Amazon AI

Amazon AI refers to Amazon's collection of artificial intelligence products and research, delivered primarily through Amazon Web Services (AWS). It encompasses Amazon's own Nova foundation model family, the Amazon Bedrock managed service for running third-party models, agentic tooling such as the Strands framework, and Amazon's multi-billion-dollar partnership with Anthropic.^[1]

History

Amazon applied machine learning to product recommendations, search and logistics for two decades before the generative AI era, and launched AWS AI services — including speech and image APIs — from 2016. As generative AI accelerated, Amazon invested up to $4 billion in Anthropic, one of the largest single corporate bets on a model developer, gaining a privileged role as Anthropic's primary cloud and training-infrastructure partner.^[2]

At its re:Invent conference in late 2024 Amazon introduced Amazon Nova, a family of text, image and video generation models positioned on price-performance. In 2025–2026 the portfolio was consolidated: reporting in July 2026 indicated Amazon was winding down most of its flagship Nova models — including the high-end Premier and Omni tiers, the Reel video model and the Canvas image model — while retaining Nova 2 Sonic and Nova 2 Lite as its core foundation models, Nova Forge for customer model building, and Nova Act for browser agents. In the same period Amazon's Strands Labs released the open-source Strands Decider 2B, a small model that judges whether a proposed agent action should proceed.^[3][4]

Key Concepts and Technology

Amazon's AI strategy differs from rivals in two respects. First, through Amazon Bedrock, AWS offers a curated catalogue of models from Anthropic, Meta, Mistral, AI21 Labs and others alongside Amazon's own, so customers can switch models without changing infrastructure — Bedrock is the leading example of a model-agnostic serving platform. Second, Amazon's in-house models are tuned for cost and latency over benchmark leadership, targeting high-volume workloads such as contact-centre speech, search ranking and recommendation.^[1]

The Strands agent SDK and Nova Act reflect Amazon's bet that AI value accrues less from a single frontier model than from reliable agentic workflows — software that plans, calls tools and verifies its own steps. Strands Decider's design, a 2-billion-parameter model whose only job is to approve or block another agent's action, exemplifies the emerging pattern of small, specialised guardrail models wrapping larger ones.^[4]

Applications and Impact

Amazon's AI reaches consumers through Alexa and retail personalisation, and enterprises through AWS services for contact centres (Amazon Connect), document processing, contact-centre analytics and code generation. Bedrock has become a common enterprise entry point in regulated industries precisely because it avoids lock-in to one model vendor. Amazon's investment in Anthropic also makes AWS the de facto home for Claude workloads, giving the cloud a frontier-model drawcard without requiring Amazon to win model benchmarks itself.^[2]

>See Also

• Amazon Bedrock
• Anthropic
• Claude Language Model
• AI Agents
• Sovereign AI