Design and analysis of shape-based analog computing (S-AC) circuits employing margin-propagation method are primary subjects of the current research. It explores key features of S-AC circuits, particularly their scaling potential, which is evaluated in precision, speed, and power efficiency compared with digital designs. Machine learning (ML) architectures with mathematical functions are employed to develop S-AC circuits. Input/output characteristics are mapped from a CMOS process for circuit simulations. S-AC-based neural network’s accuracy has no impact by temperature changes. When accuracy of fundamental S-AC process increases with several splines, it remains scalable. This paper also focuses on the design margin and shape analysis. The design parameter S and machine learning applications both factors are crucial to ensure that the system replicates the desired function form. Instead of using traditional design methods, S-AC design lets the user select the proto-shape-based on the application’s requirements and concentrate on obtaining the appropriate functional forms.