
Microsoft's $3 billion Oracle computing power lease falls through due to security compliance issues; Goldman Sachs warns: AI infrastructure has reached a "vulnerable moment."
Microsoft abandons a $3 billion computing power leasing agreement with Oracle due to security compliance issues, which Oracle denies. This incident highlights the shortage of computing power and intensified competition amid the surge in demand for AI infrastructure. Goldman Sachs warns that AI infrastructure is entering a "fragile moment."
According to reports, Microsoft (MSFT.US) has abandoned a $3 billion partnership agreement to lease cloud computing power from Oracle (ORCL.US) due to security concerns. In response, Oracle stated on Tuesday that the report is inaccurate. Microsoft declined to comment.
Insiders previously revealed that Microsoft had recently negotiated with Oracle to lease the latter's cloud infrastructure, but the deal ultimately fell through due to security and compliance issues. Insiders indicated that the value of the deal could exceed $3 billion.
The breakdown of negotiations highlights an increasingly severe reality brought about by the booming development of artificial intelligence (AI): even the world's largest technology companies are facing a shortage of computing power. As demand for AI services surges, competition among cloud service providers like Microsoft intensifies, as they compete not only for customers but also for the infrastructure and capacity needed to run their own products.
This scramble for computing power has led to a rare wave of collaborations, power-sharing agreements, and infrastructure deals worth hundreds of billions of dollars.
Microsoft recently disclosed that its capital expenditures will reach $190 billion in 2026, primarily to expand data center capacity. The company has sought assistance from Amazon to increase capacity for its GitHub code development business in response to recent service interruption issues.
Insiders revealed that Microsoft is seeking to reach one or more agreements with other cloud service providers to prioritize its customers' access to Azure cloud computing resources. One insider stated, "We are looking around for leaseable computing resources."
Reports indicate that Microsoft originally planned to migrate some business loads to Oracle Cloud Infrastructure (OCI), but Oracle's public cloud has not passed the Federal Risk and Authorization Management Program (FedRAMP) certification. This security framework is designed to ensure that cloud services are secure enough to handle U.S. government data. Insiders noted that Oracle was unwilling to invest in the necessary modifications to obtain this certification at that time.
A senior Oracle executive reportedly stated that adding FedRAMP to Oracle's public cloud (as opposed to its government cloud, which already complies with the framework) would be a significant engineering challenge.
An Oracle spokesperson stated, "The details mentioned in the report are not accurate," but did not specify what was inaccurate. The spokesperson added, "Microsoft is both an OCI partner and our customer. We have a highly collaborative and productive relationship and frequently discuss how to further expand our existing projects."
Insiders revealed that Microsoft is still evaluating and exploring various options for leasing cloud infrastructure. Amazon and Google's public clouds have passed FedRAMP certification.
Other large tech companies are also engaging in similar deals. SpaceX and Google recently announced a new agreement in which Google will pay SpaceX $920 million per month for AI computing power from October 2026 to June 2029.
As of Tuesday's market close, Oracle was down 2.24%, and Microsoft was down 1.48%.
It is worth noting that the market has been highly sensitive to news regarding AI infrastructure recently. The previous week, AI infrastructure company Crusoe paused its 1.8GW data center project in Wyoming at the request of a client, leading to a significant drop in related stocks Goldman Sachs partner Rich Privorotsky warned that as market positions, leverage, and AI capital expenditures become deeply intertwined, the risks of the AI ecosystem "can no longer be ignored." Even minor delays, postponements, or changes in priorities are enough to compel investors to reassess their assumptions about future demand
