OpenAI Breaks Azure Lock in Microsoft Deal Rewrite

Microsoft drops exclusive OpenAI IP rights and ends its revenue-share payments as OpenAI gains freedom to deploy on Google Cloud, AWS, or any provider.

OpenAI Breaks Azure Lock in Microsoft Deal Rewrite

OpenAI was already running ChatGPT workloads on Google Cloud. The Microsoft deal signed in 2023 didn't stop that - it just made the arrangement legally uncomfortable. The amendment announced today, April 27, resolves that friction by formally ending Azure's exclusive position in the partnership.

Deal Changes at a Glance

TermBeforeAfter
Microsoft IP licenseExclusiveNon-exclusive (through 2032)
Cloud hostingAzure first and onlyAny provider (Azure still preferred)
Revenue: Microsoft to OpenAIYesEnded
Revenue: OpenAI to MicrosoftYes, open-endedThrough 2030, capped
AGI milestone clauseTied to paymentsRemoved

How the Old Deal Was Structured

The 2023 investment gave Microsoft three things: exclusive access to OpenAI's IP, a revenue-share arrangement flowing in both directions, and preferential cloud hosting rights that made Azure the default compute layer for OpenAI's products.

That structure made sense when Microsoft was writing large checks and OpenAI had one product category. It got complicated fast. Amazon's $50 billion deal promised OpenAI committed Trainium capacity, which directly challenged the Azure-first arrangement and prompted Microsoft to weigh legal action earlier this year. The AGI milestone clause - which tied certain payment terms to whether OpenAI had achieved artificial general intelligence - created another ambiguity nobody wanted to litigate.

The Exclusivity Problem

OpenAI needed infrastructure flexibility to honor commitments to multiple cloud providers. Every time OpenAI spun up capacity outside Azure, the exclusivity terms created a compliance risk. ChatGPT's reliance on Google Cloud infrastructure - confirmed by multiple sources before today's announcement - was technically in tension with what Microsoft had contracted for.

The Revenue Asymmetry

The original deal had money flowing in both directions. Microsoft paid OpenAI for commercial access; OpenAI paid Microsoft for cloud services and as a percentage of revenue. With OpenAI's annual recurring revenue now above $30 billion, those flows had grown large enough to complicate both companies' financial reporting and draw scrutiny.

What the Amendment Actually Does

Three things changed.

Non-Exclusive IP License

Microsoft keeps its license to OpenAI models and products through 2032, but that license is no longer exclusive. OpenAI can now license its models to other companies, cloud providers, and enterprises without any constraint from the Microsoft contract. Microsoft's own MAI model lineup - its effort to build independent AI capabilities - becomes more relevant now that it can't rely on exclusive OpenAI access to hold its competitive position.

Simplified Revenue Structure

Microsoft stops making revenue-share payments to OpenAI entirely. OpenAI continues paying Microsoft through 2030, at the same percentage as before, but subject to a total cap. The AGI milestone clause is gone. Removing it is cleaner than it sounds: nobody had agreed on what AGI meant in a legal context, so the clause was either unenforceable or a future dispute waiting to happen.

Multi-Cloud Freedom

The core operational change: OpenAI can now serve any product to any customer through any cloud provider. Azure remains the preferred platform and gets first access to new product launches - unless Microsoft "cannot or chooses not to" support the required capabilities. That carve-out is significant. It gives OpenAI a path to route workloads elsewhere whenever Azure's offering is technically insufficient.

What Developers Actually See

For developers building on the OpenAI API, the immediate change is nothing. Your API calls still hit the same endpoints.

from openai import OpenAI

client = OpenAI(api_key="...")
response = client.chat.completions.create(
    model="gpt-5",
    messages=[{"role": "user", "content": "What changed today?"}]
)

The computation backing that request can now legally run on Google Cloud, AWS, Azure, or any combination OpenAI chooses - without routing approval from Microsoft or customer opt-in. What matters to developers is what this unlocks over time: more infrastructure competition for OpenAI workloads means OpenAI has more bargaining power when negotiating compute costs. Analyst Max Weinbach noted the change could "bring down prices for customers." Whether OpenAI passes those savings through its API pricing is a separate question.

OpenAI and Microsoft partnership announcement Microsoft and OpenAI announced the deal revision on April 27 via the Official Microsoft Blog. Source: blogs.microsoft.com

Who Changes Position

The amendment reshapes relationships for four parties at once.

PartyGainsLoses
OpenAIMulti-cloud freedom, no revenue payments from MicrosoftAzure exclusivity as a negotiating chip
MicrosoftRevenue certainty through 2030, IP license through 2032Exclusive position, inbound revenue from OpenAI
Google CloudLegitimized as OpenAI infrastructure partnerNothing material
AmazonAWS deal no longer in conflict with Microsoft exclusivityNothing material

Google Cloud benefits most cleanly. It was already running ChatGPT workloads. The amendment turns that from a legal gray area into a formal arrangement. Amazon's $50 billion Trainium commitment to OpenAI, which was the original trigger for Microsoft's lawsuit consideration, is now unencumbered.

Microsoft's stock declined on the announcement. The market's read: what Microsoft gives up is worth more than what it gets.

Microsoft stock decline on deal news Microsoft shares fell after the announcement as investors priced in the loss of exclusivity. Source: finance.yahoo.com

Where It Falls Short

The deal is presented as mutual flexibility, but the structure still favors Microsoft in the medium term. OpenAI owes Microsoft revenue through 2030. Azure gets first access to new products, which means the default for enterprise customers building on OpenAI via Azure stays intact. And Microsoft's non-exclusive IP license through 2032 means it can keep building Azure AI services on OpenAI's models while competing against OpenAI's direct sales.

The "cannot or chooses not to" carve-out for routing workloads elsewhere sounds useful, but it requires OpenAI to show Microsoft's technical inadequacy before exercising it - which is the kind of claim that invites disputes. In practice, routing workloads away from Azure's preferred status will take operational choices, not just legal permission.

There's also the shareholder angle. Microsoft remains a major OpenAI investor. That relationship creates alignment incentives that outlast any specific contract term and constrain how aggressively OpenAI will compete with Azure directly.


The test for this amendment is whether OpenAI now routes meaningful production traffic away from Azure. That would show up in Azure's AI revenue disclosures over the next two quarters - and it'd tell you whether this is an infrastructure shift or a paperwork one.

Sources:

Sophie Zhang
About the author AI Infrastructure & Open Source Reporter

Sophie is a journalist and former systems engineer who covers AI infrastructure, open-source models, and the developer tooling ecosystem.