Class A—This keeps the original addressing structure; First octet contained the network, all other bits describe the node. To a machine, a class A address was easy to identify; the first bit of the address was always 0. That meant that only the remaining 7 bits were left to define a class A network. 7-bits yield 128 possibilities. Since digital systems tend to start at zero instead of one, class A network addresses have IP addresses where the first octet has a range between 0 -127. That leaves 16,777,216 possible addresses on each of those networks. Only the largest entities have class A addresses. You’ll find a lot of the original big players such as AT&T, HP, GE, IBM, Xerox, MIT, Apple, Level 3, a number of U.S. government entities, and the U.K. hold class A addresses. Some that have disappeared, like DEC, have been inherited by others, HP in this case.
Class B—A machine would look at the first bit and know this address can’t be a class A because the first bit is a one. It then looks to see if the second bit is a 0 or 1. If the second bit is zero it knows it is dealing with a Class B address. This gives a first octet with a range from 128-191. Class B addresses use the first two octets to identify the network. But again, it doesn’t have all 16 bits to use for unique networks, as the first two were used to identify the address as class B. So a class B address can identify 16,384 addresses. That leaves 65,536 addresses for nodes on each network. Class B addresses point to many universities, other large players that weren’t there at the beginning like Cisco, Tektronix, Sun, individual armed services, NASA, most defense contractors, a myriad of government agencies around the world, large IP service providers, and most hi-tech companies of any size.
Class C—As with class B addresses, the machine would see that the first bit is one. In this case, the second bit is one also, so it confirms that the third bit is zero. If so, then this is a class C address, which gives the first octet a range of 192-223. A class C address uses the first three octets, minus the first three identifying bits, so the remaining 21-bits from the first three octets yield 2,097,152 network addresses, allowing 256 nodes on each of those networks.