OpenAI has formally ended its long-standing exclusive cloud service arrangement with Microsoft by restructuring their partnership agreement, turning instead to embrace other platforms such as Amazon. This move aims to break free from infrastructure constraints and fully compete for the enterprise market.
Earlier reports noted that on Monday, Microsoft and OpenAI announced revisions to their collaboration agreement. Key terms include Microsoft ceasing revenue share payments to OpenAI and converting its intellectual property license for OpenAI models from exclusive to non-exclusive. This historic loosening of ties immediately stirred market reactions: Microsoft’s stock fell nearly 4% in pre-market trading before partially recovering, while Amazon.com shares rose about 1% on expectations of benefits from a multi-cloud strategy.
On the same day, Amazon.com announced it had reached a deal to bring OpenAI to its Amazon Web Services cloud platform, which will introduce new services specifically designed to run AI agents.
For OpenAI, embracing platforms like Amazon.com is a critical step in catching up with rival Anthropic and overcoming growth bottlenecks in the enterprise sector. Previously restricted by exclusive terms with Microsoft Azure, some enterprise clients reluctant to migrate cloud platforms were shut out. Ending the exclusivity allows OpenAI to align with the industry trend toward multi-cloud deployment and reawaken broad latent demand.
Although OpenAI has gained vital operational autonomy and is eager to expand its revenue base through Amazon Web Services, the company faces serious commercial challenges in the multi-cloud era, as many enterprise customers have shifted to other highly competitive AI models over the past few years.
The termination of the exclusive agreement reshapes the cloud services value chain. The revised terms mark a significant restructuring of the financial relationship between Microsoft and OpenAI. Under the new arrangement, Microsoft will no longer pay a share of revenues to OpenAI, eliminating a fixed cost for Microsoft. In exchange, however, Microsoft loses its moat in AI infrastructure, as its exclusive IP access is broken. Meanwhile, OpenAI will continue reverse revenue-sharing payments to Microsoft until 2030, maintaining the original proportion but subject to a total cap.
At the equity level, Microsoft retains a 27% stake in OpenAI following last year’s for-profit restructuring and will remain a major shareholder in its growth. It was reported that the two sides engaged in months of tense negotiations last year over ambiguous AGI terms in the agreement. The new deal removes the mechanism that would have ended Microsoft’s exclusive access upon reaching AGI thresholds, eliminating a major uncertainty in their relationship. Market interpretation of the revision reflects cautious assessment of Microsoft’s ability to maintain its core position, while Amazon.com is widely expected to benefit directly from the move to non-exclusivity.
OpenAI’s expansion onto AWS is essentially a necessary response to growing computational demands and fierce competition in the enterprise market. Amazon.com previously announced an initial investment of $15 billion in OpenAI, with plans to introduce OpenAI technology on AWS’s Bedrock service featuring a "stateful" runtime environment that retains interaction details for more continuous use. AWS will also offer OpenAI Frontier to help businesses deploy AI assistants.
OpenAI executives noted in a memo that demand for the new AWS products has been "staggering." A spokesperson also emphasized that enterprise clients seek solutions that deliver tangible ROI, and the new offerings will enable businesses to natively deploy agent workflows in Amazon Bedrock. To further promote the collaboration, AWS CEO Matt Garman and OpenAI leadership plan to host an event for AWS customers on April 28, focusing on agentic AI technology.
OpenAI is currently racing to catch up with competitor Anthropic in the enterprise space. Disclosed figures show Anthropic’s recent annualized revenue reaching $30 billion, slightly ahead of OpenAI. In response, OpenAI leadership has directed employees to reduce peripheral projects and concentrate on winning more enterprise business.
Despite Amazon.com’s active promotion of OpenAI’s inclusion, some AWS customers have responded tepidly. In the three years since OpenAI ignited the AI boom, many enterprises have become deeply reliant on other models. Six companies or consultants working with AWS expressed satisfaction with cost-effective options such as Anthropic and Amazon.com’s own model Nova, available through Bedrock.
Adam Sandman, CEO of testing software firm Inflectra, remarked that the announcement would have been significant a few years ago, but today the market’s real focus is on models like Qwen and DeepSeek. He emphasized that in practical tasks such as coding, Claude already performs better. Inflectra previously had about 10% of its clients using Microsoft Azure to access OpenAI, but that share has since fallen to less than 1%, with the vast majority returning to AWS-managed services using Bedrock models.
Data control and cost efficiency also pose significant barriers to customer adoption of OpenAI. Phil Christianson, Chief Product Officer at IT software company Xurrent, pointed out that transferring data to Microsoft Cloud conflicts with client requirements to keep data within AWS. As leading models narrow in capability, incentives to switch platforms are limited. Cris Daniluk, CEO of AWS consultancy Rhythmic Technologies, added that while the new runtime offers technical advantages, being tied to OpenAI products can sometimes be frustrating. As multi-cloud setups demonstrate advantages in reducing latency and optimizing performance, OpenAI must still prove its indispensable business value in a market already penetrated by competitors.
Comments