git update --init by default performs a normal git clone. This is less favorable when bandwidth or traffic is limited. I'm going to show a workaround for this, the core idea is to only fetch the needed commit.

The following steps used a small repository that contains submodule: node-libtidy.

1. shallow clone the submodule repo

$ git submodule update --init --depth=1

It is normal to see a error here, don't panic.

remote: Total 0 (delta 0), reused 0 (delta 0), pack-reused 0
error: no such remote ref abb03163aeafb8b7fc1efd2413d9f077bcdbeed9
Fetched in submodule path 'tidy-html5', but it did not contain abb03163aeafb8b7fc1efd2413d9f077bcdbeed9.
Direct fetching of that commit failed.

2. find the needed commit SHA1

If the previous command contains the SHA1, just use it.

There are other ways if you missed it:

$ git submodule status
-abb03163aeafb8b7fc1efd2413d9f077bcdbeed9 tidy-html5

# OR

$ git ls-tree HEAD:
100644 blob d32ca78ff42b7cf2b532a1eb1dd35bc81b584046    .gitignore
...
160000 commit abb03163aeafb8b7fc1efd2413d9f077bcdbeed9  tidy-html5

(background: commit sha1 of submodule is stored in a git tree object)

3. fetch that commit

# cd into the bare repo cloned for the submodule
$ cd .git/modules/tidy-html5/

# fetch by sha1, the full sha1 is required
$ git fetch origin abb03163aeafb8b7fc1efd2413d9f077bcdbeed9

4. checkout the correct commit in submodule

# back to main repo
$ cd ../../../

# update the submodule again without fetch, it should succeed this time
$ git submodule update --no-fetch