I've recently collected some data very closely approximating the level of veterans at HBS... basically membership numbers for the Armed Forces Alumni Association (the veteran club on campus) over the past decade or so. It shows that the average number per year (this includes a few non-US veterans) to be 40, with a range of 27-50. It's probably fair to assume about 4-6 of the veterans are from non-US militaries, often from conscripted services but not always.
Below is a chart of the AFAA strength since 2000:
If one was to look for meaning in the patterns though, it can be deceiving. For one thing, one has to look at the year of application (assuming the Fall of the year counts for round 1 and round 2, and thus the bulk of the applicants). If one therefore subtracts 3 years from the graduation year to look at the application year, the following is observed:
The fewest number of veterans were admitted in late 2002 (and early 2003). But what does this mean? Does this mean that fewer military people applied that year? Does that mean it was more competitive that year? Or perhaps it was less competitive because so few applied? Is the reverse true in 2003-2004 when the most were admitted? I don't think one can conclude any such position given just this information.
The data shows a mean of 40 with a standard deviation of 9. Most likely, the minor changes year to year are mostly random, and reinforces the fact that HBS does not have "hard" quotas.
Please also note that I don't have data on HBS class sizes going back all those years, so that could have a consequence if military numbers change with class size. To the best of my knowledge, HBS class sizes don't change very often.
This data may be interesting to provide historical context, but shouldn't be interpreted to mean much more than that.