Orient Securities Company Limited (DFZQ) has released a research report suggesting that the industrial significance of Physical AI lies in advancing AI from "generating content" to "completing tasks," and upgrading from a "software tool" to "delivering outcomes." This shift is expected to accelerate the adoption of physical world entry points such as robotic systems, intelligent inspection, industrial automation, and edge intelligence, potentially reshaping the business models of traditional manufacturing and industrial services. The key to its implementation is building a closed-loop capability encompassing "AI models + industrial data + process know-how + automated equipment + supply chain delivery." The firm believes Physical AI is likely to become a crucial direction for future AI applications, with sectors like industrial manufacturing, intelligent inspection, and industrial simulation standing to benefit.

Transition from Digital to Physical World

Historically, AI applications have primarily focused on digital scenarios such as text, code, images, and office collaboration, essentially aimed at enhancing human efficiency. The core shift with Physical AI is its entry into real-world physical scenarios like manufacturing, semiconductors, robotics, warehousing and logistics, and industrial inspection, where it completes a closed loop of perception, planning, execution, and delivery. The recent concept of AI-Native Service proposed by Silicon Valley venture capitalists further strengthens this direction: AI companies are no longer just selling software subscriptions or tool licenses but are directly delivering business outcomes for clients, such as completing PCB design and manufacturing, delivering metal components, finalizing chip design, executing industrial inspections, or optimizing factory operations. The report states that the industrial significance of Physical AI is to propel AI from content generation to task completion and from a software tool to outcome delivery, driving accelerated adoption of physical world entry points like robotic systems, intelligent inspection, industrial automation, and edge intelligence, with the potential to reshape traditional manufacturing and industrial service business models.

Key Implementation Areas

Unlike general-purpose office software, scenarios such as manufacturing, semiconductors, PCBs, industrial materials, and metal processing are characterized by high complexity, high customer value, high delivery barriers, and a strong results orientation, making them naturally suitable for the implementation of AI-native service models. Taking PCB and hardware manufacturing as an example, in the traditional model, clients need to separately procure design tools, find engineers, interface with factories, and repeatedly conduct prototype verification, leading to long processes, high costs, and significant delivery uncertainty. In contrast, AI-native manufacturing services can integrate design, simulation, process optimization, supply chain management, and production delivery into an end-to-end capability. Clients would then pay not for software features or labor hours but for the final delivered outcome of PCBs, components, or chips. The firm believes the key to Physical AI implementation lies in constructing a closed-loop capability that combines AI models, industrial data, process know-how, automated equipment, and supply chain delivery. Once such companies successfully navigate a vertical scenario, they are likely to establish a strong data flywheel and achieve economies of scale.

Shift in Business Model

The difference between AI-native services and traditional SaaS or industrial equipment companies is that revenue is no longer primarily derived from software licensing, equipment sales, or billed labor hours, but rather from fixed project fees, pay-per-delivery, cost-saving shares, or outcome-based shares. From the client's perspective, this model reduces upfront capital expenditure and trial-and-error costs, transforming the service from a cost item into a profit increment. From the supply side, the higher the proportion of work undertaken by AI, the greater the opportunity for revenue growth to decouple from increases in employee headcount, and marginal profit rates are expected to improve with scale. Therefore, the investment focus in the Physical AI direction should shift from simply asking "does it have robots/AI models" to examining whether a company possesses real-world scenario data, automated delivery capabilities, process expertise, and a commercial closed loop. The firm considers future areas worth watching to include AI-driven industrial design and manufacturing, intelligent inspection, robotic execution systems, industrial simulation, edge AI, and industrial vision. Related companies are expected to continuously benefit as AI transitions from the digital to the physical world.

Risks mentioned include intensifying industry competition and slower-than-expected industry advancement.