Even the strongest partnership cracks in front of a big enough check. The exclusive deal Microsoft signed when it invested $1 billion in OpenAI in 2019 was the foundational formula of the AI industry for seven years: OpenAI models run only on Azure. That formula dissolved on April 27, 2026, with a single contract amendment. But read this as a simple change in cloud deployment policy and you miss the point.
What Happened: The Formal Dissolution of a Seven-Year Exclusivity
OpenAI and Microsoft officially announced a revised partnership agreement on April 27, 2026. There are three core changes. First, OpenAI can now deploy its models and products to every cloud provider, including AWS and Google Cloud. Microsoft remains the priority partner for new product launches, but the exclusive lock is gone. Second, the 20% revenue share OpenAI pays Microsoft stays in place through 2030, but for the first time a total cap has been set, with the exact figure undisclosed. Third, Microsoft's license to OpenAI intellectual property holds through 2032 but has now converted to a non-exclusive license. In addition, the AGI clause from the original contract, which gave Microsoft obligations and rights tied to any OpenAI declaration of AGI, was also removed in this revision. Behind it all is Amazon's capital. Amazon agreed to invest up to $50 billion in OpenAI and signed terms expanding the existing $38 billion AWS contract by an additional $100 billion over eight years. That dwarfs the $13 billion total Microsoft has invested in OpenAI since 2019. Microsoft weighed legal action over breach of exclusivity but chose the negotiating table instead.
Why This Matters More Than People Think: The Collapse of the Azure-as-Default Myth
For seven years, the most common choice when enterprise CTOs designed AI infrastructure strategy was simple: run OpenAI models on Azure. This was not technical optimization but structural coercion, because Azure was effectively the only option for using OpenAI models properly. Now that structure is gone. Once the same GPT-family models can run on AWS Bedrock and Google Cloud's Vertex AI, enterprises will choose a cloud based on cost, latency, existing infrastructure, and data governance policy. Google is already actively evaluating a similar OpenAI deployment arrangement. According to Bloomberg, Microsoft will also stop collecting part of the revenue share it had received from OpenAI, which affects Microsoft's near-term financials.
Hidden Insight: The Real Question the Deleted AGI Clause Raises
The change that passed most quietly in this amendment is the most telling: the deletion of the AGI clause. The original contract had a provision granting Microsoft specific rights and obligations if OpenAI declared it had reached AGI. That this clause is gone implies two things. One, OpenAI may believe it is closer to AGI and does not want to hand Microsoft special authority at that moment. Two, Microsoft may have concluded the clause held no real value, or conceded it during negotiations. In the wider picture, this is the end of the formula that OpenAI equals Microsoft's AI asset. OpenAI now stands as an independent platform that can pit AWS, Azure, and Google Cloud against one another. Paradoxically this threatens all three clouds, because each finds it harder to justify investment in its own AI models. The bear case, however, is real: critics argue OpenAI's freedom comes at the cost of its cheapest, most committed capital partner, and that spreading across three clouds multiplies operational complexity and security surface while Microsoft, now unbound, has every incentive to accelerate its own non-OpenAI models and starve the very partner it just freed.