Microsoft was recently in talks with Oracle about leasing the company’s cloud infrastructure, but the deal fell through due to security and compliance concerns, according to people familiar with the matter.
One of the people said that the deal could have been worth more than $3 billion.
The failed talks highlight a growing reality of the AI boom: even the world’s largest technology companies are running short on computing power. As demand for AI services soars, cloud providers like Microsoft are increasingly competing not just for customers but for the infrastructure and capacity needed to run their own products.
That scramble is driving an unusual wave of partnerships, capacity-sharing agreements, and multibillion-dollar infrastructure deals as companies race to secure enough computing resources to support the next generation of AI.
Microsoft recently projected that its capital expenditures for the 2026 calendar year will reach $190 billion, largely to expand data center capacity. The company has already turned to Amazon to add capacity for its GitHub code development business to address recent outages.
Microsoft is seeking a deal or deals with other cloud providers to prioritize its own Azure cloud computing resources on customers, the people said. “We are shopping for capacity everywhere,” one of the people said.
The plan was to move some Microsoft workloads to Oracle Cloud Infrastructure, but Oracle’s public cloud did not have the Federal Risk and Authorization Management Program (FedRAMP), a standardized security framework that ensures cloud services are secure enough to handle U.S. government data. Oracle was not willing to add this framework, one of the people said.
“The details mentioned in the article are inaccurate,” an Oracle spokesperson said, declining to specify the inaccuracies. Microsoft is both an OCI partner and a customer. We have a tremendously collaborative and fruitful partnership, where we often talk about ways we can expand upon our ongoing work together.”
Microsoft declined to comment.
An Oracle executive told Business Insider that adding FedRAMP to Oracle’s public cloud (versus its government cloud, which already meets it) would be a massive engineering lift.
Microsoft is still evaluating and exploring options for leasing cloud infrastructure, the people familiar with the talks said. Amazon and Google’s public clouds have FedRAMP.
Other big tech companies have been making similar deals.
SpaceX and Google recently disclosed a new deal in which Google will pay SpaceX $920 million a month for AI compute capacity from October 2026 to June 2029. That emerged just two months after Google’s own cloud business agreed to sell AI compute capacity to Anthropic.
Have a tip? Contact this reporter via email at astewart@businessinsider.com or Signal at +1-425-344-8242. Use a personal email address and a nonwork device; here’s our guide to sharing information securely.
