Publishing scientific research is a highly profitable business—unsurprisingly because publishers are given the research, which is highly valuable, for free. Robert Maxwell—a notorious British business bully who fell off his yacht and drowned after robbing the pension fund of his employees—got rich not through publishing the Daily Mirror but through publishing science.
Reed-Elsevier, the world's largest publisher of scientific research, publishes over 1800 journals. In 2005 it made an adjusted operating profit of £1142m on a turnover of £5166m. The scientific part of the company contributed 39% of the profit (£449m) from 28% of the business (£1436). In other words, the scientific part of the company (which is predominantly medical) makes a large part of the profits and in some ways supports the non-scientific parts.
It is hard to come by the accounts of an individual journal, but I have seen many over the years. There are many comparatively minor journals that have an annual income of a million pounds. (I have never seen the accounts of the New England Journal of Medicine, but its annual income will be tens of millions of dollars and probably close to $100m.) The million pound journal might well have a gross margin (income minus direct costs like paper and printing) of £600 000. After subtracting the overheads the profit might well be £350 000. By no means all journals are so profitable, but this is a much more profitable enterprise than most.
... I should make clear that I am describing the classic research journal, which is comprised almost entirely of original research. Many of
the world's 10 000-20 000 biomedical journals (nobody knows the true number) still take this form, although many are now adding other features, such as review articles.