Why does MIT have a /8 IPv4 block?

The key thing to understand is that the Internet started out as an experiment in interconnecting networks. The creators never expected it to grow into what it is today without major redesign work.

In the early days, 8 bits were used to identify the network and 24 bits to identify the host within the network.

As the Internet grew, there was concern about running out of network addresses. The original vision of the Internet had been linking a small number of large networks, but the reality was tending more towards a large number of small networks.

Classes were introduced by RFC 791 in 1981 to improve address utilisation. Class A addresses had an 8 bit network number and a 24 bit host number. Class B addresses had a 16 bit network number and a 16 bit host number. Class C addresses had a 24 bit network number and an 8 bit host number. The class of an address was determined by its most significant bits (0 for A, 10 for B, 110 for C).

When they again started to run out of addresses, CIDR was introduced in 1993, allowing allocations on any power of two blocksize (though, by policy, the smallest blocks routed on the internet were /24, the equivalent of the old class C) and deprecating the division of addresses into classes.

Some existing allocations remained in place through these processes. Others fell by the wayside as networks were reorganised.

MITs allocation is mentioned in IEN 115 which is dated 1979, so it came from the first phase where every network got what we would now call a /8.  A couple of other Universities also appear to have had allocations in the pre-classes era. If we look at RFC 776 which seems to be the last allocation list from the pre-classes era, we see at least three Universities - MIT, Stanford and UCL. MIT actually had TWO allocations for different networks.

As far as I can tell, two Universities kept /8 allocations from the pre-classes era into the ICANN era - MIT and Stanford. Stanford gave theirs back in 2000, apparently out of altruism.

In 2017, MIT split up their allocation and sold off parts of it. Specifically, they seem to have sold a bunch of /16s to Amazon.

Trying to find a definite answer—there are a few out there like this one and this one on Quora—but basically MIT was one of the earliest/first users of the Internet and because classless inter-domain routing was still not a common “thing” in the years prior to 1993/94.

FWIW, the 18.0.0.0/8 range was assigned to MIT from at least July 1990 according to the details gleaned in RFC 1166 and related IANA and RIR records show that MIT was recognized as controlling the whole 18.0.0.0/8 range in January 1994.

MIT’s acquisition of the range was based on the needs of the classful network architecture that was common at the time.

The simplest and most succinct answer I can find comes from Greg Skinner on Quora who states:

MIT's assignment (18.0.0.0/8) is one of the original assignments created during the early research phase of the Internet. It was originally assigned to MIT-LCS (Laboratory for Computer Science). You can find more information in RFC 796 - Address mappings, and other documents that it references.

And going even further into the history of the Internet, the first RFC referencing the assignment of network 18 to MIT-LCS (Laboratory for Computer Science)—the lab at MIT that did the work on the Internet at the time—can be found in the Network Working Group’s RFC 739 which is dated November 1977; look for “LCS Network” under the heading, “Assigned Network Numbers.”

From the Wikipedia article, “List of assigned /8 IPv4 address blocks”:

Some large /8 blocks of IPv4 addresses, the former Class A network blocks, are assigned in whole to single organizations or related groups of organizations, either by the Internet Corporation for Assigned Names and Numbers (ICANN), through the Internet Assigned Numbers Authority (IANA), or a regional Internet registry.

Each /8 block contains 16,777,216 addresses.

As IPv4 address exhaustion has advanced to its final stages, some organizations, such as Stanford University, formerly using 36.0.0.0/8, have returned their allocated blocks to assist in the delay of the exhaustion date.

Someone asked why they did not give back the block, or at least part of it, this is an interesting question and apparently Stanford has done this.