Microsoft’s January 21, 2025 announcement did not end its OpenAI partnership or move OpenAI’s API off Azure. It changed who could provide new cloud capacity: Microsoft received a right of first refusal, giving it the first opportunity to meet OpenAI’s needs while leaving OpenAI room to add capacity elsewhere. Later agreements went further. As of August 18, 2026, Microsoft says it remains OpenAI’s primary cloud partner, but OpenAI can serve its products across other clouds and Microsoft’s license to OpenAI intellectual property is nonexclusive.
What Microsoft announced in January 2025
Microsoft’s January 21, 2025 announcement changed the arrangement for additional computing infrastructure, not the whole relationship. Microsoft said it would have a right of first refusal (ROFR) on OpenAI’s new cloud capacity and approved OpenAI building additional capacity, particularly for research and model training. OpenAI said it would continue increasing its Azure consumption. Microsoft’s announcement
The word “new” matters: the announcement did not say existing OpenAI workloads were moving off Azure. Nor did it state that Azure was technically unable to meet a particular need, disclose a capacity shortfall, or identify a pricing dispute. It described a more flexible way to obtain future capacity.
What a right of first refusal meant
Operationally, OpenAI could bring a request for additional capacity to Microsoft first. Microsoft would have the opportunity to provide it; if Microsoft could not or did not provide what OpenAI needed, OpenAI could pursue other arrangements. That priority was not the same as an unconditional Microsoft monopoly over all future compute.
#1 Best Overall
The public announcement did not specify response deadlines, pricing, capacity thresholds, or the precise conditions under which Microsoft could decline. The agreement’s full mechanics were not made public.
What remained in place at the time
Microsoft said the OpenAI API would remain exclusive to Azure, run on Azure, and be available through Azure OpenAI Service. It also said its access to OpenAI intellectual property for products such as Copilot and the revenue-sharing arrangements would continue. The announcement said core partnership elements extended through 2030. These were distinct from the new-capacity arrangement: API exclusivity could remain even as OpenAI gained a route to obtain additional infrastructure elsewhere. Microsoft’s January 2025 statement
Rank #2
- Server 2022 Standard 16 Core
Why Stargate made the capacity change significant
OpenAI announced Stargate on the same day as Microsoft’s partnership update. It described an intention to invest up to $500 billion over four years, beginning with $100 billion, in U.S. AI infrastructure. Those figures were an announced plan, not evidence that the full amount had already been spent or deployed. OpenAI named SoftBank, OpenAI, Oracle, and MGX as initial equity funders, and Arm, Microsoft, NVIDIA, Oracle, and OpenAI as initial technology partners. It said it would continue increasing Azure consumption alongside the additional compute. OpenAI’s Stargate announcement
Microsoft was therefore not excluded from Stargate, but neither was it the sole infrastructure provider. The project made the practical importance of more flexible capacity easier to see: OpenAI could pursue a larger infrastructure program involving multiple partners while retaining Azure use. Being named a technology partner does not establish that Microsoft hosted every Stargate workload, and the announcements do not provide a complete allocation of workloads among providers.
How the partnership changed after January 2025
| Date | What the parties said | Why it matters |
|---|---|---|
| October 28, 2025 | OpenAI said it had contracted to purchase an incremental $250 billion of Azure services and that Microsoft no longer had a right of first refusal to be OpenAI’s compute provider. It also said Microsoft’s model and product IP rights were extended through 2032, and described Azure API exclusivity until AGI under that agreement. OpenAI’s October 2025 announcement | The January ROFR arrangement was no longer the stated compute framework. The October agreement also described third-party product development, non-API products being served on any cloud, and the possibility of qualifying open-weight models; these were terms of that stage, not a substitute for the later amendment. |
| February 27, 2026 | A joint statement described Azure as the exclusive cloud provider for stateless OpenAI APIs and said OpenAI’s first-party products would continue to be hosted on Azure, while acknowledging flexibility to commit to additional compute elsewhere. The joint statement | This was an interim description of the relationship before the April amendment. It should not be read as the latest terms. |
| April 27, 2026 | OpenAI said Microsoft remained its primary cloud partner; OpenAI products would ship first on Azure unless Microsoft could not or chose not to provide the required capabilities; and OpenAI could serve all products across any cloud. Microsoft’s license to OpenAI IP became nonexclusive. Microsoft no longer pays a revenue share to OpenAI, while OpenAI’s revenue-share payments to Microsoft continue through 2030, subject to a cap. OpenAI’s April 2026 announcement | This is the latest major amendment described in the official announcements available as of August 18, 2026: Microsoft is primary, not exclusive, and OpenAI has broad multicloud serving rights. |
OpenAI has also described Stargate as involving data-center partnerships with Oracle, SoftBank, CoreWeave, and others, while saying Microsoft would continue providing cloud services, including through Stargate. That does not establish that every Stargate workload runs on Oracle or that workloads can be transferred freely among providers. OpenAI’s Stargate and Oracle update
What the changes mean for Azure customers and developers
The January 2025 announcement did not announce a customer migration or a change to Azure OpenAI Service controls. Azure OpenAI Service remains a relevant route for customers who want Azure-native integration; OpenAI’s direct products and API are separate routes, and the current agreement also allows OpenAI products to be served across clouds. Rights between OpenAI and its infrastructure partners do not, by themselves, tell a customer which products, models, or regions can be purchased.
Choose the route based on the workload, not the headline
- Azure OpenAI Service: Consider it when Azure identity, networking, governance, compliance arrangements, procurement, or billing are central to the deployment. Confirm the model, quota, region, and service details for the specific offer.
- OpenAI directly: Consider direct OpenAI access when the first-party product or API relationship is the priority and Azure-specific controls are not a requirement.
- Another cloud or infrastructure provider: Consider this when the workload is already integrated there or the provider offers required capacity, geography, networking, or specialized infrastructure. Confirm that the exact OpenAI product is available under current terms.
Do not assume that a product being served on another cloud means identical pricing, latency, model versions, data handling, regional availability, quotas, safety controls, or service guarantees. Check the terms and technical documentation for the exact service and deployment before committing.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What it means for Microsoft and competing providers
Analysis: Microsoft retains a substantial commercial and technical relationship with OpenAI, a primary-cloud position, Azure integration, and a continuing IP license. But it no longer has the same control over every unit of OpenAI compute that sole-provider status would imply. A nonexclusive license also means its access to OpenAI IP is not exclusive. Other infrastructure providers may compete for OpenAI workloads, but public announcements do not establish who hosts each workload or that one provider is a drop-in replacement for another.
Recommended Free Tools
That makes “Did Microsoft lose OpenAI?” the wrong binary question. Microsoft lost its sole-provider position for new capacity in January 2025, initially kept priority through ROFR, and OpenAI said in October 2025 that the ROFR had ended. By April 2026, Microsoft remained the primary cloud partner while OpenAI had broad multicloud serving rights. The partnership persisted, but its boundaries changed.
What the public announcements do not settle
The company statements provide summaries, not the full contracts. They do not establish the January ROFR’s detailed procedures, capacity allocation, pricing formulas, or service-level obligations. They also do not provide a workload-by-workload map across Azure, Oracle, or other providers, or guarantee that each OpenAI service is available on every cloud. Customers should verify availability, data handling, regional coverage, and contractual commitments for the specific product they plan to use.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
