AWS rolls out OpenAI models as Microsoft exclusivity ends

OpenAI models are now available on Amazon Web Services a day after the company ended its exclusive cloud deal with Microsoft. The move expands OpenAI’s deployment options and gives Amazon a new AI product line to sell to enterprise customers.

AWS announced a slate of OpenAI model offerings, adding an agent service that developers can call from their own applications. The announcement came without detailed pricing or performance benchmarks, but it signals that Amazon is ready to host the same large language models that have powered ChatGPT on Azure.

Amazon Web Services adds OpenAI models

The AWS press release listed the OpenAI models as part of its broader AI portfolio. The list includes the standard GPT‑4‑style models and a newly‑named “agent” service that can orchestrate multi‑step tasks. Amazon did not disclose whether the agent runs on dedicated hardware or leverages existing Inferentia chips.

Amazon’s cloud team positioned the offering as a plug‑in for existing SageMaker pipelines. Existing SageMaker customers can import the OpenAI endpoints with a few API calls, according to the documentation. The integration promises lower latency for workloads that already run on AWS, but the claim has not been benchmarked publicly.

Analysts note that AWS now competes directly with Azure for the lucrative enterprise OpenAI market. The competition could drive down prices, but it also forces customers to evaluate two very different billing models and data‑ residency guarantees.

Amazon launches audio Q&A on product pages

In parallel, Amazon introduced a “Join the chat” widget on its retail site. The feature lets shoppers ask natural‑language questions about a product and receive spoken answers generated by an AI model.

The audio response plays automatically in the browser, turning a text‑only FAQ into a conversational experience. Amazon says the service runs on the same OpenAI models now available on AWS, though the exact model version is not disclosed.

Early user testing showed mixed results. The voice engine handled factual queries about specifications well, but it occasionally produced vague or repetitive answers for subjective questions. Amazon has not released usage metrics, so the impact on conversion rates remains unclear.

The rollout raises privacy questions. The widget records voice input, sends it to the cloud for processing, and returns an audio file. Amazon’s privacy notice states that data may be used to improve its AI services, but it does not specify retention periods.

The shifting OpenAI‑Microsoft partnership

OpenAI and Microsoft announced a restructured partnership that ends the former’s cloud exclusivity. Microsoft remains the “primary cloud partner,” and OpenAI products will ship first on Azure unless Microsoft cannot support a capability. The language makes Azure the default but not the sole host.

The new agreement also changes revenue sharing. Microsoft no longer pays a share to OpenAI for using its models, but OpenAI will continue to remit a 20 % share of revenue to Microsoft through 2030. Microsoft retains a non‑exclusive license to the models through 2032.

The restructuring follows months of tension over a rumored $50 billion Amazon investment in OpenAI. That investment, reported as part of a $110 billion funding round that also included SoftBank and Nvidia, would give Amazon a sizable stake in the AI startup.

OpenAI’s ability to sell services on AWS and Google Cloud now opens a path to an IPO, which market watchers expect later this year. The prospect of a public listing adds pressure on Microsoft to keep the relationship productive, even as the exclusivity clause disappears.

Market and technical implications

The immediate effect is a broader choice for enterprises that need to run OpenAI models at scale. Companies can now match their existing cloud contracts with model hosting, avoiding data‑transfer costs and multi‑cloud complexity.

From a technical standpoint, the agent service on AWS could simplify building autonomous workflows. However, without clear performance data, developers must test latency and cost against Azure’s well‑documented pricing.

Amazon’s audio Q&A experiment illustrates how retailers are experimenting with conversational AI at the point of sale. If the feature proves to boost conversion, other e‑commerce platforms may follow suit, increasing demand for low‑latency inference.

The competitive pressure may also force Microsoft to revisit its pricing or add new capabilities to retain first‑mover status. Meanwhile, Google Cloud, which has not announced a comparable OpenAI offering, could seek its own partnership or double down on its own Gemini models.

What to watch

Track the adoption rate of AWS’s OpenAI endpoints in the next quarter, especially among Fortune‑500 firms that already run workloads on SageMaker. Watch for any pricing announcements from Amazon that could undercut Azure’s current rates. Finally, monitor OpenAI’s SEC filings for clues about the timing of a potential IPO, which would lock in the financial stakes for both Microsoft and Amazon.