BACKGROUND & AIMS: Recent data in mice have identified de novo ceramide synthesis as the key mediator of hepatic insulin resistance (IR) that in humans characterizes increases in liver fat due to IR (‘Metabolic NAFLD’ but not that due to the I148M gene variant in PNPLA3 (‘PNPLA3 NAFLD’). We determined which bioactive lipids co-segregate with IR in the human liver. METHODS: Liver lipidome was profiled in liver biopsies from 125 subjects that were divided into equally sized groups based on median HOMA-IR (‘High and Low HOMA-IR’, n=62 and n=63) or PNPLA3 genotype (PNPLA3(148MM/MI), n=61 vs. PNPLA3(148II), n=64). The subjects were also divided into 4 groups who had either IR, the I148M gene variant, both of the risk factors or neither. RESULTS: Steatosis and NASH prevalence were similarly increased in ‘High HOMA-IR’ and PNPLA3(148MM/MI) groups compared to their respective control groups. The ‘High HOMA-IR’ but not the PNPLA3(148MM/MI) group had features of IR. The liver in ‘High HOMA-IR’ vs. ‘Low HOMA-IR’ was markedly enriched in saturated and monounsaturated triacylglycerols and free fatty acids, dihydroceramides (markers of de novo ceramide synthesis) and ceramides. Markers of other ceramide synthetic pathways were unchanged. In PNPLA3(148MM/MI)vs. PNPLA3(148II), the increase in liver fat was due to polyunsaturated triacylglycerols while other lipids were unchanged. Similar changes were observed when data were analyzed using the 4 subgroups. CONCLUSIONS: Similar increases in liver fat and NASH are associated with a metabolically harmful saturated, ceramide-enriched liver lipidome in ‘Metabolic NAFLD’ but not in ‘PNPLA3 NAFLD’. This difference may explain why metabolic but not PNPLA3 NAFLD increases the risk of type 2 diabetes and cardiovascular disease.