DOBOT has released its interim results for the first half of 2026, revealing robust financial performance. During the period, the company generated approximately RMB 320 million in revenue, a year-on-year increase of 106.6%. Gross profit reached roughly RMB 150 million, up about 100% year-on-year, with the gross margin holding steady at 47.4%. Furthermore, in the first half of 2026, DOBOT invested around RMB 100 million in research and development, marking a 148.4% increase year-on-year and representing over 30% of total revenue.

More striking than the doubling revenue growth is the qualitative shift in the revenue mix: revenue from its embodied AI business surpassed RMB 45 million, growing over 20-fold year-on-year and accounting for nearly 15% of total revenue. Data shows that DOBOT has expanded its embodied AI customer base to 231 clients, with a pipeline that fully covers all stages of commercial conversion, from contract signing and sample validation to small-batch delivery and large-scale deployment. Among these, nearly 100 customers are in industrial manufacturing, positioning DOBOT as one of the industry leaders with the broadest industrial client coverage.

In terms of application scenarios, DOBOT's embodied AI products have been deployed at end-user facilities such as Valeo Group, Leapmotor, and KEBODA, operating continuously in specific processes including material handling, loading and unloading, quality inspection, and precision assembly across sectors like automotive and new energy, 3C, equipment manufacturing, and daily chemicals. From industrial collaborative arms to full-size humanoid robots, DOBOT has built a multi-form embodied AI product matrix, placing its overall commercialization progress in the industry's top tier.

The financial report indicates that DOBOT's revenue from industrial manufacturing reached approximately RMB 200 million in the first half, up 142.4% year-on-year and accounting for 62.8% of product sales revenue, making it the largest revenue source. Notably, in May 2026, DOBOT launched its self-developed world action model, the "Kongyi Dobot WAM Embodied Large Model." On the LIBERO standard benchmark for embodied intelligence, this model achieved an average completion rate of 99.25% across four standard task suites, surpassing all mainstream public models with published LIBERO results as of 2026. Additionally, the Kongyi Dobot WAM model topped both the Set of Marks and Natural Language tracks in the UNOBench Challenge simultaneously.