Microsoft and OpenAI forge new path in AI partnership

Microsoft and OpenAI have finalised a significant update to their strategic partnership, building on a foundational relationship that started in 2019. This new definitive agreement strengthens their collaboration and clarifies the operational rules for the next phase, especially concerning the future of Artificial General Intelligence (AGI).

Recapitalisation and Microsoft's investment

The most immediate change involves OpenAI's governance and ownership structure. Microsoft supports the OpenAI board’s move to form an OpenAI Group Public Benefit Corporation (PBC) and recapitalise the organisation. Following this action, Microsoft's investment in OpenAI Group PBC now stands at an approximate valuation of USD 135 billion. This figure represents roughly 27 percent of the company on an as-converted diluted basis, which includes all stakeholders: employees, investors, and the OpenAI Foundation. For context, before recent funding rounds, Microsoft held a 32.5 percent stake in the OpenAI for-profit entity.

Core partnership elements remain

The agreement preserves the essential elements that have defined their successful collaboration. OpenAI remains Microsoft’s frontier model partner, meaning Microsoft continues to be the primary corporate user for OpenAI's most advanced AI models. Furthermore, Microsoft retains exclusive IP rights and Azure API exclusivity until a formal declaration of AGI.

Key changes to the operating agreement

The new agreement introduces several modifications that provide greater operational independence and extended rights for both parties:

• AGI Verification: The declaration of AGI by OpenAI will now be subject to verification by an independent expert panel.

• Extended IP Rights: Microsoft’s Intellectual Property (IP) rights for both models and products are extended through 2032, and now include models developed post-AGI, with appropriate safety measures in place.

• Research IP Rights: Microsoft's IP rights to research, the confidential methods used to develop models and systems, such as internal deployment models, will remain until the expert panel verifies AGI or through 2030, whichever occurs first. Importantly, this research IP exclusion does not cover model architecture, weights, inference code, fine-tuning code, or hardware/software IP, which Microsoft retains.

• Hardware Exclusion: Microsoft's IP rights now exclude OpenAI’s consumer hardware products.

• Product Development Flexibility: OpenAI can now jointly develop some products with third parties. Any API products developed this way must beexclusive to Azure. However, non-API products may utilise any cloud provider.

• Independent AGI Pursuit: Microsoft now holds the right toindependently pursue AGI, alone or with other partners. If Microsoft utilises OpenAI’s pre-AGI IP to develop their own AGI, the resulting models must meet specified, high compute thresholds.

• OpenAI's Azure Commitment and Cloud Provider Choice: OpenAI has contracted to purchase an incremental USD 250 billion of Azure services. In a related provision, Microsoft will no longer have a right of first refusal to be OpenAI’s sole compute provider, suggesting greater long-term flexibility for OpenAI.

• Government Customers: OpenAI can now offer API access to US government national security customers, irrespective of the cloud provider they use.

• Open Weight Models: OpenAI now has the ability to release open weight models that satisfy necessary capability criteria.

The revenue share agreement remains in place until the expert panel verifies AGI, although payments will be made over a longer duration.

These changes position both Microsoft and OpenAI to continue their respective pursuits of AI advancement while managing the critical transition toward AGI and providing clear operational boundaries for growth.