The global robot AI chipset market size is expected to reach $866 million in 2028, which will help GenAI's popularity in robotics, according to a report by market research firm Omdia.

Omdia notes that since Google launched its transformer RT-1 for robotics applications in 2022, multiple enterprises have made a big push for GenAI to be widely used in robotics. In addition to Google, companies like Meta, OpenAI, and Toyota are trialing or testing various foundational models in their robotics applications. Chinese service robot vendors, such as CloudMinds and OrionStar, have successfully developed their own basic models and plan to integrate these models with client software systems.

However, GenAI is resource intensive technology. In most industries, GenAI is typically deployed in the cloud as models require large clusters of graphics processing units (Gpus) for training and inference. At the opposite end of the spectrum, robots prefer local processing because the tasks and business-critical applications in which they participate often prioritize real-time control and ultra-low latency responses.

Although NVIDIA's Gpus remain the AI chipset architecture of choice for cloud infrastructure and robotics, said Lian Jie Su, principal analyst for applied Intelligence at Omdia, But non-GPU vendors such as Qualcomm, Intel, and AMD have launched AI System-on-chip (SoC) or dedicated AI chipsets for dedevice robotic applications such as machine vision, navigation and mapping, and functional safety.

Encouragingly, the widespread adoption of GenAI has also contributed to the popularity of humanoid robots. Humanoid robots are the closest robots to human form, so many robotics experts believe their combination with humanlike GenAI would be a match made in heaven. In this wave, companies such as Agility Robotics, Boston Dynamics, Figure, Tesla, Fourier Intelligence, Datamari, and Ubiselect have launched a variety of humanoid robots for industrial and service applications. However, the technology is still in its infancy and is unlikely to be widely deployed in the next five years. Automated Guided Vehicles (AGVs) and Autonomous Mobile robots (AMR) are still more mature forms of application for GenAI.

Instead of focusing on hype, the industry should focus on building its data and technology base, Mr Su said. For their part, robot vendors need to extend the capabilities of low-power GenAI in robots through various model optimization techniques, emphasize real-time control and performance, and become more adept at leveraging the advantages brought about by the convergence of computation and connectivity. For robot users, the development of domain-specific GenAI models together with rigorous scrutiny of ethics, security, safety, and performance will greatly facilitate the adoption of GenAI-enabled robots.