Prof. Karl Kleine at Fachhochschule Jena in Germany found and scanned some documents relating to the distribution of early versions of Unix, and kindly made them available.
The first (in PDF form) is the license issued to Katholieke Universiteit in Nijmegen, The Netherlands, in December 1974. They were one of the early educational users, and probably the same license was used for all the educational organizations at that time. Despite the name of the file, from the date of the contract, the license probably refers to the Fifth Edition system; the Sixth Edition manual is dated May, 1975. It's quite possible, however, that it was the 6th Edition that was actually delivered.
The license is full of boilerplate, but probably the important operative clause is that of 4.05, which effectively allows free use within the university, provided the users do not disclose outside the organization. Section 2.01 grants use "for educational and academic purposes only;" 4.05 requires the licensee not to disclose the software "or methods and concepts utilized therein," to anyone except employees or students as necessary for purposes granted. I believe that the wording allowed John Lions to teach his Unix course and prepare his famous Unix commentary, but that the terms were tightened up later to be more restrictive by the time of the Seventh Edition. However, I understand that the restriction against disclosing methods or concepts (as distinct from actual source code) caused ill-ease to some university lawyers. This restriction was indeed a bit peculiar: the concepts had already been published, for example in the C. ACM paper.
This license was royalty-free, but there was a $150 administrative fee.
Kleine also found a couple of price lists for software as of about a decade later. In particular, there is a schedule for academic/educational prices as of September 1983; they're uniformly $400 for System III, System V, 32V, Seventh Edition. These were available for either the PDP-11 or Vax 11/780 as applicable. There's no mention of separate charges for multiple CPUs.
From few months later (February 1984), he found a fuller (and more expensive) price list for both commercial and educational licenses for System V release 2, plus various unbundled add-ons. The basic educational charge for the OS has increased to $800, and there is now a charge for additional CPUs, even educationally. I suspect that the 1983 educational schedule still applied, however, for SIII, SVr1 and the 7th Edition-related releases.
One other thing is evident in the later price list: there is a noticeable ($16K) charge for administrative use even by universities. This was intended to handle universities that wanted to use the system for their internal business (e.g. student registration) as distinct from teaching and research.
Kleine maintains an excellent history page, containing scanned documents of early manuals and specifications for Fortran, Algol, Pascal, and other languages, as well as documentation on several early machines and technologies.