ChatGPT Enterprise is available through AWS Marketplace

On November 9, OpenAI announced that eligible organizations can purchase ChatGPT Enterprise through AWS Marketplace. The new purchasing channel uses the billing relationship the company already has with Amazon Web Services. OpenAI is the seller in the catalog and manages the service. The company says the offer includes the models and features of the plan it sells directly, with product updates reaching both purchasing channels at the same time.

Who sells the service and where the cost appears

The AWS Marketplace listing describes ChatGPT Enterprise as cloud software delivered as a service. Purchasing and billing go through the marketplace, and OpenAI provides the service. The company says eligible spending can be applied toward the customer's AWS spending commitments. The listing mentions EDP and PPA commitments and makes this subject to the terms of the organization's agreement.

The product page directs businesses to a private offer and describes pricing as specific to their requirements and eligibility. In AWS documentation, private offers are associated with the buyer's account and have their own acceptance deadlines. A user with the required permissions can see the product, seller and expiration date in the marketplace console. Accepting an offer creates an agreement that can later be viewed alongside the subscription details.

OpenAI names the customer's AWS or OpenAI account team as the contact for confirming eligibility. The announcement separately mentions ChatGPT for Healthcare and ChatGPT for Financial Services for eligible customers. The two sector-specific offerings have their own terms. Announced availability of the Enterprise plan and confirmation of an organization's eligibility are separate stages in this purchasing channel.

The AWS console separates available offers from those that were accepted or expired. For an accepted offer, the agreement identifier can be used to look up the subscription. This information belongs to the account managing the purchase. For assistance, the listing directs product questions to OpenAI and points to AWS Support for AWS infrastructure.

What the Enterprise workspace includes

The product listing includes ChatGPT conversations, ChatGPT Work, Codex and agents called dots. Conversation and analysis features support work with information and documents. ChatGPT Work covers tasks spanning several steps, such as research and preparing a deliverable. Codex is intended for work on code. Dots can track updates and continue tasks between conversations within the access and approval rules that apply to them.

The plan is purchased at organization level. OpenAI documentation describes a centrally administered workspace in which members gain access through invitations or the company's identity system. Access to tools depends on seat type, role and workspace policies. For example, where an eligible agreement offers a dedicated Codex seat, it provides access to Codex, while a standard seat also includes ChatGPT conversation features. The organization's agreement determines its seat types and billing terms.

OpenAI lists SAML authentication, member administration and control over the tools available in the workspace among the Enterprise controls. The company also describes control over connected internal sources and data retention for the Enterprise plan. According to OpenAI, business data is encrypted in transit and at rest, and is excluded from model training by default.

Permissions in the purchasing console and roles in ChatGPT are managed separately. AWS uses its own permissions to determine who can see marketplace offers. Within ChatGPT, OpenAI documentation distinguishes members, administrators and owners. The company using the service controls access to work conversations and files through its workspace policies, and the seat type determines which features are available to a member.

Who processes the data

The AWS Marketplace listing identifies OpenAI as the data processor for ChatGPT Enterprise. AWS is one of the subprocessors used to support the service, under the data processing terms referenced in the listing. AWS's role in purchasing and billing is separate from the storage and processing conditions selected for the Enterprise workspace.

The listing distinguishes where data is stored from where inference runs, meaning the processing a model performs to generate a response. Data residency applies to eligible content stored at rest. Inference residency applies to the model's GPU calculations on eligible content in supported regions. Each option applies once enabled and within the limits documented for the region and feature. Processing on other components and data sent through external integrations are subject to separate rules.

For storage, OpenAI lists Europe, defined here as the European Economic Area and Switzerland, among the regions available to eligible workspaces. The company says processes such as extracting text from a file and routing a request across services can take place outside the region, even with inference residency enabled. Data sent to an external app follows the terms of that app's provider.

How the offer differs from Amazon Bedrock

OpenAI presents the new listing as complementary to its models available through Amazon Bedrock. ChatGPT Enterprise provides a managed workplace product that members access in the organization's workspace. Bedrock is a route for development teams to use models in applications and agents they build. In the first case, the agreement covers access to the Enterprise product. In the second, the model integration is part of the team's application.

For example, a team working with documents that needs managed access for employees uses the plan's workspace features. A team adding generated responses to its own product uses a model access service and sets the application's rules separately. OpenAI's announcement describes these as different uses within the same organization. The terms of an agreement, member access and processing conditions apply at the level of the relevant offer and service.