Amazon Research Awards Open

Amazon opened its Spring 2026 Research Awards cycle on March 25, with proposals due May 6 at 11:59 p.m. Pacific Time. (amazon.science) This round covers seven topics: artificial intelligence for information security, agentic artificial intelligence, Amazon 2030, Amazon Security, two Build on Trainium tracks, and robotics. Amazon said proposals will be judged on scientific quality, creativity, and potential impact at scale. (amazon.science) The awards are structured as unrestricted gifts to a principal investigator’s university or organization, and Amazon says it does not take intellectual property rights in the resulting work. Recipients are encouraged to publish results and release related code in open-source repositories. (amazon.science) In plain terms, agentic artificial intelligence is software built to act on goals across tools and data sources, not just answer a prompt once. Amazon’s call asks for work on low-code agent systems, multi-agent collaboration, safety, data access, and enterprise governance. (amazon.science) For the agentic artificial intelligence track, selected investigators may receive unrestricted funding averaging no more than $70,000 and Amazon Web Services promotional credits averaging no more than $50,000. Amazon said decision letters for that track are scheduled for August 2026. (amazon.science) The two Trainium tracks shift attention from model demos to the machinery underneath them: post-training, which is the tuning step after a base model is built, and kernels, the low-level code that helps chips run machine learning workloads faster. Both topics sit alongside robotics and security in a broader push toward systems research. (amazon.science) Amazon Research Awards has been running since 2015, and Amazon says it merged the program with Amazon Web Services Machine Learning Research Awards in 2020. The company now frames the program as support for academic and nonprofit research tied to its science agenda. (amazon.science) The recipient list shows 828 results across years and research areas, including 253 in machine learning, 91 in robotics, and 30 in agentic artificial intelligence. Amazon said its November 2025 cohort included researchers from 41 universities in eight countries, after a June 2025 cohort from 46 universities in 10 countries. (amazon.science) For researchers working on the plumbing behind artificial intelligence systems, the next date that matters is May 6. Amazon has already opened the portal, and this cycle closes in just over three weeks. (amazon.science)