The next chapter of the Microsoft–OpenAI partnership (openai.com)

Microsoft and OpenAI announced a new definitive agreement and the formation of an OpenAI Group public benefit corporation (PBC) that recapitalizes ownership—after the deal Microsoft holds about a $135 billion stake (≈27% on an as‑converted diluted basis). The pact preserves Microsoft as OpenAI’s frontier model partner and keeps Azure API exclusivity and Microsoft’s IP rights in place through 2032 (including post‑AGI models with safety guardrails), while adding an independent expert panel to verify any OpenAI declaration of AGI. Microsoft can now pursue AGI independently or with others, and the revenue‑share and Azure purchase commitments remain (OpenAI contracted to buy an incremental $250B of Azure services), though Microsoft no longer has a right of first refusal for OpenAI’s compute provider. For the AI/ML community the agreement clarifies governance, IP boundaries and cloud dynamics. “Research” IP tied to development methods stays with Microsoft until AGI verification or 2030 (whichever first) but explicitly excludes model architecture, weights, inference/finetuning code and datacenter hardware/software—items Microsoft still retains rights to as non‑Research IP. OpenAI may jointly build products with third parties (API products remain Azure‑exclusive; non‑API products can run on any cloud), can provide API access to U.S. national security customers, and may release open weights that meet capability criteria. There are also large compute thresholds on any AGI built using OpenAI IP prior to declaration—meant to shape the compute trajectory of the AGI race. Overall, the deal balances long‑term commercial ties, more multi‑cloud flexibility, and added oversight mechanisms that will influence openness, competition, and safety norms in advanced AI.