AWS Launches AI Agent Marketplace With Anthropic On July 15

AWS launches AI agent marketplace with Anthropic, rolling out Nvidia Blackwell‑powered servers and making some major upgrades to SageMaker. AWS is betting big on infrastructure as a strategic differentiator.

1. AWS Launches AI Agent Marketplace with Anthropic

At AWS Summit New York on July 15, AWS will debut its long‑awaited AI agent marketplace, a one‑stop shop for AI tools and assistants built by startups and enterprises. Anthropic, backed by AWS, will be a featured partner. The platform allows enterprises to browse, purchase, install, and integrate AI agents designed to act autonomously, making decisions or executing multi‑step software tasks using generative AI engines.

AWS is expected to take a modest commission on agent sales, with startups setting their own SaaS‑style pricing. To further accelerate adoption, AWS may allow eligible startups and enterprises to apply their AWS credits toward these marketplace purchases, effectively subsidising the initial cost of deploying sophisticated AI agents. Though similar marketplaces already exist, like Google Cloud’s AI Agent Marketplace and Microsoft’s Agent Store within Copilot, AWS aims to attract a broader audience by leveraging its massive cloud customer base. For Anthropic, whose annualized revenue recently reached $3 billion, the marketplace offers an even bigger growth channel.

2. Nvidia Blackwell Superchips Arrive on AWS

In a parallel infrastructure play, AWS has introduced new EC2 instances powered by Nvidia’s Blackwell GB200 Grace Superchips, including the B100 and GB200 GPUs built for massive-scale generative AI workloads, especially inference. These new instances tap into AWS’s EC2 UltraClusters featuring 20k+ GB200 Superchips and 414 exaflops of compute in co‑developed Nvidia–AWS supercomputers.

Graphics‑heavy workloads, including multi‑trillion‑parameter LLMs, stand to benefit from up to 30× faster inference and 25× energy savings compared to previous-gen GPUs. AWS also offers these GPUs through Nvidia DGX Cloud, optimized for secure, high‑performance LLM training and deployment, further cementing AWS’s position in the GPU‑accelerated AI race.

3. Amazon Launches Kiro

Amazon Web Services has launched Kiro, a new AI-powered IDE designed to bring structure and clarity to “vibe coding.” Building off VS Code, Kiro emphasizes specification-driven development, allowing developers to define project requirements in a structured spec file. It automatically generates user stories using the EARS format, builds design docs including data flow diagrams, TypeScript interfaces, and API endpoints, and breaks everything down into task lists complete with testing and accessibility steps.

What sets Kiro apart are its event-driven “hooks”, which act like background agents. These can trigger actions like code scanning or updating documentation when files are saved or modified, ensuring consistency, security checks, and adherence to team standards. Initially launching in preview across Linux, macOS, and Windows with Amazon’s Sonnet models powering it, Kiro aims to bridge the gap between rapid prototype coding and production-ready software. AWS plans to continue expanding model support and introduce paid tiers after the preview.

4. SageMaker Evolves into Full Data‑to‑AI Hub

AWS’s flagship AI platform, Amazon SageMaker, is also receiving a substantial upgrade: it’s evolving into a unified data, analytics, and AI studio, consolidating tools like SageMaker Studio, EMR, Redshift, AWS Glue, and Bedrock into a cohesive interface 

Key new features include:

• SageMaker Lakehouse: unified access to all data, spanning from S3 lakes to Redshift warehouses and federated sources, making data queryable via SQL or analytics engines.
• Unified Studio: a single environment to prepare data, build and fine‑tune models, utilize foundation models via Bedrock, and collaborate, using tools like Q Developer Assistant for SQL and code generation
• Integration with Bedrock & Apache Iceberg: enabling zero‑ETL data flow into generative AI workflows and compatibility with open‑source standards.

This positions SageMaker not just for model development but as an end‑to‑end “data‑to‑AI” platform, mirroring trends seen in Microsoft Fabric but still with AWS’s flexibility across services.

Why Do These Moves Matter?

Collectively, AWS is staking out two distinct but complementary fronts in the AI war:

1. Marketplace Innovation

AWS Launches AI Agent Marketplace to tip the scale toward a new economy of autonomous AI agents, giving third parties a direct path to enter the AWS ecosystem.

1. Infrastructure Superiority

By scaling both proprietary Nvidia‑powered compute and enhanced data/AI tooling, AWS aims to match and outpace rivals like Microsoft and Google.

Underpinning this strategy is AWS’s heavy investment in custom Trainium‑based servers, which are part of “Project Rainier”, built specifically to power partners like Anthropic. That team is training Claude models on hundreds of thousands of Trainium 2 chips, representing AWS’s dual‑track approach: harnessing both proprietary silicon and best‑in‑class partner technology.

AWS CEO Adam Selipsky emphasized the co‑innovation with Nvidia as part of their strategy “to make AWS the best place to run Nvidia GPUs in the cloud,” while Anthropic and other startups gain unique distribution and monetization routes.

What’s Next?

• July 15: 

AWS Launches AI Agent Marketplace to kick off a broader agent ecosystem.

• Q3–Q4 2025: 

NVIDIA Blackwell instances and SageMaker upgrades go live.

• Long term: 

AWS’s integrated strategy includes

• Custom chips
• Partner alliances 
• Open compute

This places it in close contention with Microsoft’s Copilot offerings and Google’s Vertex AI.

AWS Launches AI Agent Marketplace to gain a competitive edge in the AI arms race: offering customers cutting‑edge compute, seamless tooling, and a vibrant marketplace across silicon, software, and services.