Amazon Web Services (AWS) and OpenAI have announced a significant expansion of their partnership, integrating OpenAI’s frontier artificial intelligence models and agents into AWS infrastructure. The collaboration is designed to provide enterprises with advanced AI capabilities supported by the security, operational maturity, and data governance required for production workloads.
Three new offerings are currently available in limited preview, allowing customers to utilize OpenAI models on the AWS cloud platform.
Key Offerings in Limited Preview
1. OpenAI Models on Amazon Bedrock
For the first time, AWS customers can access the latest OpenAI models directly through Amazon Bedrock. This integration allows organizations to evaluate and deploy OpenAI models alongside existing options from providers like Anthropic, Meta, Mistral, Cohere, and Amazon.
Key benefits of this integration include:
- Unified Environment: Access, fine-tune, and orchestrate models through a single, consistent service.
- Enterprise Security and Governance: OpenAI models on Bedrock inherit existing enterprise controls, including IAM-based access management, AWS PrivateLink connectivity, encryption (at rest and in transit), and comprehensive logging via AWS CloudTrail.
- Financial Consolidation: Customers can apply their OpenAI model usage toward existing AWS cloud commitments, simplifying procurement and financial governance without requiring a new security model or infrastructure configuration.
2. Codex on Amazon Bedrock
Codex, an AI coding agent utilized by over four million developers weekly to automate, write, and refactor code, is now available within AWS environments. This allows enterprise teams to operate the OpenAI coding agent at scale using their existing infrastructure.
- Integrated Workflow: Customers can authenticate using AWS credentials and process inference directly through Amazon Bedrock infrastructure.
- Accessibility: Codex on Bedrock is accessible via the Bedrock API, beginning with support for the Codex CLI, the Codex desktop application, and the Visual Studio Code extension.
- Billing Integration: Similar to the base models, Codex usage can be applied toward existing AWS cloud commitments.
3. Amazon Bedrock Managed Agents, Powered by OpenAI
To support complex, multi-step tasks that require memory, specific skills, and permission enforcement, AWS has introduced Amazon Bedrock Managed Agents. This service combines OpenAI’s agentic capabilities with AWS’s global infrastructure.
- Optimized Architecture: The service utilizes the OpenAI agent harness, engineered to maximize the performance, reasoning, and task-steering capabilities of OpenAI models.
- Built-in Security: Every agent operates with a unique identity, logs all actions for auditability, and runs inference entirely within the customer’s Amazon Bedrock environment.
- Scalability: The architecture is designed to support the deployment of hundreds of thousands of agents, leveraging AWS’s global scale and proximity to existing enterprise data and applications.
Infrastructure Support: Bedrock AgentCore
To support the deployment of these agents, AWS highlights the role of Bedrock AgentCore, an open platform designed to build, connect, and optimize agents at scale across various models and frameworks.
For users building on Bedrock Managed Agents, AgentCore serves as the default compute environment. As an organization’s agent footprint grows, AgentCore provides critical enterprise capabilities, including authorization policy enforcement, agent discovery, observability, and evaluation metrics.
Future Outlook
This limited preview marks the beginning of a deeper technical collaboration between AWS and OpenAI. As OpenAI continues to develop new reasoning and agentic capabilities, the two companies plan to bring these ongoing advancements directly to Amazon Bedrock, ensuring enterprise applications benefit from future AI breakthroughs.