As humanoid robots move into highly integrated systems and move from industrial applications to home environments, AI models are being deployed to meet the growing demand for back-end understanding and interaction capabilities, according to TrendForce’s latest research. It has been reported that training is becoming increasingly important. The global LLM market in robotics, including AI training and AIGC solutions, is expected to exceed USD 100 billion by 2028, at a CAGR of 48.2% from 2025 to 2028.
At CES 2025, NVIDIA announced the Cosmos platform, which integrates the World Foundation Model (WFM). This platform is expected to be a major driver of market growth.
The Cosmos platform is closely related to robot development, leveraging real-world and AI-generated data to build a digital twin environment for training. Benefits of this approach include cost savings, process simplification, customized training, and more realistic interaction simulations.
With Cosmos, NVIDIA aims to accelerate the development of general-purpose robotics technology, recognizing the evolution of AI from generative AI to agent AI and ultimately physical AI. This advancement is critical for self-driving cars and robots to advance the training process and achieve broader commercial deployment.
One of the major challenges in robot training is the simulation-reality gap, or the difference between the virtual training environment and real-world conditions. Deep learning and neural networks rely primarily on simulated data, which can lead to inefficient real-world performance due to lack of consistency with physics, sensor data, and actuator dynamics. There is a gender.
NVIDIA introduced the NaVILA model in December 2024 to fill this gap. This allows robots to understand natural language commands without prior training or mapping and navigate complex terrain using visual systems and LiDAR technology.
Currently, AI training and AIGC solutions account for 10% of the cost of humanoid robots, with the software side accounting for the largest cost share. Other cost factors include operating system (4%) and autonomous navigation (3%).
Several key companies in the NVIDIA ecosystem are driving advancements in robot simulation technology. Companies such as Siemens, Vention, and Hexagon specialize in robot simulation, while Rockwell Automation, PTC, and Cadence focus on computer-aided engineering (CAE) and reality capture. Additionally, Taiwanese startup MetAI provides innovative solutions to the ecosystem.
TrendForce highlights that NVIDIA’s Cosmos platform is poised to address training gap challenges and highlights the importance of AI large-scale model simulation. This development represents an important step towards the commercialization and integration of humanoid robots into everyday life.
Leading humanoid robot developers such as Figure AI and 1X are expected to be early adopters of the Cosmos platform, setting the stage for a new era of AI-driven robot innovation.
For more information about TrendForce’s robot reports and market data, click here or email our sales department at SR_MI@trendforce.com.
For more insights from TrendForce analysts on the latest technology industry news, trends and predictions, visit https://www.trendforce.com/news/.
Next article
European and Japanese IDMs strengthen collaboration with Chinese foundries to capture ‘China for China’ opportunities, TrendForce announces
