Full Breakdown
OpenAI Ends Microsoft Exclusivity and Launches Models on AWS Bedrock
4/30/2026, 1:09:01 AM
Core Event
On April 27 2026 OpenAI and Microsoft announced a revised partnership that ends Microsoft’s exclusive right to sell OpenAI’s models. The amendment allows OpenAI to offer its products on any cloud provider. One day later, Amazon Web Services (AWS) unveiled that OpenAI’s latest foundation models, the Codex coding agent, and a new “Bedrock Managed Agents” service will be available on AWS’s Bedrock platform.
Background, Key Figures & Timeline
- 2016-2024: OpenAI’s large-scale training runs on Microsoft Azure; Microsoft invests $13 billion and holds a 27 % stake in the for-profit arm.
- Nov 2025: OpenAI announces a $38 billion commitment with AWS.
- Feb 2026: Amazon invests $50 billion in OpenAI; OpenAI agrees to spend an additional $100 billion on AWS compute, including 2 GW of Trainium chips.
- Apr 27 2026: Revised Microsoft-OpenAI agreement removes exclusivity, keeps Azure as the primary cloud and retains a non-exclusive IP license through 2032; revenue-share payments to Microsoft continue through 2030 under a capped total.
- Apr 28 2026: AWS launches OpenAI models on Bedrock in limited preview, with GPT-5.4 now available and GPT-5.5 slated for the coming weeks.
Key actors: OpenAI CEO Sam Altman, CRO Denise Dresser, Microsoft CEO Satya Nadella, Amazon CEO Andy Jassy, AWS CEO Matt Garman, UBS analysts, and OpenAI head of business Steve Sharpe.
Data & Statistics
- Amazon’s investment: $50 billion (largest single external investment for OpenAI).
- OpenAI’s spend on AWS: $100 billion over eight years, including 2 GW of Trainium accelerator capacity.
- Codex serves >4 million weekly active users.
- AWS AI services generated > $15 billion annualised revenue in Q1 2026, with growth > 20 % YoY.
- Microsoft’s Azure commitment from the 2025 recapitalisation: $250 billion of Azure services.
Why It Matters
The shift dismantles a decade-long de-facto monopoly on frontier AI models, intensifying competition between the two largest cloud providers. Enterprises can now access OpenAI’s models within the security, governance, and cost-control framework of AWS, reducing data-silo concerns and procurement friction. The move also pressures Microsoft to broaden its AI portfolio (e.g., deeper ties with Anthropic) while preserving Azure’s “first-dibs” status.
Official Statements & Responses
- OpenAI emphasized that the new arrangement “expands access to our models for customers where their data and workloads already reside.”
- AWS highlighted that the integration “delivers the best models in the best environments for customers to be successful.”
- Microsoft framed the amendment as “greater predictability and flexibility” while retaining Azure as the primary launch platform.
- Amazon called the partnership “very interesting” and promised rapid rollout of the models on Bedrock.
- Denise Dresser reiterated that the prior Microsoft partnership “has also limited our ability to meet enterprises where they are.”
Criticism & Opposition
UBS analysts warned that “Microsoft appears to have made more concessions than gains,” suggesting the revised terms may weaken Microsoft’s strategic position. Industry observers noted that OpenAI recently missed internal revenue and user targets, raising doubts about the startup’s ability to meet its massive AWS spending commitments. Capacity constraints remain a concern; AWS executives acknowledge ongoing demand outpacing available compute.
Conflicting Reports & Gaps
Sources differ on the exact mechanics of the revenue-share cap: some describe a fixed-percentage cap through 2030, while others note the cap amount is undisclosed. The timeline for full public availability of GPT-5.5 on AWS is described as “in the next couple of weeks” by AWS leadership, yet other reports leave the date unspecified. Details on how Azure will retain “first-dibs” on new products are vague across statements.
Verbatim Quotes
- “This is what our customers have been asking us for for a really long time,” — Matt Garman, AWS CEO
- “I wish I could be there with you in person today, my schedule got taken away from me today,” — Sam Altman, OpenAI CEO (recorded message)
- “The two are not related in any way,” — Denise Dresser, OpenAI CRO
- “on an extremely steep growth curve across consumer, enterprise and developers.” — Steve Sharpe, OpenAI head of business and financial communications
What’s Next
OpenAI plans to roll out GPT-5.5 on Bedrock within weeks, expand Trainium-based compute capacity, and explore additional collaborations with other cloud providers such as Google Cloud. Microsoft will continue to prioritize Azure for initial releases while negotiating further AI-related investments with Anthropic and other partners.
