Samsung Electronics is overhauling its flagship Global Research Outreach (GRO) program around physical artificial intelligence (AI), introducing new research areas including biosignal-based technologies and photonics.
The company is shifting GRO's focus away from semiconductor manufacturing processes toward securing next-generation technologies such as AI, robotics, and human-machine interfaces (HMI).
According to industry sources on Aug. 9, Samsung has placed AI-related fields at the forefront of this year's GRO research agenda, including AI for science, physical AI computing, and AI system architecture.
Launched in 2009, GRO is Samsung's annual global research collaboration program, which invites and supports research proposals from university researchers worldwide. From 2022 to 2025, its primary themes included semiconductor materials, processes, metrology, equipment, new computing, and machine intelligence. Starting this year, however, the program has pivoted toward securing technologies expected to be critical to the future.
The move is seen as aligned with Samsung's broader strategic shift. This year, the company established the RX (Robotics eXperience) Business Promotion Office directly under the CEO, while investing heavily in physical AI through initiatives including a robot data factory and infrastructure for mass-producing humanoid robots in Gumi.
As a result, research into real-time processing in edge environments—where computing takes place directly on devices rather than in remote data centers—has become increasingly important.
One notable addition to this year's GRO is human-machine interface technology. Rather than relying on keyboards or voice commands, HMI technologies are designed to identify a user's intent by reading bodily signals, such as brainwaves or muscle activity, in real time.
HMI is particularly useful in industrial settings where conventional input devices such as keyboards and mice are difficult to use. As robots increasingly work alongside people, there is a growing need for technologies that can interpret biosignals and understand human intent.
The ultimate objective is to enable robots to recognize a person's condition and respond accordingly, even without an explicit command.
Samsung has also expanded its focus on physical AI computing. New research topics include embodied intelligence built on vision-language-action (VLA) models, which allow AI systems to perceive their surroundings, interpret information, and take action.
The research will explore the co-design of computing architectures, hardware, and software needed to run AI models on edge devices with limited power and space.
In AI system architecture, Samsung has added research on photonic fabrics—ultra-high-speed interconnect networks that use light—as well as benchmarks for measuring how effectively AI agents perform when coordinating multiple tasks.
The goal is to establish standards that can quantitatively analyze bottlenecks involving CPUs, memory, input/output (I/O), and GPUs during AI-agent execution.
“Samsung's reorganization of GRO around AI and robotics signals where it is placing priority in securing future core technologies,” an industry source said. “It reflects a strategic decision to work with universities in laying the technological foundation that will determine competitiveness three to five years from now.”
