Solomon: Expanding AI Vision Module System to Reduce Humanoid Robot Data Learning

With advancements in humanoid robot technology, Solomon stated that this year it is focusing on reducing the data learning required for humanoid robots.

Solomon was invited to participate in the AI-themed performance presentation held by the Taiwan Stock Exchange this afternoon. Looking ahead to the generator and energy agency businesses, Solomon Chairman Chen Cheng-lung pointed out that this year's growth momentum comes from the construction applications of semiconductor fabs and artificial intelligence (AI) data centers. The expansion of global AI data centers, coupled with Taiwanese companies investing in overseas factories, has led to a shortage of generator sets.

Chen Cheng-lung stated that the order scale for generator sets in the first quarter of this year reached NT$2 billion to NT$2.5 billion, a significant increase from the average order scale of NT$1.5 billion in the past two years, indicating strong growth momentum. Orders are mainly from Taiwanese semiconductor manufacturers, and AI data center applications are also promising.

He pointed out that the delivery period from US original manufacturers has also been extended due to strong market demand. The average delivery period used to be about 6 to 8 months, but now it has been extended to 10 months or even 15 months.

Regarding the layout of robot applications, Chen Cheng-lung stated that Solomon continues to deploy robot AI vision software product applications. AI vision products are also used in non-robot fields such as AI defect detection, and the company is also deploying wearable devices and mobile AI camera applications.

Chen Cheng-lung pointed out that Solomon, based on AI vision software, integrates optical technology to develop modular system equipment, entering into robot inspection, system integration solutions, and AI inspection equipment.

Speaking of the progress in humanoid robot applications, Chen Cheng-lung stated that breakthroughs are not yet obvious. The actual application needs in factory production lines are different from entertainment performances. The current challenges faced by humanoid robots include insufficient stability of mechanical components and hardware systems, vision only being able to recognize at close range, and large, time-consuming, and costly physical data learning, making application difficult.

He pointed out that the balance of bipedal humanoid robots needs to be overcome. In addition, technologies such as battery life and hand dexterity in humanoid robots also need improvement. In the future, humanoid robots must enhance their autonomous capabilities to have the opportunity for commercial application.

Observing the proportion of product categories, according to data, the electromechanical energy business accounted for about 40% of Solomon's overall performance in 2025, automated products accounted for 29%, optoelectronic manufacturing accounted for 21%, and electronic channels and others accounted for about 10%.