January 22, 2025 – Microsoft announced an adjustment to its partnership with OpenAI yesterday, January 21st, allowing OpenAI to utilize computing resources from competitors. This modification is aimed at accommodating OpenAI’s escalating demand for computational power and easing tensions between the two parties arising from compute supply issues.
Previously, OpenAI’s compute requirements were exclusively reliant on Microsoft’s Azure cloud infrastructure. Due to the substantial compute investment necessary for OpenAI’s model training and user base expansion, Microsoft’s supply speed had caused dissatisfaction among some OpenAI shareholders. This dissatisfaction had prompted OpenAI to partner with Oracle in June, with Microsoft’s permission.
According to the announcement, the revised agreement primarily addresses compute exclusivity. Microsoft retains a Right of First Refusal (ROFR), implying that while OpenAI can obtain compute from other providers, Microsoft has the priority to fulfill its needs.
Microsoft has authorized OpenAI to independently build additional compute capacity, primarily for model research and training, further bolstering OpenAI’s research and development capabilities.

The foundation of the partnership, which extends until 2030, remains largely unchanged. Microsoft maintains exclusive rights to utilize OpenAI technology for products like Copilot, and OpenAI’s API will continue to be exclusive to Azure. The existing revenue-sharing model, reportedly with Microsoft receiving 20% of OpenAI’s income, will also persist.
This agreement adjustment coincides with the announcement of a $500 billion joint venture between OpenAI, SoftBank Group, and Oracle. The venture aims to establish a cloud computing datacenter named “Stargate” in the United States. Microsoft has expressed its intention to provide technical support for the joint venture but will not invest financially.