Preeclampsia is a pregnancy specific syndrome characterized by hypertension and proteinuria after 20 weeks of gestation. To reveal the relationship between lipids and preeclampsia, we conduct lipidomic profiling of maternal serums of 44 severe preeclamptic and 20 healthy pregnancies from a multi-ethnic cohort in Hawaii. Correlation network analysis shows that oxidized phospholipids (OxPLs) have increased inter-correlations and connections in preeclampsia, while other lipids, including triacylglycerols (TAGs), have reduced network correlations and connections. Thirty-one lipid species from various lipid classes demonstrate predominantly reductions and causal relationships with preeclampsia. They include phosphatidylglycerol (PG), TAG, diacylglycerol (DAG), phosphatidylcholine (PC), cholesterol esters (CE), phosphatidylethanolamine (PE), sphingomyelin (SM), ceramides (Cer-NS), hexosyl ceramides (HexCer-NS), lysophosphatidylcholine (LPC), lysophosphatidylethanolamine (LPE), and free fatty acid (FFA). Many of these lipids are also selected as important features by a linear discriminant analysis (LDA) classifier with high predictive accuracy (F-1 statistic 0.941 and balanced accuracy 0.88), indicating their potential to serve as biomarkers for severe preeclampsia. Our study supports the hypothesis of a phospholipid (PL) centered, dysregulated lipidomic metabolic atlas. That is, severe preeclampsia may be originated from hypoxia, which induces the accumulation of OxPLs through oxidative stress whereas reduces many other lipids (eg. reduced PCs, TAGs and ceramides). These molecular changes coherently lead to dysregulated biological functions, such as insulin signaling and inflammation/infections. Moreover, the lipid changes may also be responsible for the comorbidity between preeclampsia and gestational diabetes, a clinically known risk factor for preeclampsia.