Microsoft and OpenAI: A Reinvented Partnership Amid Tensions

A Reinvented Partnership Between Microsoft and OpenAI

The relationship between Microsoft and OpenAI has always been complex, marked by disagreements and contractual reorganizations. Despite these tensions, the two tech giants recently announced a revision of their partnership that, against all odds, appears amicable. This announcement surprised many observers who expected a more contentious breakup.

Microsoft revealed updates to its agreement with OpenAI, the most notable being that OpenAI can now offer its products and services on all cloud providers, not just Azure. The following day, OpenAI announced it would make its AI models, such as Codex, available on AWS, Microsoft’s main competitor in cloud computing.

An Inevitable Agreement with Amazon

This development seemed inevitable since Amazon signed a $50 billion agreement with OpenAI earlier this year. This agreement aimed to make AWS a third-party provider for OpenAI Frontier while collaborating on custom models for Alexa. Microsoft, unhappy with this deal with its main competitor, had considered legal action before renegotiating the terms.

Clearly, Amazon and OpenAI were looking to strengthen their collaboration, especially as Amazon tries to catch up in the AI space. OpenAI had informed its employees that the agreement with Microsoft limited its ability to meet the needs of businesses, particularly those using Bedrock, Amazon's service.

Financial Benefits for Microsoft

Although Microsoft is frustrated to see OpenAI's models available to Amazon, the company could still benefit from it. According to the revised agreement, Microsoft will continue to receive 20% of the revenue generated by OpenAI with ChatGPT and its API platform, including revenue from competing cloud platforms like AWS. Even though there is a revenue cap, this arrangement seems profitable for both parties, despite Microsoft’s initial irritation regarding the agreement with Amazon.

Microsoft remains OpenAI's "primary cloud partner" and will launch new OpenAI products first on Azure. However, the lack of exclusivity means these models may be available shortly thereafter on competing platforms. It is also possible that OpenAI's models could appear on Google’s Gemini Enterprise Agent service in the future.

An Extended License and Unidirectional Revenue

Microsoft's non-exclusive license for OpenAI's models and products has been extended until 2032, two years beyond the previous deadline of 2030. Revenue sharing is now unidirectional, with Microsoft no longer required to pay 20% of its Azure OpenAI revenue to OpenAI. Additionally, Microsoft will not share revenue with OpenAI if its advertising revenue increases by 15% year-over-year.

Microsoft holds about 27% of OpenAI's for-profit branch, making it a major shareholder in OpenAI's future growth. This amicable, albeit partial, separation was made possible by the removal of the clause regarding artificial general intelligence (AGI), which had long dictated the partnership between OpenAI and Microsoft.

The Removal of the AGI Clause

The removal of the AGI clause means that Microsoft will no longer lose access to OpenAI's most advanced future models once AGI is achieved. For its part, OpenAI is no longer incentivized to declare AGI to free itself from the agreement. The two parties had debated this clause for over a year, and tensions peaked when OpenAI considered acquiring the AI coding tool Windsurf for $3 billion last year.

OpenAI wanted Windsurf to be exempt from the agreement with Microsoft and even contemplated accusing Microsoft of anti-competitive behavior. Ultimately, OpenAI abandoned its acquisition plans, but this led to months of tense negotiations over OpenAI's controversial for-profit restructuring last year.

Growing Distrust and Internal Tensions

Microsoft's distrust of OpenAI increased when Sam Altman was ousted from his position as CEO in 2023 by the board of the nonprofit organization. This distrust prompted Microsoft CEO Satya Nadella to recruit Mustafa Suleyman, co-founder of DeepMind, to lead its AI efforts. Altman was reportedly "angry" that Suleyman joined Microsoft, especially as Suleyman's team began to engage directly with OpenAI engineers.

Tensions between Suleyman's team and OpenAI reached a boiling point when Microsoft's AI lead shouted at former OpenAI CTO Mira Murati during a 2024 meeting, demanding access to OpenAI's code. While Microsoft co-founder Bill Gates, CTO Kevin Scott, and Nadella had all delicately nurtured the agreement with OpenAI, Suleyman was brought in to shake things up and create his own AI models so that Microsoft would be less dependent on OpenAI in the future.

A Strategy for Independence for Microsoft

Some changes to the agreement also seem designed to distance Microsoft from the chaos surrounding OpenAI and Sam Altman. OpenAI is increasingly juggling public controversies, strategic shifts, and a multitude of executive reshuffles. The atmosphere appears tense, and the contract update comes in the same week that Elon Musk and Altman find themselves in court discussing the future of the AI company. Microsoft has already begun to get drawn into this drama.

Beyond Musk and Altman's legal antics, Microsoft continues to work quietly on creating models that are powerful enough to replace those from OpenAI. It has achieved impressive results with models like MAI-Transcribe-1, a speech transcription model that works in 25 different languages. However, most of Microsoft's AI models are highly task-specific or smaller in scale compared to OpenAI's large language models.

Diversifying AI Partnerships

To bridge this gap, Microsoft is increasingly turning to options from Anthropic, which are more powerful than OpenAI's models for certain tasks. Sources at Microsoft claim that the company has even considered using Google's Gemini models for some products, just as it has quickly begun testing Anthropic's Claude Code.

Microsoft is prepared to use the best AI model for each task, aiming to keep businesses tied to its enterprise software offerings and fend off competition from AI startups and even OpenAI itself. Part of the tensions throughout the partnership between Microsoft and OpenAI has been related to the collision of the two parties, each trying to convince the same mutual clients to purchase their AI services.

A Future of Financial Collaboration

All of this indicates that the future of AI at Microsoft will be a mix of models developed in-house and by OpenAI, Anthropic, and others. This leaves Microsoft and OpenAI with a partnership that is much more financially focused than on close collaboration regarding the future of AI. They will remain closely linked, but the situational partnership seems much less complicated than before.