Microsoft renews OpenAI partnership with $135 billion investment stake

New Delhi: Microsoft and OpenAI have signed a new definitive agreement that marks the next phase of their long-standing collaboration, first established in 2019. 

The renewed deal builds on what both companies describe as one of the most successful partnerships in the technology sector, setting the framework for continued cooperation and independent innovation.

Under the new structure, Microsoft will continue to support OpenAI’s transition to a public benefit corporation (PBC) following its recapitalisation. The technology company now holds an investment in OpenAI Group PBC valued at around $135 billion, representing roughly 27% on an as-converted diluted basis, including employees, investors, and the OpenAI Foundation. Excluding the latest funding rounds, Microsoft’s earlier stake in OpenAI’s for-profit entity stood at 32.5% on an as-converted basis.

The renewed agreement maintains the core elements of the partnership, including OpenAI’s position as Microsoft’s frontier model partner and the continuation of Microsoft’s exclusive intellectual property (IP) and Azure API rights until the point Artificial General Intelligence (AGI) is reached.

At the same time, the deal introduces new provisions designed to allow both organisations to pursue innovation independently. Once OpenAI declares AGI, that declaration will be verified by an independent expert panel. Microsoft’s IP rights for models and products are now extended through 2032, covering post-AGI models with safety guardrails in place.

Microsoft’s rights to confidential research methods used in model development will continue until either AGI is verified by the expert panel or through 2030, whichever occurs first. These research IP rights apply to internal or research-only models and exclude elements such as model architecture, model weights, and code for inference or finetuning. Microsoft retains non-research IP related to those areas, as well as to data centre software and hardware.

The agreement clarifies that Microsoft’s IP rights do not include OpenAI’s consumer hardware. OpenAI, meanwhile, gains the ability to jointly develop products with third parties, with any API-related products remaining exclusive to Azure. Non-API products can now be hosted on other cloud providers.

Microsoft will also have the ability to pursue AGI independently or through partnerships with other entities. If it develops AGI using OpenAI’s IP before AGI is declared, the models will be subject to compute thresholds significantly larger than those currently required to train leading models.

The revenue share agreement between the companies remains in place until AGI is verified by the expert panel, with payments now spread over a longer period.

OpenAI has committed to purchasing an additional $250 billion worth of Azure services, although Microsoft will no longer hold the right of first refusal to act as OpenAI’s compute provider.

The new terms also allow OpenAI to offer API access to US government national security clients regardless of the cloud provider and to release open-weight models that meet capability criteria.