Southeast Asia Carves Its Own Niche in the Global Market with Physical AI

Southeast Asia may be far from the premier Artificial Intelligence (AI) innovation hubs of the United States and China, but in its own way, the region is rapidly building out AI research and innovation centers to attract investment, cultivate talent, and build solutions that leverage AI and Machine Learning (ML):

• Indonesia: In July 2025, Indonesia’s Ministry of Communication and Digital announced the launch of its Indonesia AI Center of Excellence. The center is designed to support ethics, infrastructure, data governance, investment, talent, and Research and Development (R&D). Indonesia's effort in AI are being further stimulated by the foundation of the NVAITC Joint Center at Universitas Gadjah Mada (UGM) in August 2026.
• Malaysia: The National AI Centre of Excellence is being established. Microsoft is playing a prominent role in establishing the national AI Centre of Excellence (COE). The institute will work closely with AI Malaysia Berhad and the Malaysia AI Safety Institute to spearhead AI R&D, commercialization, and innovation.
• Philippines: In February 2026, the Philippines's government established the National Artificial Intelligence Center for Research and Innovation (NAICRI) as the country’s central institutional anchor for AI research, advanced computing, and innovation.
• Thailand: In August 2025, Thailand’s NSTDA announced preparations to establish its AI Thailand Hub, which is being led by Chulalongkorn University and ETDA.
• Vietnam: In March 2021, Vietnam’s government launched its first ever AI center of excellence, the HUST–Naver AI Center. It is a jointly owned and operated international research center between Hanoi University of Science and Technology and the Korean software developer, Naver. In December 2024, NVIDIA announced that it was collaborating with the Vietnamese government to establish a Vietnam R&D Center focused on AI to serve the needs of Vietnamese startups, universities, government agencies, and researchers.
• Singapore: The country has been a very proactive regional evangelist of R&D in AI, with multiple AI innovation hubs: AI Singapore (AISG) coordinates and stimulates a range of AI programs, bringing together research institutions to grow talent, build the ecosystem, and drive national AI capabilities. Notable key stakeholders include the NUS AI ecosystem (including the NUS Artificial Intelligence Institute (NAII), A*STAR’s AI research stack (Centre for Frontier AI Research (CFAR), the AI-heavy work at A*STAR IHPC, and A*STAR I²R) and AIMfg (Sectoral AI Centre of Excellence for Manufacturing), which was jointly developed by MTI and A*STAR in September 2024.

These centers of excellence are working on a wide range of AI initiatives—including pharmacological drug discovery and genomics, AI for smart cities, and digital assurance—with a prominent cluster of use cases being tested, prototyped, and deployed in the “Physical AI” domain. The region may not be able to match the R&D spend of the United States or China, but its enterprises and governments are keen to ensure the Return on Investment (ROI) from Physical AI is harnessed.

Across the region, a range of physical AI use cases are being deployed. Many of these are on the implementation/inference side of the market, but these stakeholders are striving to enhance the training/brain aspect for future gains:

• Autonomous Security Patrols: In Singapore, Certis is using AI-enabled autonomous robots for security patrols and inspection workflows in complex environments. The goal is to augment frontline manpower, improve patrol consistency, and strengthen response coverage.
• Residential Autonomous Shuttles and Delivery Robots: In Singapore, since April 2026, Grab says its Ai.R autonomous shuttle has been operating in the Punggol residential neighborhood.  Furthermore, the company has autonomous delivery robots operating in the region. The shuttle had logged ">40,000 km" and "more than 3,800 public rides" without incident.
• Port Logistics Orchestration: The PSA (Singapore) has been trialing Autonomous Prime Movers (aPMs) for horizontal container transport. The trial has been running 24/7. PSA also has announced an AI-based transport planning platform, OptETruck, which uses AI scheduling and asset pooling to reduce empty trips for haulers.
• Smart Hospitals: Siriraj Hospital in Bangkok, Thailand, has deployed "5G-connected self-driving vehicles" for contactless delivery of drugs and medical supplies. Separately, Thai healthcare deployments are expanding into bedside support: Ichitan’s Dinsaw Mini Robot can provide symptom monitoring, medication reminders, and remote communication.
• Smart Agriculture AI Done Support: In Malaysia, Aonic has reported it has deployed autonomous drones, AI-driven plantation mapping and software, for precision spraying and site monitoring.
• Drone Inspection of Infrastructure: Singapore-based Garuda Robotics offers a drone infrastructure system for remote surveillance and critical-asset inspection in Malaysia and Singapore. The solution uses AI to detect faults such as cracks, corrosion, and rust. Garuda reports the system can be "2X more efficient than human patrolling."
• Smart Mining and AI Safety Systems: The Indonesian Ministry of Energy and Mineral Resources has reported that Berau Coal’s AI-integrated remote monitoring utilizes a combination of Closed-Circuit Television (CCTV) and analytics to monitor human and heavy-equipment movement.
• Cloud-Native, Robotics-Ready Manufacturing: In Vietnam, Kärcher’s manufacturing facility in Vietnam was launched as a cloud-native factory template running from within the "AWS Singapore region." The deployment is a combination of automated cloud, plant operations, and an increasing number of autonomous robots.

Based on this scan of the Southeast Asian AI and Physical AI landscape, R&D commitment is building around these operational areas:

• The deployment of autonomy in defined demarcated areas such as hospitals, business districts, airports, and seaports.
• Hardware and ambient condition inspection in hazardous or hard-to-reach environments such as in the energy, telecoms, utilities, and plantations sectors.
• Logistics and yard automation, where highly repetitive tasks or movement of goods are needed, such as in seaports, warehouses, air cargo, and industrial parks.
• Labor task acceleration and mitigation in service-intensive sectors such as security, healthcare, and operational facilities.
• Ensuring personnel safety and operational performance of equipment in heavy industry sectors such as mining and large-scale industries.

What will the Total Cost of Ownership (TCO) and ROI from Physical AI be? Unfortunately, the answer is not clearcut and depends on multiple variables. Based on ABI Research’s ongoing research, ROI will depend more on the amount of upfront Capital Expenditure (CAPEX) required and efficiency gains versus the Operational Expenditure (OPEX) profile. This is quite different from enterprise Agentic AI in the cloud. Using Agentic AI, AI implementations break even and start generating positive ROI between year 1 and year 2, driven by efficiency gains and associated human capital reductions. For cloud-based models, AI agent running costs make up a much higher proportion of 5-year TCO than at the edge. Human resources-related use cases create the highest year-over-year ROI.

These are exciting times for Physical AI in Southeast Asia, but it will also be critical for these Physical AI deployments to be able to “scale.” Physical AI changes the scale and purpose of connectivity. As AI expands beyond factories into ports, transport, energy, logistics, cities, and remote operations, its value comes from coordinating distributed systems across large, dynamic environments. That will require connectivity that is secure, resilient, programmable, and pervasive. Unlike legacy communications networks built mainly for human use, Physical AI will depend on machine-to-machine coordination, greater sensor uplink traffic, stricter latency, higher resilience, and stronger security. Physical AI operators will use fiber-optic, satellite, and Wi-Fi, but 5G-Advanced, perhaps even 6G architecture and coverage, will have a role to play.