Hah! Found it: BYTE 08-08, that is, August 1983, in an issue with the theme "The C language", Jerry Pournelle goes through the popular languages of the day to figure out what would be the most likely sweepstake language in the microcomputer world. He believes in Modula-2, thinks that "most useful" to learn right then would be a compiled basic (CB80) or Pascal, and has a small side bet on PL/1 taking the world by storm (after all, Gary Kindall likes it). For C, he notes that
" However, I don't expect it ever to become a highly popular language, and if I had to bet on its future, I'd say that it will take
a respectable niche, after which its growth will be quite slow in comparison to the microcomputer world in general ."
I love computing archaeology
The reasons being that the language is difficult to read, does not enforce good programming habits, and that the compilers generate excessively large executables. He compiles a Hello World-thing with Lattice-C and ends up with an 11 kB executable, which is clearly out of line. He explains why this is (basically, it links a large standard library to "Make up for th missing UNIX environment" that the language expects. Further talk about systems programs in C, rewritten by an irate user in Assembler instead.
"There might be implementations that don't generate superfat code, but we haven't come across one."
I think PL/1 had some traction as a systems programming language around then still, but writing stuff in pure assembler was quite obviously an option. And this is written 2 years
after the original MS-DOS came out (and had a total size almost the same as that Hello World program).