The COVID-19 pandemic has prompted many school districts to turn to distance or at-home learning. Studies are emerging on the negative effects of distance learning on educational performance. Less is known, however, about the socio-economic, geographic, and demographic characteristics of students exposed to distance-learning across the United States. We introduce a U.S. School Closure & Distance Learning Database that tracks in-person attendance across more than 100,000 schools from January through October 2020. We measure year-over-year change in in-person attendance for each school throughout 2020 to estimate whether the school is engaged in distance learning after the onset of the pandemic. Our findings reveal large socio-economic, geographic, and demographic disparities in exposure to distance learning. In October 2020, schools recording the lowest third-grade math scores are nearly twice as likely to be closed compared to schools recording the highest math scores. The average racial composition of closed schools is 25 percentage points less white compared to schools operating in-person (40% versus 65%). Moreover, closures are more common in schools with a higher share of students who experience homelessness, are of limited English proficiency, are eligible for free or reduced-price school lunch, live in single-parent families, or are racial/ethnic minorities. Distance learning is more common in high schools and middle schools relative to elementary schools, but disparities in exposure to distance learning are comparable across school type. Given the potential negative consequences of school closures on educational performance, the vast inequalities in exposure to distance learning portend rising disparities in learning outcomes.