In this paper, we present a new approach for realistic worst-case analysis of VLSI circuit performances and a novel methodology for circuit performance optimization. Circuit performance measures are modeled as response surfaces of the designable and uncontrollable (noise) parameters. Worst-case analysis proceeds by first computing the worst-case circuit performance value and then determining the worst-case noise parameter values by solving a nonlinear programming problem. A new circuit optimization technique is developed to find an optimal design point at which all of the circuit specifications are met under worst-case conditions. This worst-case design optimization method is formulated as a constrained multi-criteria optimization. The methodologies described in this paper are applied to several VLSI circuits to demonstrate their accuracy and efficiency. Keywords Worst-case analysis, worst-case design optimization. I. Introduction I NEVITABLE fluctuations in the manufacturing proces...