • 能源与算力协同是出海新机遇:全球智算中心爆发导致能源与温控成为瓶颈,物理AI在能效优化上的优势使其具备强大的海外市场需求。企业应抓住这一窗口期,通过技术代差优势切入中东、北美等高价值市场,重塑全球能源基础设施。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Funding 融资 Robotics 机器人 Product Launch 产品发布