In Healthcare 4.0, AI-driven telesurgery systems offer the potential for high-quality, real-time surgical services to remote locations. Achieving ultra-low latency is crucial for these systems. Traditional Internet-based telesurgery faces high latency, which fails to meet the stringent demands of Healthcare 4.0. This paper presents a telesurgery system architecture utilizing the Tactile Internet. We employ network calculus to model the system’s ultra-low-latency performance. Additionally, we propose a telesurgical force feedback prediction model based on a robot behavior bidirectional long short-term memory network (RB-BiLSTM). This model uses the upper bound of end-to-end delay as a threshold to predict the robotic arm’s next action, enabling precise compensation for feedback delays and ensuring timely haptic feedback. Experimental results show that our model effectively keeps the surgeon updated on the patient’s status, thus enhancing the overall telesurgery process.