Digital advertising is now a commonplace feature of political communication in the United States. Previous research has documented the key innovations associated with digital political advertising and its consequences for campaigns and elections. However, a comprehensive picture of political spending on digital advertising remains elusive because of the challenges associated with accessing and analyzing data. We address this challenge with a unique dataset (N=3,639,166) derived from over 13 million expenditure records reported to the Federal Election Commission (FEC) between 2004 and 2020. Employing a machine learning model to classify expenditures into nine categories including digital ads and services, this paper makes four key observations. First, 2020 was a watershed election in the growth of digital campaign spending. Second, there are clear partisan differences in the resources allocated to digital advertising. Third, platform companies play a central role in an otherwise partisan market for digital ads and services. Fourth, digital platforms and consultants occupy a distinct ideological niche within each party.