Tencent Cloud launched a data processing, model training and inference stack for robots and self-driving systems at the big data session of the 2026 Tencent Global Digital Ecosystem Summit on 30 October, Pandaily reported. The company also said it would release TCLake XPhysis, a platform intended to cover the wider physical AI engineering process.
Robots and self-driving systems produce large volumes of video and sensor data that must be prepared for repeated training and simulation. Tencent Cloud’s stack uses DataBuddy workflows and its AI DLC data lake compute service to connect those jobs. Xpark and Meson prepare and check the data, TCRay handles distributed training and inference, and TCLake manages data and AI assets. A problem found in simulation can then go back into data preparation for another training round. Tencent Cloud says the approach has cut data preparation from 15 days to two, increased GPU utilisation by nearly 200 percent and reduced average daily cost per GPU by 35 percent.
Driving a car could eventually draw on a self-driving system whose simulation failures have been fed back into training, if that repeated process improves its performance on the road.
Zhuoyu Technology is among the named users of the stack. BioMap uses Tencent Cloud’s EMR DIaaS platform and TCRay to process and train on multi-omics data for its virtual cell work. Tencent Cloud has also made a private deployment option available for DataBuddy, its enterprise data and AI workbench, for finance and government customers.
