Amazon and OpenAI Form New Alliance: OpenAI Models Coming to AWS

Two tech giants, OpenAI and Amazon Web Services (AWS), have announced a new partnership. Through this collaboration, two new OpenAI open-weight reasoning models will be made available on the AWS platform. Starting Tuesday, these models will be accessible to users via AWS’s Bedrock and SageMaker AI services.

This move marks a significant strategic step for Amazon, as AWS had not previously offered OpenAI models directly. Although these models are downloadable through Hugging Face, Amazon will provide them as its own service.

This partnership gives Amazon a competitive edge by adding OpenAI to its roster of primary model providers. Previously, AWS primarily hosted Anthropic’s “Claude” model, a major rival to OpenAI.

Via the Bedrock platform, customers can build and host generative AI applications using their preferred models, while SageMaker allows users to train or create their own models.

Meanwhile, Microsoft remains OpenAI’s largest cloud partner and has already deployed optimized versions of these new models for Windows devices.

Amazon CEO Andy Jassy sees Microsoft’s success as a major challenge. Recently, Wall Street analysts questioned why Amazon lags behind in AI, to which Jassy responded by calling Microsoft a “second local competitor,” making a remark about their size.

At the same time, Oracle, another competitor, recently signed a $30 billion data center deal with OpenAI, positioning itself as a powerful OpenAI partner compared to Amazon.

For OpenAI, this new partnership strengthens its position amid its long-term collaboration with Microsoft. It also enables AWS’s vast corporate customer base to easily build AI applications using OpenAI models.

Additionally, OpenAI has released these new models under the Apache 2.0 open-source license. This contrasts with Meta’s “SuperIntelligence” models, which are unlikely to be fully open-source, as Meta recently acknowledged.

This joint initiative sets a new direction in the technology and cloud computing landscape.