Rather than the size of the directory files¹, it looks like you want the cumulative size or maybe disk usage of those directories and all the the files that can be found through their traversal.
Then you likely want:
fdfind --type d 2023 -X du -sh --
(here using the fdfind name of that command as found on Debian and derivatives as there's a pre-existing fd command which is for something completely unrelated; replace with fd if that's how it's called on your system).
That's for disk usage; add a --apparent-size option to du for size instead.
Bear in mind that du does some deduplication and counts the disk usage of a file only once if it's found (via any of its hardlinks or via the same paths) in more than one directory. For instance, if there's a 2023/2023 directory, you'll see 0 for the disk usage of the subdirectory, because it and all the files within have already been accounted for in the report for the upper 2023 directory.
You can disable that deduplication with --count-links.
You may also want to pass a --prune option to fdfind to stop looking for more files in the directories that match.
Note the -- which you need to separate options from files (without it, you'd run into issues for a file called -2023, or --files0-from=2023/etc/shadow for instance (leading to a potential information disclosure vulnerability). Same would apply for stat or most commands.
And note the -X instead of -x for du to be passed as many file paths as possible instead of being run once in a new process for each file. Note that because du may end up being called more than once if there's a large list of matching files, that could make the deduplication mentioned above unreliable. A better way to pass the file list to du would be via its stdin using the --files0-from option. Then you could even add the -c option to get a cumulative total line at the end:
fdfind --type d --print0 2023 | du --files0-from=- -sch --
-h (--human-readable), --count-links, --files0-from and --apparent-size are all non-standard extensions of the GNU implementation of du. -h is commonly found in other implementations these days, the other ones (including the --human-readable long form of -h) more rarely so.
¹ think of a directory as a phone directory for instance. On many file systems, that's a bit like a CSV file that contains a list of file names in one column and where to find them (the inode number) on disk in another (and sometimes a third column to indicate the type of the file). And like any file that directory/csv file has a size and a disk usage of its own and that's the one gstat -c %s or gfind -printf %s or ls -l report.
