OpenAI, MS finalize contract... Ending cloud monopoly and moving toward 'multi-cloud'

OpenAI and Microsoft ($MSFT) have once again adjusted the core terms of their artificial intelligence alliance. With this revision, OpenAI is no longer locked into a specific cloud service and can directly offer its products and models through multiple cloud providers.

This change is at least the third contract modification since 2025. The partnership between the two companies began with Microsoft’s $1 billion investment in June 2019, after which OpenAI has primarily used Microsoft Azure as its de facto core infrastructure. However, after the exclusivity cloud terms were abolished in January of this year, the main rights and revenue sharing structure largely remained intact.

The most notable change is the distribution structure of the OpenAI application programming interface (API). Previously, only Microsoft could offer this API on its cloud, but under the revised contract, OpenAI can now provide its products to customers via “all cloud providers.” This effectively means a multi-cloud strategy has been officially established.

This change was immediately reflected in the market. Amazon ($AMZN) CEO Andy Jassy announced on X that OpenAI models will be available to Amazon Bedrock customers within weeks. This is interpreted as OpenAI expanding its distribution channels in the generative AI market and beginning to formally cooperate with other major tech giants beyond Microsoft.

OpenAI also introduced a supporting “stateful runtime environment.” This service is a tool to help AI agents more easily manage data used for task automation. From an enterprise perspective, it also helps manage agents to comply with internal security policies. This is seen as OpenAI trying to expand from merely providing models to offering integrated enterprise-grade operational environments.

However, Microsoft’s priority has not disappeared entirely. Both companies stated that if OpenAI products meet certain conditions, they will be prioritized for release on Azure. But if Microsoft fails to support or chooses not to support the required features, this principle does not apply. While exclusivity has weakened, strategic priority collaborations still remain.

The intellectual property structure has also changed. Initially, Microsoft held an “exclusive” license to OpenAI models and some products, but the new contract adjusts this to a “non-exclusive” license valid until 2032. This broadens OpenAI’s space to sign separate technology-sharing agreements with other companies and opens up potential additional revenue streams.

This is the second case where Microsoft’s rights to OpenAI’s intellectual property have been weakened. Previously, when OpenAI restructured into a nonprofit under the public benefit corporation (PBC) structure last October, some rights were already reduced. At that time, intellectual property related to consumer hardware in development was excluded.

The revenue sharing structure has also been adjusted. According to CNBC, under the original contract, Microsoft could receive 20% of revenue from OpenAI products like ChatGPT subscriptions. This percentage will continue until 2030, but with a cap. Meanwhile, Microsoft decided not to share a portion of its own AI revenue with OpenAI anymore. While the cooperation remains on the surface, in essence, the profit-sharing structures are becoming more clearly differentiated.

This contract revision also aligns with Microsoft’s efforts to reduce dependence on OpenAI technology. Earlier this month, Microsoft publicly released its own AI models optimized for image generation and speech transcription. At the same time, the two companies maintain close transactional relationships. It is reported that last year, OpenAI committed to spending $250B on Azure services, roughly equivalent to 368.875 trillion Korean won.

Ultimately, this revision appears more like a “realignment” rather than a “breakup.” The pattern is: OpenAI moves toward a broader cloud market, while Microsoft, maintaining priority rights, develops its own AI capabilities. This is seen as a landmark scenario indicating that the generative AI market is shifting from early alliance-centered competition to a landscape of multi-party cooperation and platform expansion.

TP AI Notes: This article uses a language model based on TokenPost.ai for summarization. The main content may be incomplete or differ from actual facts.