Background In some cross-sectional epidemiologic studies the shape of the association between serum concentrations of perfluoroalkyl acids (PFAAs) and lipids suggests departures from linearity. Objectives We used statistical approaches allowing for non-linearity to determine associations of prenatal exposures of perfluorooctane sulfonic acid (PFOS) and perfluorooctanoic acid (PFOA) with lipid concentrations. Methods PFAAs were measured in serum from pregnant women collected in 1991–1992 at enrollment in the Avon Longitudinal Study of Parents and Children and lipids in serum from their daughters at ages 7 (n = 111) and 15 (n = 88). The associations of PFAAs with lipids were first explored by cubic splines, followed by piecewise linear regressions by tertiles to obtain regression coefficients (β) and their 95% confidence limits (95% CL) (in mg/dL per 1 ng/mL). Results At age 7, total cholesterol was positively associated with prenatal PFOA concentrations in the lower tertile (β = 15.01; 95% CL = 2.34, 27.69) but not with PFOA concentrations in the middle (β = − 3.63; 95% CL = − 17.43, 10.16) and upper (β = − 1.58; 95% CL = − 4.58, 1.42) tertiles. At age 15, a similar pattern was noted as well. Positive associations between LDL-C and prenatal PFOA concentration in the lower tertile were observed in daughters at ages 7 (β = 14.91; 95% CL = 3.53, 28.12) and 15 (β = 13.93; 95% CL = 0.60, 27.26). LDL-C was not associated with PFOA concentrations in the middle or upper tertile at any age. Neither HDL-C nor triglycerides was associated with prenatal PFOA exposure. Non-linear patterns of association of total cholesterol and LDL-C with prenatal PFOS were less consistently noted. Conclusion Exposure to low levels of PFOA during prenatal development may alter lipid metabolism later in life. Given the small sample size further replication of the association in large independent cohorts is important.