The antifungal agent amphotericin B (AmB) is a polyketide produced by Streptomyces nodosus. The synthetic precursors of the amphotericin macrolactone skeleton are acetyl-CoA, malonyl-CoA and methylmalonyl-CoA. The genome sequence of the wild type S. nodosus ATCC14899 revealed a type II polyketide synthase (PKS) competing for malonyl-CoA. The same competitive branch was sequenced and verified in a mutant named S. nodosus ZJB2016050 (S. nodosus N3) screened in our lab. The transcriptome of the secondary metabolic synthetic gene cluster comparisons suggested that type II PKS (PKS5) competition is a factor in low production. The deletion of the PKS5 gene led to the titer of AmB improved from 5.01 g/L to 6.32 g/L while the by-product amphotericin A (AmA) reduced from 0.51 g/L to 0.12 g/L. A sequence of genes including PKS amphA, acc1, mme and mcm were overexpressed in a ΔPKS5 mutant, resulting in improved production AmB from 5.01 g/L to 7.06 g/L in shake flasks at 96 h. The yield of AmB and AmA in a 5 L bioreactor at 144 h was 15.6 g/L and 0.36 g/L, respectively. The intracellular reducibility of the wild type, mutagenesis type and genetically engineered type were detected, which was first found to be related to the by-product AmA. The increment of skeleton biosynthesis may consume more NADPH and reduces AmphC ER5 domain reduction. This study can be implemented for other polyketides in industrial production.