What new AWS AI capabilities were announced at the summit?

AWS introduced three main capabilities: a "harness" feature for Amazon Bedrock AgentCore that reduces the number of API calls required for agents to interact with models, a new AWS Context service that stores metadata using Apache Iceberg to help agents access business data, and improvements to agent support for handling context across multiple data types (public data, internal documents, and proprietary information).

Why did AWS invite both OpenAI and Anthropic to speak at the same event?

The consecutive keynotes on June 25 and 26 appear designed to signal that AWS is adopting a model-agnostic platform strategy. Rather than competing directly with these AI vendors, AWS is positioning itself as a neutral infrastructure layer where businesses can use whichever models and agents best fit their needs.

What is Amazon's financial commitment to Anthropic?

Amazon has announced an investment of up to $330 million（約530億円） in Anthropic, and is providing 100,000 of its AI training chips (Trainium) to support Anthropic's work on Claude and other development.

Back to articles

AWS shifts strategy to model-agnostic AI: OpenAI and Anthropic keynote on consecutive days

ITmedia AI＋2d ago5 min read

Key takeaway

AWS is reshaping its AI strategy to support multiple vendors' models and agents—not just its own—by unveiling tools that make it easier for businesses to combine different AI models on AWS. The company hosted both OpenAI and Anthropic as keynote speakers at its Japan summit, signaling that it will profit from orchestrating competing AI systems rather than competing with them directly.

Summaries like this, in your inbox every morning.

3 Key Points

What happened
At AWS Summit Japan 2026 (June 25–26), OpenAI and Anthropic each delivered keynote speeches on consecutive days. AWS announced new capabilities for its AI agent service (Amazon Bedrock AgentCore), including a "harness" feature that reduces the number of required API calls, and a new service called AWS Context that stores metadata using Apache Iceberg to help agents work with business data more flexibly.
Why it matters
AWS is signaling a pivot away from proprietary model lock-in toward a "best-of-breed" approach where businesses can mix models and agents from different vendors on AWS infrastructure. By hosting both OpenAI and Anthropic at the same event, AWS appears to be positioning itself as a neutral platform for AI workloads rather than pushing its own models, which may appeal to enterprises seeking multi-model flexibility.
What to watch
AWS is emphasizing four layers—model availability (via Bedrock), agent orchestration (via AgentCore), contextual data (via AWS Context), and model-agnostic identity and security—as its competitive advantage. The company is betting that managing how different models and agents interact, rather than owning the models themselves, is where the real value lies.

FAQ

What new AWS AI capabilities were announced at the summit?: AWS introduced three main capabilities: a "harness" feature for Amazon Bedrock AgentCore that reduces the number of API calls required for agents to interact with models, a new AWS Context service that stores metadata using Apache Iceberg to help agents access business data, and improvements to agent support for handling context across multiple data types (public data, internal documents, and proprietary information).
Why did AWS invite both OpenAI and Anthropic to speak at the same event?: The consecutive keynotes on June 25 and 26 appear designed to signal that AWS is adopting a model-agnostic platform strategy. Rather than competing directly with these AI vendors, AWS is positioning itself as a neutral infrastructure layer where businesses can use whichever models and agents best fit their needs.
What is Amazon's financial commitment to Anthropic?: Amazon has announced an investment of up to $330 million（約530億円） in Anthropic, and is providing 100,000 of its AI training chips (Trainium) to support Anthropic's work on Claude and other development.

Discussion

No discussion yet for this article

Stay ahead with AI news

Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.

Get Started Free

Free · takes 30 seconds · unsubscribe anytime

1 minute a day. The AI essentials.

200+ sources · Email / LINE / Slack

Get it free →