“It’s not about robots in labs anymore someone just bought a thousand of them.” This has recently become the message that has resonated throughout the world’s robotics community, as received recently from the Galaxy Universal company with their all-time record-breaking order for the supply of 1,000 units of the company’s G1 series humanoid, at a price that broke all price records at 700 million US dollars. This corresponds to 700,000 RMB per humanoid, which in turn positions the company at the top of the humanoid robot market, surpassing even the annual revenues of the largest companies in the China robotics market.
This acquisition comes in the wake of the company’s recent funding, which has broken records by exceeding $300 million, giving the company a valuation of over $3 billion. This is the first time that such figures have been attained within the same category of the humanoid robotics challenge in the Chinese market. The company has financing from the state-owned financial institutions, including the financial wings of China Mobile ‘Chain-Leader Fund, CICC Capital, Chinese Academy of Sciences Fund, Suzhou Venture Capital, and other financial institutions. It also has financing from the financial institutions of the Middle Eastern countries. Singapore investments also form a component of the company’s financing. It also receives financing from the sovereign wealth funds, whereby the funds primarily invest in the company’s robotics segment, which primarily involves AI. Gulf countries, including the Vision 2030 initiative of Saudi Arabia and the AI Strategy 2031 of the UAE, invest heavily in the embodied AIs.
In techie words, the technology underlying the G1 series robots in the Galaxy Universal is grounded on two major innovations: NavFoM and DexNDM. NavFoM is the first ever large-scale globe-spanning navigation model for cross-ontology and round-the-clock fully autonomous navigation. Unlike the traditional path-following approaches that rely on static maps for path routing, NavFoM supports hour-long range navigation with dynamic obstacle avoidance even in complex industrial and retail environments. This requires innovation in autonomous navigation algorithms that include multimodal sensor fusion, precise environment mapping, and reinforcement learning-based adaptive control, as seen in Physical AI.
DexNDM, the hand neurodynamics model designed by the authors, focuses on one of the most difficult tasks for humanoid robots delicate manipulation tasks in uncertain physical interactions. This is because traditional controlling algorithms through trajectories become futile when such controlling algorithms are used for elastic material interactions, as well as when controlling algorithms have to deal with manipulanda that require nuanced control of forces. However, the approach takes into consideration the latest breakthroughs in the development of foundation models for neuro-dexterous manipulation.
Galaxy Universal’s learning process follows to a significant extent the ‘Sim-to-real’ approaches, creating virtual environments to validate the learning of policies in a virtual environment prior to use in the real world. Though very helpful in preventing problems such as safety and enabling fast learning in the process of training learning in a significant manner, the approach experiences the same challenges in depicting reality in a virtual environment. An approach such as in the CSAIL’s RialTo system would be applicable in reducing the challenges of the ‘sim-to-real’ reality gap. Relevant data from the industrial customer base is also very enlightening.
Apart from the notable agreement, Galaxy Universal has managed to sign agreements with multi-unit clients such as CATL, Bosch Group, Toyota Motor, Hyundai Motor, BAIC, SAIC, Geely Zeekr, and Great Wall Motor. These agreements highlight the direction being taken by firms such as Hyundai’s Boston Dynamics in the development of humanoid robots such as Atlas designed for mass deployment in the factory setting. Finally, in these cases, the robots are moving from the laboratory to the factory setting where they would be capable of carrying out tasks like assembling, inspecting, and transporting material using human-like limbs.
On a strategic level, the involvement of Middle Eastern funds in the funding of Galaxy Universal proves, in addition to other aspects, that the value embodied in the high value embodied intelligence assets is in high demand by these sovereign funds in the Gulf States for geoeconomic needs, specifically for the purpose of developing the corresponding components in the said economies into world-class tech hubs. The Public Investment Fund in Saudi Arabia, the Mubadala Fund in the UAE, and Singaporean firms are not passive in receiving investments. Chinese robotic firms are being diversified, even protected against domestic constraints, as a result of this funding injection. The presence of Galaxy Universal also indicates the acknowledgment that the “real economy” concept factor in China’s master plan in the future includes value embodied in AI.
By integrating highly advanced AI models within physical infrastructures, China seeks to address concerns in its demographics, improve the productivity of its manufacture sector, as well as aim to become world-class in next-generation export-oriented industrial automation services. The existence of the Suzhou industrialization bases as well as the “3 centers +1 base” R&D framework makes it ready to manufacture on a massive scale once it reaches maturity. On the level of tech investors, robotic engineers, and manufacturers, the G1 purchase order amounting to $700 million is more than just a sales success since it proves that humanoid robots, until now confined to exhibitions, pilot projects, & pilot implementations, are finally about to enter into the operational use level in a sustainable fashion. The coincidence of innovation in navigation & manipulation, high-quality sim-to-real transfer boundaries being cleared, & capital alignment on a single stage dictates the start of an era in which size, certainty, & ecosystems shall provide the foundations for differentiation in the adoption level in the industry’s use level in AI satisfaction.
