A CPU, or central processing unit, is the general-purpose processor that runs a computer's operating system and its apps. It is built to do many different tasks, one after another, very fast. Every server, laptop and desktop has at least one. A processor is a chip that computes: a small, flat piece of silicon with billions of transistors, microscopic electrical switches.
A CPU is made of cores. Each core reads one instruction after another from memory and carries it out: add two numbers, compare them, fetch data, or jump to another part of the program. A laptop CPU has about 8 to 16 cores. A server CPU has 64 or more. Each core runs at a clock speed of about 3 GHz, which means about three billion steps every second. The set of instructions that a CPU understands is called its instruction set architecture (ISA). Most laptops and servers use the x86 instruction set. A growing share use Arm designs, which chip companies obtain through processor IP licensing.
Why not use GPUs for everything? Most software is full of decisions: if the password is wrong, stop; if the item is out of stock, show a message. Each step depends on the result of the step before. A GPU does the same sum on thousands of numbers at once, and it is slow at long chains of decisions. A CPU core is large because it is built for these chains. It guesses which way each decision will go and prepares the next instructions in advance.
Consider an online grocery app in Kolkata at 8 p.m., its busiest hour. Each order is a chain of small, different tasks. The server checks that the customer is logged in, reads the prices of 14 items from the database, checks stock at the nearest store, applies a coupon, works out the delivery fee and saves the order. A server with two 64-core CPUs can handle a few thousand such orders every second, with each core working through one request after another. The CPUs keep the data they are working on in DRAM, the server's main memory.
In an AI server, the CPU has a supporting job. It runs the operating system, loads data from storage and sends work to the GPUs. A slow CPU can leave expensive GPUs waiting. CPUs also help protect servers. Many contain a hardware root of trust, which checks that the first software to start has not been changed. Some can keep data encrypted even while it is being processed, which is called confidential computing.
Laptops and desktops bought by Indian offices use the same kinds of CPU, in smaller versions that use less power. Newer laptop CPUs add a small AI engine, the feature behind the AI PC. The leading CPU makers run some of their largest design centres in Bengaluru and Hyderabad. Engineers there design and test these chips, which are then made in factories abroad.