Amazon opens Spring 2026 research awards across agentic AI, robotics, security, and Trainium

New call for proposals widens the scope of Amazon Research Awards, with funding and AWS credits available for academic work spanning AI infrastructure, security, robotics, and open-source research.

Amazon has opened its Spring 2026 Amazon Research Awards call for proposals, inviting academic researchers globally to submit work across seven areas including agentic AI, robotics, AI security, and machine learning infrastructure.

The submission window runs from March 25 to May 6, 2026, with decisions expected in August. Amazon says proposals will be evaluated on scientific quality, creativity, and potential for impact at scale.

The program funds academic research and contributions to open-source projects, with awards structured as one-time unrestricted grants to institutions.

Expanded focus on agentic systems and AI security

This cycle includes seven tracks: AI for Information Security, Agentic AI, Amazon 2030, Amazon Security, Build on Trainium: Accelerating Post-Training, Build on Trainium: Kernels for ML Acceleration, and Robotics.

The AI for Information Security track includes topics such as “trustworthy and reliable agentic AI for security operations,” “AI-powered incident response and red-teaming,” and “vulnerability detection and remediation using agentic AI.”

The Agentic AI track focuses on multi-agent systems, human-AI collaboration, and enterprise-scale agent operations. It also includes research into no-code and low-code tools for deploying AI agents, alongside governance, safety, and data integration.

Amazon Security proposals include areas such as secure code generation, authentication frameworks for AI agents, and large-scale anomaly detection in cloud environments.

Trainium and post-training research take priority

Two tracks are dedicated to research on AWS Trainium, Amazon’s AI chip platform for large-scale model training.

The post-training track focuses on reinforcement learning, alignment techniques, reward modeling, and distributed training systems. The kernels track targets lower-level research, including automated kernel generation, optimization, debugging tools, and cross-framework translation.

Applicants in these tracks are encouraged to request AWS promotional credits, typically ranging from $50,000 to $250,000 or more depending on the scope of the project.

Funding structure and eligibility

Funding varies across tracks. AI for Information Security and Amazon Security awards may provide unrestricted funds averaging up to $80,000, while Agentic AI awards average up to $70,000. The Amazon 2030 track offers up to $100,000 in funding alongside AWS credits.

Robotics proposals may receive up to $50,000 on average, plus credits.

All funded projects are assigned an Amazon research contact and receive access to AWS training resources, including tutorials and hands-on sessions with Amazon scientists and engineers.

Researchers from around the world are eligible to apply. Each proposal must be submitted by a primary investigator, with one co-PI permitted.

Amazon says proposals should typically be no longer than three pages, excluding appendices, and may include requests for AWS promotional credits alongside a detailed research budget.