Select Language:
Both robotics companies based in Hong Kong reported significant revenue growth last year as the commercialization of embodied intelligence picks up speed. Despite the rapid sales increases, both firms are still operating at a loss due to ongoing substantial investments in research and development.
One company recorded a revenue of approximately $71.4 million, reflecting a 32% increase from the previous year, while its net loss narrowed by roughly $1.6 million to about $12.2 million. The other firm announced revenue exceeding $290 million, a 53% rise, with a net loss decreasing by just over $370 million to around $790 million.
Their latest financial statements reveal that last year, one company launched several humanoid robots, along with quadruped and hexapod bionic robots. The other company boosted sales of full-size intelligent humanoid robots to nearly 1,080 units from only a few in the prior year.
This second company generated approximately $120 million (CNY820 million) from sales of their full-size intelligent humanoid robots and related solutions, marking a 22-fold increase year-over-year. This segment now constitutes 41% of their total revenue, making it their fastest-growing business area.
The first company entered the embodied intelligence sector relatively late, with last year’s revenue from these robots reaching about $3 million (CNY20 million), representing 4% of total revenue. However, this segment has more than quadrupled compared to the previous year.
Both companies are continuing to ramp up investment into R&D for embodied intelligence. One company’s R&D spending increased by roughly 60%, reaching about $17.6 million (CNY115 million), much of which was focused on embodied intelligence. The other company’s R&D costs increased by 6%, totaling about $77 million (CNY508 million), with $41 million (CNY270 million) dedicated to developing full-sized humanoid robots.
Heavy investment in embodied intelligence remains the primary reason for the first company’s ongoing losses. Removing over $6 million (CNY40 million) spent on R&D-specific to embodied intelligence, the company nearly broke even last year. For the second company, while increased sales of humanoid robots boosted revenue, accounts receivable remains high, exceeding 90% of annual revenue and reaching about $262 million (CNY1.8 billion), up 40% from the prior year.
In discussions following their financial disclosures, executives from both companies indicated that humanoid robots with embodied intelligence are likely to first be deployed in industrial and commercial contexts before making their way into household services.
One CEO noted, “Last year, research was the dominant market focus, but the next step will be to enter industrial and commercial sectors,” adding that robots’ locomotion capabilities have largely met practical thresholds. Future development will likely center on operational scenarios in manufacturing and interactive applications within homes and service environments.
They also mentioned that embodied intelligence robots are now capable of handling components in industrial settings, with several orders already secured. Beginning in the second quarter, some basic operational tasks are expected to be gradually implemented.
Once breakthroughs are achieved in industrial operational capabilities, the potential for a trillion-dollar market could emerge. The company anticipates shipping tens of thousands of quadruped robots within one to two years, aiming for about 10,000 units sold this year.
The CEOs highlighted ongoing challenges related to data collection and model development. One mentioned that last year, the industry accumulated fewer than 30,000 hours of valuable data. He noted that hundreds of thousands of hours of less-valuable data only support pre-training and hinder efforts to improve operational accuracy and generalization. The company is exploring data feedback mechanisms and aims to gather scenario-specific data through their annual production of 100,000 robotic arms.
Regarding model development, efforts to combine vision-language-action frameworks with reinforcement learning were already underway last year. The ultimate goal is to develop a comprehensive world model, which will require tens of millions or even over 100 million data points—an endeavor expected to take three to five years.
The management of the other firm also confirmed that they are advancing validation of their vision-language-action pathways and developing robust world models. Their R&D budget is projected to increase to roughly $103 million (CNY700 million) by 2026, with a focus on foundational embodied AI models, world models, and product enhancements.
ChatGPT
Perplexity
Gemini AI
Grok AI
