Taipei—Korea IT Times Reporting Team
COMPUTEX 2026 was not simply another IT exhibition. The message running through this year’s show was clear: the center of gravity in AI is moving from cloud-based generative models toward the physical world.
Held in Taipei from June 2 to 5 under the theme “AI Together,” COMPUTEX 2026 brought together more than 1,500 companies from 33 countries across 6,000 booths. The scale confirmed its position as Asia’s leading platform for AI, robotics, semiconductors and smart manufacturing.
Across the exhibition halls, one trend was hard to miss. AI is no longer confined to screens. It is beginning to operate factories, control robots and make decisions inside industrial environments.
NVIDIA’s Blueprint for Physical AI
No company had a larger presence at COMPUTEX 2026 than NVIDIA.
Jensen Huang’s keynote and the GTC Taipei sessions laid out NVIDIA’s industrial strategy for the Physical AI era. The company introduced the GR00T N1.7 humanoid foundation model, the Blackwell-based Jetson Thor edge AI platform and the Isaac GR00T Reference Humanoid Robot, setting a new benchmark for the robotics industry.
The reference design, combining a Unitree H2 Plus chassis, Sharpa tactile five-finger hand, Jetson Thor and GR00T software, was seen as an attempt to narrow the gap between research prototypes and commercial deployment.
Robotics industry officials at the venue said companies are moving beyond the stage of building AI from scratch for their robots. The new question, they said, is which layer they can differentiate within NVIDIA’s ecosystem.
The fact that leading humanoid players such as Agility Robotics, Figure, 1X, AGIBOT and NEURA Robotics are already adopting or testing GR00T-based technologies suggests that NVIDIA is becoming a de facto platform operator for the robotics industry.
AI Moves Into the Factory
Another clear shift was visible across Hall 1’s AI Computing & Tech Zone and Hall 2’s Industrial IoT & Embedded Systems Zone: AI is moving directly into industrial sites.
Aetina’s Physical AI and agentic AI demonstrations, AAEON’s DEEPX NPU-based edge AI platform, and industrial vision and control solutions from ADLINK and ARBOR were no longer positioned as laboratory technologies. Most were aimed at real manufacturing lines, logistics systems and smart factory operations.
The focus of the AI industry is beginning to move away from large language model performance alone and toward measurable industrial productivity.
Taiwanese companies were particularly active in combining AI with manufacturing supply chains. This approach was closely aligned with the broader Physical AI trend seen across COMPUTEX. Taiwan is increasingly treating AI not as a standalone software sector, but as a means to strengthen manufacturing competitiveness.
Startups Enter the Commercialization Race
InnoVEX also looked different this year.
With about 500 startups participating, InnoVEX functioned less as a display area and more as a commercialization hub connecting startups with investors, corporations and platform companies.
Programs such as the Top 15 Pitch Contest, global investor meetings and the Kyushu Global Innovation Pitch from Japan were structured to connect startups with real customers and capital.
Notable companies including RLWRLD, DIDEN ROBOTICS and Avalanche Computing shared one common feature: many were already linked to NVIDIA Inception or global accelerator programs. The scene showed that ecosystem access, not technology alone, is becoming a competitive asset.
Government Support Becomes Part of the Competition
National competition in startup support was also visible across COMPUTEX 2026.
Nine countries, including Korea, Canada, France, Israel and Japan, operated national pavilions at InnoVEX to support their startups and SMEs in overseas expansion.
Canada’s Ontario model drew particular attention. Rather than simply offering shared booth space, the program connected selected companies with market briefings, export consultations and business matching. It was seen as a practical model for turning exhibition participation into business outcomes.
Korea also showed strength through its experience in large-scale pavilion operations and a high level of company participation. Even so, several observers noted that Korea will need a more focused and strategic support model going forward.
Remaining Challenges
Several issues also surfaced. Jetson Thor developer kits and production modules have already entered the supply stage, but some distribution channels are showing signs of pressure as demand increases. After the announcement of the GR00T reference design, supply-chain management has become a more important variable.
There was also criticism that some of the many forums and sessions remained focused on introducing new technologies rather than sharing concrete commercialization strategies or operational experience from the field.
Korean companies still face positioning challenges. Although the number of participating firms was meaningful, many remained focused on components, embedded systems and hardware. Relatively few were directly addressing upper-layer AI technologies such as GR00T or foundation models.
What It Means for Korea
COMPUTEX 2026 offered Korea both opportunity and warning.
The opportunity is real. As NVIDIA opens GR00T and connects its ecosystem with Hugging Face and LeRobot, the entry barrier for Korean robotics and AI startups has become lower than in the past. The fact that companies such as DIDEN ROBOTICS are linking with global platforms is a positive signal.
But competition is intensifying. As Ontario’s model shows, exhibition support alone is no longer enough. Korea needs long-term programs that move from targeted company selection to focused support, post-event matching and global partnership development.
At the same time, Korea must develop datasets and local AI models optimized for its own manufacturing and service environments. The task is to use global platforms while building strengths that cannot be easily replicated elsewhere.
The Starting Point of the Physical AI Era
COMPUTEX 2026 was one of the first major exhibitions to show AI moving beyond polished demonstrations and into real industrial environments.
NVIDIA’s full-stack strategy, combined with the rapid response of global partners, suggests that 2026 and 2027 could become a turning point for the commercialization of humanoids and Physical AI.
Korea now faces a strategic choice. It can take part more actively in the global ecosystem while building differentiated solutions for its own industrial environment, or it can remain a late participant in a platform-led competition.
COMPUTEX 2026 already offered a direction to that question.
“AI Together” is beginning to lead to “Physical AI Together.” COMPUTEX 2026 has ended, but the Physical AI era is only just beginning.
Editorial Contact: Korea IT Times Technology Team (monica@koreaittimes.com | info@koreaittimes.com)
