AI and AC for silicon leadership

Chip manufacturing gets more complex every year, and fundamental physics dampens our very real enthusiasm for continued scaling. The solutions to this problem often involve more precise modeling and intricate optimization strategies, across multiple modules in the Fab. Whatever may be the “law” that best describes the plodding state of scaling today, it is clear that the law that seems to govern the annual increase in Fab computation has an exponent greater than Moore’s Law ever did. Accelerated Computing (AC) helps solve this problem by drastically reducing the cost and energy required for these computations, thereby making possible previously prohibitive techniques and precision. On top of that, Artificial Intelligence (AI) methods, with the dramatic rise in their ability to describe the physical world, have opened the door to tackling age-old problems in the Fab with a new toolbox. This talk will describe the problem, some solutions, and give examples of how AC and AI are moving into production in advanced Fabs.