For doing DE analysis (on gene level!) with Candidas and Saccaromyces (genome size ~10-12 MB) we usually sequence 20-25 M reads, after mapping we usually get 90-95% of unique mapping, so the "effective" reads are around 17-20 mln (if you do trimming you also get rid of some data).
If you only plan to do DE, you can go even lower, say 10-15 mln, but if you add a bit more you also can do some additional things, e.g. variant calling, allele-specific expression, etc.
EDIT: if you have enough money, I would invest money to perform as many replicates as possible, with minimum of 3 per condition.
This is a very nice work comparing depth vs replicates. + some self-advertisement from the field of fungal transcriptomics.