Flint changed its public water source in 2014, causing severe water contamination. We estimate the effect of in utero exposure to polluted water on health at birth using the recent Flint water crisis as a natural experiment. Matching vital statistics birth records with various sources of data, we employ a Difference-in-Differences (DID) approach as well as a Synthetic Control Method (SCM) to identify its causal impact on key birth outcomes. Our results suggest that the crisis modestly increased the rate of low birth weight (LBW) by 1.1-1.8 percentage points but had little effect on length of gestation or prematurity. The effects are larger for black or less educated mothers. Children born to disadvantaged mothers demonstrated 1.2-2.0 percentage points (or 10.4-17.4 percent) and 0.2- 0.6 percentage points (or 9.5-28.6 percent) rise in LBW and VLBW, respectively. We find little evidence that the Crisis increased fetal death, suggesting that the scarring effect in utero may dominate the channel of mortality selection. These results survive a rich set of placebo and falsification tests. Finally, our results lend support to three mechanisms at work linking water contamination and birth outcomes, i.e. biological effect, maternal stress, and avoidance actions.