Oracle and Microsoft recently discussed a deal that would see the companies rent servers from each other should they run out of capacity for large-scale artificial intelligence cloud customers.

The Information reports that a decision has not been made on the deal, aimed at addressing server shortages using high-end GPUs.

The two companies announced a partnership in 2019 to allow customers to run computing jobs across both clouds, necessitating the build out of direct connection fiber links between data centers of both Oracle and Microsoft.

That existing infrastructure could be used to share GPU server capacity, a person familiar with the matter told The Information.

The publication previously reported that Microsoft has rationed access to AI servers for some of its internal teams, to ensure customers have access to the compute.

OpenAI is one of those companies. Microsoft has invested billions on the ChatGPT and Dall-E creator, in return gaining an exclusive license to offer its services over Azure cloud.

In addition to the discussions over sharing resources, Oracle has asked Microsoft if it could resell OpenAI’s software to Oracle customers similar to how it resells Microsoft’s 365 productivity suite.