Authenticated encryption with associated data (AEAD) has become prominent over time because it offers authenticity and confidentiality simultaneously. In 2018, the National Institute of Standards and Technology (NIST) initiated a competition to standardize lightweight AEAD and hash functions, with Ascon as the final winner among the 10 finalists. Numerous prior works evaluated their performance on FPGA and ASIC, but not on a parallel architecture like GPU, which is a common accelerator already found in many existing cloud servers. In this work, the first GPU implementation of the NIST AEAD finalists is proposed. Several GPU implementation techniques applicable to all AEAD schemes are presented, along with novel techniques for some specific schemes to enhance throughput performance. Experimental results show that all NIST AEAD finalists can achieve high throughput (up to 111.53M AEAD per second), approximately 142.19 and 72.65 improvement compared to unoptimized GPU version, and the investigated FPGA results respectively. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.