SK Telecom provides telecom services regardless of whether it's an iPhone or a Galaxy model. Physical AI won't build the hardware robots itself, but we plan to offer a platform that can train a variety of robots based on Digital Twin technology. It's like becoming a robot military academy.
Cho Ik-hwan, 48, SK Telecom vice president in charge of Physical AI, said in an interview with ChosunBiz on the 7th, Physical AI is a market that is just emerging, but within three years we aim for it to account for more than 10% of SK Telecom's AI B2B (business-to-business) sales.
SK Telecom is putting forward Digital Twin and a robot training platform as the core pillars of Physical AI. A Digital Twin is a technology that replicates real-world factories, logistics warehouses, and data centers in digital space and links real-time data to analyze, predict, and simulate operating conditions, and it is regarded as a next-generation industrial technology that can dramatically raise productivity and operational efficiency.
Cho has worked across Metaverse, Digital Twin, and Physical AI at SK Telecom. After starting a career in 2008 as a principal researcher in Samsung Electronics' Wireless Business Division and DMC Research Center, Cho joined SK Telecom in 2015 and oversaw the development of Augmented Reality (AR) and Virtual Reality (VR) platforms and technologies. After serving as AR/VR development team leader and head of Metaverse development, Cho has led the Digital Twin Lab since 2024, laying the groundwork for the Physical AI business. Since Nov. last year, Cho has headed the Physical AI division under SK Telecom's AI CIC.
Going forward, we intend to provide a Physical AI platform based on Digital Twin technology not only to traditional manufacturing such as semiconductors, logistics, automobiles, and shipbuilding, but also to AI data centers (DC), which have emerged as a new kind of manufacturing in the AI era. The following is a Q&A with Cho.
― It's been about a year since a dedicated Physical AI organization was established under AI CIC. What are the results and plans?
The first phase of the business, the Digital Twin phase, digitizes space, processes, and equipment and builds identical environments to virtually simulate new attempts with reduced risk to achieve production optimization. Over the past year, we've found clues that we can create value through this. The second phase is a business that trains robots based on Digital Twin technology. Think of it as a robot military academy. To train robots, you have to teach them what tasks to do, and collecting this data is quite difficult. We are generating data and training in a Digital Twin that replicates the real environment. Robotics training on top of the Digital Twin is the business area we are currently advancing. In the third phase, we aim to complete a virtuous cycle on top of an AI DC where, based on GPUs, data generation, training, application, and inference take place, and the newly collected data is trained so that everything operates agentically.
― What stage are you at now?
We are in phase two. We are at the stage of integrating individual technologies within manufacturing corporations into a single pipeline so customers can achieve productivity innovation through our Physical AI platform. The Digital Twin-based robot training platform has not yet been commercialized. We have verified it by conducting pilot tests with SK Group affiliates and external corporations. We build platforms within corporations so that Physical AI robots can replace tasks people used to do, such as moving and delivering goods, and once a robot is created and deployed to a customer, that's not the end; it must be advanced further. Around the second half of this year, we will explain the Digital Twin-based robot training platform to customers and move to full-fledged commercialization starting next year.
― You implemented SK hynix's semiconductor plant as a Digital Twin using Nvidia's Omniverse, a 3D virtual design platform for Physical AI.
Semiconductor processing is a highly challenging form of manufacturing, so SK Telecom's implementation of Digital Twin technology for that corporation is an important reference showing Physical AI applied to manufacturing. We expect strong demand for Physical AI in sectors such as semiconductors, logistics, automobiles, and shipbuilding, where processes are complex and production stoppage risks are high.
― Tell us about the plan to implement a Digital Twin for an AI DC.
AI DCs are preparing to apply Digital Twin and Robotics technologies to a new kind of manufacturing in the AI era. From heat management, workload management, and power management to facility and power layout at the construction stage of an AI DC, we plan to implement a Physical AI-based Digital Twin so that when it is completed, the AI factory is optimized in operation. We are discussing with Nvidia to apply a Physical AI Digital Twin to the project recently released with the government (building 15 GW of AI DCs by 2035). Because once an AI factory is built it is not easy to change, we need to design an optimized layout through Digital Twin technology from the moment equipment is installed.
― Why is a telecom company focusing on Physical AI?
Until now, telecom companies have provided telecom infrastructure to customers, but now they must provide intelligent infrastructure that processes AI, not just simple wired and wireless data transmission. Physical AI is an opportunity for telecom companies. Telecom companies have the strength of providing stable telecom infrastructure 365 days a year. And now, in the AI era, they will provide AI-related models, services, and infrastructure as well. Of course, there are many drawbacks. Because we do not directly build hardware robots, we fall short compared to specialized corporations. But we plan to keep this part open and collaborate. The robot platform SK Telecom is pursuing is not a form dependent on specific algorithms, technologies, or hardware, but a more flexible form that can apply a variety of ever-advancing technologies. We will support the use of various collaborative robots and humanoid robots on the platform, accept technologies we do not have through open innovation, collaborate, and also embrace robot foundation models owned by other corporations, such as Nvidia, ultimately providing a platform so customers can choose which models and robots are appropriate.
― Is it okay for growth not to build hardware while doing the Physical AI business?
Nvidia doesn't make hardware either. But we are also thinking it through. SK Telecom, and at the SK Group level as well, is actively investing in robot corporations and is reviewing additional investments. We always keep in mind the possibility of entering the robot hardware market directly. That is why we cannot categorically say the possibility (of entering the robot hardware market) is not open. For now, however, we intend to create synergies through collaboration and investment to make up for what we lack.
― Are you considering overseas markets as well?
We plan to build a track record for the Physical AI platform through SK Group manufacturing and then target overseas markets. We see Taiwan and Japan as appropriate first targets. Helping domestic manufacturing corporations expand overseas is also a future business model. In that case, the United States seems suitable.