Caterpillar and CoreWeave are collaborating to accelerate the development of physical artificial intelligence (AI) for construction equipment, aiming to reduce the time required to train autonomous machines for complex job-site conditions.
The initiative addresses challenges in the construction industry, including declining productivity and a shortage of skilled machine operators. While Caterpillar has extensive experience with autonomous machinery in mining, construction sites present more unpredictable environments that require advanced AI systems capable of adapting to changing conditions.
The collaboration combines Caterpillar’s machinery expertise with CoreWeave’s high-performance computing infrastructure. Training autonomous excavators requires processing large volumes of camera footage, LiDAR data, machine telemetry and performance information, alongside extensive simulations and reinforcement learning.
Caterpillar’s digital ecosystem already contains approximately 18 petabytes of data collected from machines, dealers and customers. However, the company notes that training physical AI for construction requires significantly more data to understand real-world operating conditions.
Working with Nvidia, the companies are using AI models to automate the annotation and labelling of incoming field data. According to Caterpillar executive Brandon Hootman, this approach has reduced the feedback cycle between collecting field data and using it for simulation or training from months or weeks to hours.
The partnership highlights the growing role of AI infrastructure in advancing autonomous construction machinery and improving equipment development processes.
