OpenAI and Microsoft amended their partnership on April 27, 2026, ending the exclusivity that has defined the AI industry's most consequential alliance since Microsoft's $1 billion investment in 2019. Under the new terms, OpenAI can now serve its models across any cloud provider, while Microsoft's license to OpenAI's IP and models extends through 2032. Azure remains the 'primary cloud partner,' but the word that mattered most — exclusive — is gone.
OpenAI will keep paying Microsoft a 20% revenue share, but the total is now subject to an unspecified cap and is guaranteed only through 2030, two years shorter than the IP license. The revenue share is also now 'independent of OpenAI's technology progress,' wording that retires the original deal's AGI clause, which would have voided exclusivity the moment OpenAI declared it had built artificial general intelligence.
The amendment lands two months after Amazon and OpenAI struck a $50 billion deal that put certain OpenAI models on Amazon Web Services. The Financial Times reported Microsoft had threatened legal action over that arrangement; the new agreement effectively moots the dispute. It also confirms what the Amazon deal already implied — Microsoft's grip on OpenAI's distribution has loosened.
Key facts
- 01Microsoft's license to OpenAI IP and models is now non-exclusive but extends through 2032.
- 02OpenAI continues paying Microsoft a 20% revenue share, capped and guaranteed only through 2030.
- 03The amendment follows a $50 billion Amazon-OpenAI deal struck two months earlier.
- 04Microsoft first invested $1 billion in OpenAI in 2019, locking in Azure exclusivity.
- 05Andy Jassy said OpenAI models will reach Amazon Bedrock 'in the coming weeks.'
OpenAI Chief Revenue Officer Denise Dresser made the commercial logic explicit in a staff memo this month obtained by CNBC. The Microsoft partnership had 'limited our ability to meet enterprises where they are — for many that's [Amazon] Bedrock,' she wrote. Customer interest in running OpenAI models through AWS has been 'frankly staggering,' according to Dresser.
“OpenAI keeps paying Microsoft 20% of revenue, but only through 2030 and now capped — and the cut is no longer tied to whether OpenAI hits AGI.”— Jaeden Schafer
Amazon CEO Andy Jassy moved quickly to capitalize. 'Excited to make OpenAI's models available directly to customers on Bedrock in the coming weeks, alongside the upcoming Stateful Runtime Environment,' Jassy posted on social media Monday. 'With this, builders will have even more choice to pick the right model for the right job.'
The deal is the formal conclusion of a renegotiation AI Chat Daily covered earlier this month, when reporting first surfaced that Microsoft and OpenAI were rewriting their pact to clear the Amazon agreement. What was a behind-the-scenes legal standoff is now a public structural change to how OpenAI sells.
For Microsoft, the trade is straightforward. It loses the moat of being the only hyperscaler that can resell GPT-class models, but it keeps a 20% cut of OpenAI's revenue for four more years and an IP license that runs another six. Kevin Scott, Microsoft's CTO and EVP of AI, has spent the past two years arguing Microsoft's value comes from its full stack — silicon, infrastructure, Copilot — rather than exclusive model access. This deal forces that thesis to hold up on its own.
For OpenAI, the upside is access to enterprise buyers who have standardized on AWS or Google Cloud and were not going to migrate workloads to Azure to use ChatGPT-class models. Bedrock alone hosts a deep roster of enterprise customers that OpenAI was previously locked out of serving directly. Sam Altman has been clear for months that distribution, not just model quality, is the constraint on OpenAI's revenue.
The retirement of the AGI clause is the quieter but arguably more interesting clause. Under the original 2019 agreement, OpenAI's board could have unilaterally severed Microsoft's commercial rights by declaring AGI achieved — a definitional landmine that hung over every product decision. Replacing it with a flat, time-bound revenue share trades narrative drama for accounting clarity.
The risk for OpenAI is the cap. A capped 20% revenue share through 2030 is a different financial profile than an uncapped one, and the public announcement does not disclose the ceiling. If OpenAI's revenue scales as aggressively as its current trajectory suggests, the cap could become the most valuable concession Microsoft extracted in the renegotiation.
The broader read is that the era of single-cloud AI exclusives is over. Anthropic already runs across AWS and Google Cloud. OpenAI is now joining it. Frontier model developers have realized that locking themselves to one hyperscaler caps their addressable market faster than it lowers their compute costs — and hyperscalers, in turn, are being forced to compete on price, latency, and tooling rather than on which model they alone can host.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
