The biggest exclusive relationship in AI just ended. OpenAI and Microsoft jointly announced an amended agreement that kills Microsoft’s cloud exclusivity, clearing the way for OpenAI models to run on Amazon Web Services and other cloud providers. If you run a service business that touches AI in any way, this changes the math on where and how you deploy.
For years, if you wanted to run OpenAI’s models in your own infrastructure, you were locked into Azure. Full stop. That constraint shaped vendor decisions, pricing negotiations, and architecture choices for thousands of companies. Now the lock is off.
What happened
- OpenAI and Microsoft amended their partnership agreement. Microsoft’s license to OpenAI’s models and IP is now non-exclusive, running through 2032.
- Azure remains OpenAI’s “primary cloud partner,” but OpenAI can now serve its products to customers across any cloud provider.
- The change follows a $50 billion deal between Amazon and OpenAI announced two months ago, which included plans for OpenAI models to run on Amazon Bedrock.
- Microsoft reportedly threatened legal action over the Amazon deal, according to the Financial Times. This amended agreement appears to resolve that.
- Amazon CEO Andy Jassy confirmed that OpenAI models will be available on Bedrock in the coming weeks.
- OpenAI’s Chief Revenue Officer Denise Dresser told staff in a memo that the Microsoft partnership had “limited our ability to meet enterprises where they are.” She called interest in running OpenAI models through Amazon’s cloud “frankly staggering.”
The numbers
- Microsoft first invested $1 billion in OpenAI back in 2019.
- OpenAI will continue paying Microsoft a 20% revenue share, but that payment is now capped at an unspecified amount and only guaranteed through 2030.
- Microsoft retains its license to OpenAI’s IP through 2032.
- Amazon’s deal with OpenAI was worth $50 billion.
5 things this means for service businesses deploying AI
- Vendor lock-in just got weaker. If you built your AI stack on Azure solely because that’s where OpenAI lived, you now have options. That’s real negotiating power when your cloud contract comes up for renewal.
- Multi-cloud AI is no longer theoretical. Running OpenAI models on AWS while keeping your other workloads wherever they already live? That’s about to be a normal thing. No more forcing everything onto one provider just to access GPT.
- Pricing pressure is coming. When one cloud provider has exclusive access to the most popular AI models, they set the terms. Competition between Azure, AWS, and eventually others will push inference costs down. Your margins get better.
- The “which cloud” conversation with clients changes completely. If you advise clients on AI infrastructure, you no longer have to lead with “well, you’ll need Azure for that.” You can match the cloud to the client’s existing setup, compliance needs, and budget.
- The AGI clause drama is over (for now). The original deal had a clause that would have ended exclusivity if OpenAI achieved artificial general intelligence. The new deal makes the revenue share “independent of OpenAI’s technology progress,” which sidesteps that whole messy debate.
The hot take
Microsoft lost this one, and they know it. Keeping OpenAI locked to Azure was the single biggest competitive advantage Microsoft had in the enterprise AI race. Now that advantage is gone. Yes, Azure stays the “primary” partner, but that’s a participation trophy when your competitor just wrote a $50 billion check. The real winner here is every business that was forced into an Azure commitment it didn’t want. Competition is about to make enterprise AI cheaper and more flexible, and that’s long overdue.
The Agency OS play
If you’re running a service business and you’ve already deployed AI workflows on Azure because OpenAI required it, don’t rush to migrate. But do start planning. Pull up your Azure contract and check when it renews. That renewal conversation now includes the phrase “we could run this on AWS instead,” which is the most powerful sentence in cloud procurement. Use it.
For client-facing work, start auditing which clients are on AWS versus Azure versus Google Cloud. The ones on AWS who wanted OpenAI models but couldn’t justify an Azure migration? Call them this week. Bedrock availability is coming in weeks, not months. Be the person who tells them it’s possible before their current vendor does. That’s how you win the next contract.
Longer term, build your AI systems to be cloud-agnostic from the start. Abstract your model calls behind a simple interface so swapping providers is a config change, not a rewrite. The era of being tied to one cloud for AI access is ending. The businesses that architect for flexibility now will save real money and move faster than everyone who waits.