Oracle insists its $300B contract with OpenAI is on schedule

""Site selection and delivery timelines were established in close coordination with OpenAI following execution of the agreement and were jointly agreed. There have been no delays to any sites required to meet our contractual commitments, and all milestones remain on track," Oracle told El Reg and other outlets in a statement. "We remain fully aligned with OpenAI and confident in our ability to execute against both our contractual commitments and future expansion plans.""

"Oracle, which entered into the $300 billion cloud contract with OpenAI in September, is on the hook for roughly 4.5 gigawatts of compute capacity over the next five years or so. The project is expected to contribute $30 billion to Oracle's revenue annually beginning in 2027. The timing of the contracts means even if Oracle had pushed some datacenter projects originally penciled in for 2027 to 2028, the company could still argue it's on track to deliver on time."

"It should also be noted that Oracle, as a general rule, doesn't build or own the datacenters it operates. For example, OpenAI's flagship Stargate datacenter in Abilene, Texas, is built and owned by Crusoe, but furnished and operated by Oracle. This means Oracle's datacenter schedule isn't entirely within its control. It can only begin deploying IT infrastructure when the shells are complete and power and plumbing have been hooked up."

Oracle insists datacenter site selection and delivery timelines were jointly agreed with OpenAI and that no required sites have been delayed, with all milestones on track. Oracle entered a $300 billion cloud contract with OpenAI in September and is committed to roughly 4.5 gigawatts of compute capacity over the next five years. The project is expected to contribute about $30 billion to Oracle's annual revenue beginning in 2027. Oracle generally does not build or own datacenter shells; third parties may build and own facilities while Oracle furnishes and operates them. Datacenter deployment depends on shell completion, power, and plumbing readiness. Approximately half of modern AI datacenter cost is land and infrastructure, with the remainder split among network, storage, and compute.

Read at Theregister