AWS has identified South Korea as a market with high potential for leading the global physical artificial intelligence (AI) sector, thanks to its strengths in manufacturing, robotics, and semiconductors. While the full adoption rate of physical AI among domestic companies stands at just 6%, the number of firms conducting pilot programs and planning to implement such technologies is increasing. AWS believes that discovering clear use cases, such as autonomous robots and predictive maintenance, along with validating return on investment (ROI), will be key to expanding actual business applications.
On September 11, AWS held a press briefing at its office in Gangnam, Seoul, to discuss its report titled 'Realizing South Korea's AI Potential by 2026.' The report, conducted by global research firm Strand Partners, analyzed the current state of AI utilization among South Korean companies based on surveys of 1,000 business leaders and 1,000 general consumers.
According to the report, the AI utilization rate among South Korean companies has risen to 58%. However, there remains a gap in translating this utilization into actual product launches and ROI measurements. AWS also assessed companies' readiness and plans for adopting next-generation AI technologies, including agentic AI and physical AI.
Nick Bonstow, a partner at Strand Partners who participated in the briefing via video, noted that South Korea has a strong industrial foundation in manufacturing, robotics, and advanced technology, which presents a 'localized opportunity' for physical AI.
AWS's analysis indicates that physical AI is still in the testing and preparation stages rather than full-scale adoption. While only 6% of companies have fully adopted physical AI, 22% are currently conducting pilot programs, and 39% plan to implement it in the future, suggesting significant potential for expansion. The most frequently cited application areas include autonomous robots, predictive maintenance, human-machine collaboration, and warehousing and logistics.
Another challenge for commercializing physical AI is the acquisition of training data for robots. For instance, the South Korean startup Config has expanded a single captured data set into 88 scenes within five hours, rather than repeatedly collecting real-world data. Robots utilizing this method achieved a control success rate nine times higher in unfamiliar environments. Bonstow explained, 'Collecting real-world data continuously is time-consuming and costly.'
To expand physical AI into actual business applications, it is essential to identify clear use cases. During the Q&A session of the briefing, Bonstow emphasized, 'The key is to discover clear and compelling use cases.' He stated that creating real-world applications in areas with evident benefits, such as autonomous robots and predictive maintenance, and measuring ROI post-implementation are crucial for scaling pilot projects into full-fledged businesses.
Bonstow concluded by asserting that South Korea has ample potential to lead in physical AI on a global scale.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.