Chip design is the stage in which a chip exists only as code and drawings, before any factory makes it. It covers the ready-made designs that a chip company licenses, the software in which its engineers describe and check the chip, and the circuit board on which the finished chip will sit.
A chip company seldom designs every block of a new chip from scratch. A block is one part of the chip, such as the processor or the graphics engine. The processor core usually comes through processor IP licensing. The company pays a fee to start using the design, and later a small royalty on every chip it ships. The graphics engine and the AI engine come the same way, as GPU and NPU IP. The page on semiconductor IP licensing explains these two payments.
The company's own engineers then design the rest in electronic design automation (EDA) software. They write the circuits as code, run the design in simulation to prove that it works, and lay out billions of transistors. A transistor is a tiny electrical switch. Most of the time goes into checking. A mistake found after the chip is made means a new set of masks and another trip through the factory, which can cost months and tens of crores of rupees.
A chip is useless on its own. It must be soldered onto a printed circuit board, or PCB, the flat green board whose copper tracks carry power and signals between chips. Engineers lay out that board in separate, simpler PCB design software. Before anything is built, signal and thermal simulation checks that signals arrive cleanly through the chip's package and the board, and that no part gets too hot.
Consider a team of 300 engineers working on a chip for wireless earbuds. They may spend 18 months in this zone and never touch silicon. At the end, they send one set of layout files to a semiconductor foundry, and the work moves to fabrication and test. Bengaluru, Hyderabad and Noida hold many such teams, which design chips for companies around the world.