Silicon Whisperers: Improving Test Quality and Cost in the Age of Generative AI

The relentless scaling of process nodes, emergence of post-silicon device technologies, and stringent quality demands of automotive and medical markets have pushed integrated-circuit (IC) testing to an economic inflection point: manufacturing tests for advanced SoCs already account for nearly one-third of total silicon cost, yet latent-defect escapes remain an industry concern. In this survey, we analyze how large language models (LLMs) and broader generative AI techniques can be leveraged to break the long-standing quality-versus-cost trade-off in test. We discuss how generative AI is transitioning from pilot studies to factory-scale impact, positioning the semiconductor industry to deliver higher-quality products at lower test cost over the coming decade.