After generative AI swept the globe, driving numerous industrial developments and reforms, Qualcomm's next move is...Dragonwing IO seriesThey shared their views on developing physical AI and expanding the robotics market, while also revealing the challenges they currently face.
In an interview during CES 2026, Qualcomm Vice President and General Manager of Automation Operations and Robotics Solutions, Anshuman Saxena, and Head of Physical AI Engineering, Ahmed Torad, provided an in-depth analysis of how Qualcomm is transforming its experience and technological assets accumulated over the past 10 years in the field of Advanced Driver Assistance Systems (ADAS) into a core driving force for the standardization and scaling of the robotics industry.
Robots are like "self-driving cars" in "unstructured environments".
"If you think about it carefully, a car is actually a kind of robot." Anshuman Saxena used this to express Qualcomm's idea of entering the robotics market, which is to regard robots as a kind of "self-driving car" design, but the details involved become more complex, including many unstructured environmental factors.
Over the past decade, Qualcomm has solved many extremely complex problems in the automotive industry: such as how to perform real-time perception and decision-making while moving at high speeds, while ensuring absolute safety and low power consumption. Now, Qualcomm is applying this proven "recipe" to robots.
Anshuman Saxena points out that unlike cars which drive in "structured environments" with lane markings and traffic rules, robots often need to operate in "unstructured environments" such as factories and homes, which makes the challenges even greater.
This is why Qualcomm emphasizes that the shift from simple AI computation to "physical AI" will present more challenges. Ahmed Torad explains that generative AI processes structured information such as text and images, while physical AI must understand the physical laws of the real world, gravity, friction, and the consequences of interacting with the real world. This is not just a computational issue, but also a matter of security and reliability.
In order for robots to integrate smoothly into the real world and coexist safely with humans, robots must continuously transform and learn a lot of unstructured information in the real world. However, their operation often involves more complex calculations and energy costs, and even the convenience of actual deployment and application must be considered.
Breaking the "fragmentation" deadlock with Dragonwing IQ series chips
The current robotics industry faces a serious problem of "fragmentation." Anshuman Saxena observed that in the past, the development of a robot often involved customized hardware and software development for a single purpose (such as sweeping or carrying), which resulted in long development cycles, high costs, and difficulty in scaling up.
Qualcomm's solution, however, aims to change the status quo through a standardized platform. With Qualcomm's newly launched Dragonwing IQ series robot chips, developers can create everything from simple industrial arms and wheeled robots to complex humanoid robots on the same hardware architecture.
The core advantage of this platform lies in its integration of heterogeneous computing capabilities, combining the heterogeneous computing of CPUs, GPUs, and NPUs to simultaneously handle visual perception, path planning, and motor control. Ahmed Torad emphasized that Qualcomm's advantage lies in "scalability." The same SDK tools and software framework can be used for grippers with only a few motors, or extended to humanoid robots with dozens of joints, thereby significantly reducing the development threshold for industry players.
"Safety Fence": AI cannot have hallucinations.
In models like ChatGPT, while AI may occasionally exhibit hallucinations, these rarely have a significant impact and are often noticed and corrected by users. However, in the field of robotics, such occurrences are absolutely unacceptable.
"You can't let a robot hallucinate and then crash into a wall or injure someone," Ahmed Torad pointed out. Qualcomm has incorporated automotive-grade "safety guardrails" into its platform. This is a real-time monitoring mechanism independent of the AI model, ensuring that the robot's decisions comply with physical limitations and safety regulations.
This is also the biggest difference between Qualcomm and pure software AI companies. Qualcomm is a "systems company," which not only provides computing power but also considers heat dissipation, power management, and real-time requirements from the chip level. Especially in battery-powered robots, Qualcomm's performance per watt technology advantage accumulated in the mobile phone and automotive fields has become a moat that is difficult for competitors to cross.
Pragmatism: Humanoid robots are not the only design direction
Although humanoid robots stole the show at CES 2026 and have become one of the key areas of development in the robotics industry in recent years, Qualcomm does not believe that robots must be designed to look humanoid.
Anshuman Saxena believes that humanoid robots are indeed the ultimate form for general applications. After all, imitating human appearance and the movements of two hands and two fingers will enable them to be applied in many industrial applications or quickly transferred to different market demands, unlike robots designed for warehousing systems, which have greater limitations in deployment and application in other fields.
However, in current commercial applications, about 70% of the usage scenarios (such as logistics, warehousing, and retail) can actually meet many needs through wheeled mobile robots or simple gripper designs to pick up items. Therefore, it is clear that it is not necessary to design robots to be close to humanoid or have two fingers to complete work tasks more efficiently and at a lower cost.
Therefore, Qualcomm's market strategy is not to lock in a specific type, but to provide a universal "brain". Whether it is a humanoid robot designed to mimic human movements or a wheeled vehicle designed for a specific task, Qualcomm's platform can support it.
Analysis of viewpoints
The conversation between Anshuman Saxena and Ahmed Torad reveals that Qualcomm is replicating its successful model in the smartphone and automotive markets: providing a highly integrated, low-power, and standardized computing platform that allows partners to flourish.
As AI moves from the cloud to the edge, the robotics industry is experiencing a "iPhone moment" similar to the early development of smartphones. Qualcomm is clearly not content with being just a chip supplier; it is trying to define the computing standards for the era of "physical AI," so that robots are no longer just products of the laboratory, but practical tools that can truly enter factories and homes.
