Shared from twixb · roboticsandautomationnews.com

Dexterity says its Foresight world model unlocks physical AI performance on Nvidia hardware

roboticsandautomationnews.com·Mar 22, 2026

Dexterity.ai has significantly enhanced its Foresight AI model's performance on Nvidia hardware, achieving a 17x speedup in processing cycles and a 32x increase in data utilization, which enables complex robotic tasks like autonomous truck loading. This advancement was highlighted by Nvidia at GTC 2026 and showcased at FedEx's Investor Day, indicating a promising future for enterprise-scale physical AI applications.

Dexterity's significant improvement in processing time and data utilization using Nvidia L4 GPUs and TensorRT highlights a crucial advancement in industrial robotics. By reducing perception pipeline time from 1,508 ms to 90 ms and increasing data processing 32x, Dexterity's Foresight model allows for faster and more informed decision-making, enhancing operations like autonomous truck loading. For robotics professionals, this underscores the importance of leveraging cutting-edge hardware to optimize AI applications in complex industrial environments.

Powered by twixb

Want more content like this?

twixb tracks your favorite blogs and social media, filters by keywords, and delivers personalized key learnings — straight to your inbox.