Michael Dickens has a repository that contains a program to optimize keyboard layouts, as well as some English letter frequency data. Among other things it contains the frequency of many bigrams.
This gist contains a program that extracts those bigram frequencies into a easily usable JSON format.
It also contains the result of running that program (bigrams-all.json), as well as a version of it where the order of the letters of a bigram is not taken into account (pairs-all.json). There are also versions of those files which only include the letters A–Z (bigrams.json and pairs.json, respectively).
To regenerate the JSON files:
# Note: This is the exact commit I got the data from. You might want to use
# a more recent commit, if available.
$ curl https://raw.githubusercontent.com/michaeldickens/Typing/5a8ecbf3d0589a4694ad0d51ad618d9ccb2a7313/data/allDigraphs.txt >data.txt
$ npm install
$ node text-to-json <data.txt >bigrams-all.json
$ node bigrams-to-pairs <bigrams-all.json >pairs-all.json
$ node filter-a-z <bigrams-all.json >bigrams.json
$ node filter-a-z <pairs-all.json >pairs.json
All of the files are in the public domain.