Briefly: yes, you generate lots of temporary arrays, but it is not slow because it often allows the use of vector instructions and parallelism, it can be faster than a constant-memory loop that does not take advantage of vectorization or parallelization. Also note that most array language implementations reuse temporary arrays a lot doing in-place modification when possible.

If the necessary temporary arrays are really huge, array programmers will just do the computation in chunks that fit comfortably in memory.