Please make an MWE that reproduces the problem, and ideally report it as an issue in the CSV repository. I cannot reproduce it with a simple example, eg
using CSV, DataFrames
N = 10^8
df = DataFrame(; a = rand(Int, N), b = rand(Float64, N), c = rand('a':'z', N))
dst = "/tmp/data.csv"
@time CSV.write(dst, df)
@show filesize(dst) / 2^30 # file size in Gb
takes 45s on my laptop, creating an almost 4GB file (using an SSD). I find this reasonable.