Robot LLM Market Expected to Surpass $100 Billion, NVIDIA's WFM Platform Leading Growth
According to research by TrendForce, as humanoid robots evolve toward highly integrated systems and gradually expand from industrial environments to home life, the importance of front-end AI model training becomes increasingly evident. This is crucial to meet the growing demands for understanding and interaction in the backend.
By 2028, the global market for large-scale language models (LLMs) in robotics, including AI training and AIGC solutions, is projected to exceed $100 billion. The compound annual growth rate (CAGR) from 2025 to 2028 is expected to reach 48.2%. NVIDIA’s Cosmos platform, set to launch at CES 2025, is expected to be a key driving force behind this market growth.
According to NEPCON China 2025, it was noted that the Cosmos platform is closely linked to the development of robotics. By utilizing real-world data and AI-generated data, the platform builds digital twin environments for training. This enables cost reduction, streamlined processes, and customized training, while also improving the realism of robot-environment interaction simulations.
NVIDIA's launch of this product is driven by two main considerations. On one hand, it aims to accelerate the development of general-purpose robotic technologies. On the other, NVIDIA believes the path of AI development has shifted from generative AI to agent-based AI. The next phase will focus on using physical AI technologies to enhance the training of autonomous vehicles and robots, laying the foundation for their wider adoption and commercialization.
TrendForce points out that the gap between simulation and reality is a major challenge in robot training. Since deep learning and neural networks are typically trained in virtual environments, robots may overlook physical characteristics, sensor information, or actuator dynamics in real-world scenarios, thus reducing their performance. Therefore, tools that strengthen the simulation of real-world conditions are essential. NVIDIA’s NaVILA model, released in December 2024, allows robots to adapt to complex terrain and perform path planning using only natural language commands, a visual system, and LiDAR—without the need for training or maps.
NEPCON China 2025 also highlighted that software-based AI training and AIGC solutions currently account for 10% of the cost of humanoid robots, with operating systems contributing around 4%, and autonomous navigation at 3%. Within NVIDIA's ecosystem, companies like Siemens, Vention, and Hexagon are actively investing in robotic simulation, while Rockwell Automation, PTC, and Cadence continue to focus on computer-aided engineering and real-world capture. Taiwanese startup MetAI is also part of this ecosystem.
TrendForce believes NVIDIA's Cosmos platform is set to further address the training gap in robotics, emphasizing the importance of AI model simulations. This also marks an important step forward for humanoid robots in terms of practical commercialization and daily deployment. Notable humanoid robot developers like Figure AI and 1X are expected to be among the first users of the platform.
Pre-register for NEPCON China 2025: https://ali2.infosalons.com.cn/reg/NEPCON25SH/registeren/login?type=WE9ZIC
Join us at NEPCON China 2025 April 22-24, 2025, for an unparalleled opportunity to expand your network, engage in direct business matchmaking, and access innovative technologies that will shape the future of electronics manufacturing!
If interested in exhibiting at NEPCON China, please contact
Ms. Julia Gu
Tel: +86 21-2231-7010
Email: [email protected]
If interested in visiting NEPCON China, please contact
Mr. Walden Li
Email: [email protected], [email protected]
Mobile/WhatsApp/WeChat: +86 136-5125-1335
For TAP service, please contact:
Mr. Alan Li
Email: [email protected]
For media partnerships, please contact:
Ms. Yuri Yu
E-mail: [email protected],[email protected]
Ms. Chopin Mo
E-mail: [email protected]
Follow us on:
Facebook: @NEPCONASIA
Twitter: @Nepcon_Official
LinkedIn: @NEPCON ASIA