Despite Wall Street jitters and reports to the contrary, Oracle insists its $300 billion datacenter deal with OpenAI is on track and proceeding on schedule.
Oracle denied a Bloomberg report that Big Red had delayed some of its datacenter projects until 2028 — a full year later than originally planned.
"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 co…
Despite Wall Street jitters and reports to the contrary, Oracle insists its $300 billion datacenter deal with OpenAI is on track and proceeding on schedule.
Oracle denied a Bloomberg report that Big Red had delayed some of its datacenter projects until 2028 — a full year later than originally planned.
"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.
As we understand it, roughly half of the overall cost of a modern AI datacenter is land and infrastructure, with the remainder being split between network, storage, and compute — the latter being by far the largest.
For Oracle’s gamble to work, it needs to rent that compute out to OpenAI or other customers for more than it costs to finance the hardware and pay its datacenter leases.
- Oracle raises AI spending estimate, spooks investors
- AI datacenter boom could end badly, Goldman Sachs warns
- AI mania to swell datacenter capex to $1.6T by 2030 – if the bubble doesn’t pop first
- Amazon’s Trainium3 is the latest to conform to Nvidia’s mold
The up-front costs required to realize a profit later are significant. As of the second quarter of Oracle’s 2026 fiscal year, the company was sitting on more than $100 billion in debt, which is expected to grow considerably before all is said and done.
On the company’s Q2 earnings call last week, Oracle revealed its 2026 capex spend would be $15 billion higher than previously forecast at $50 billion. This clearly was not what investors wanted to hear and Oracle’s share price plunged in after hours trading as a result.
Oracle’s rampant spending has led some financial analysts to raise alarm bells. In late November, Morgan Stanley suggested that investors should short Oracle stock.
The cloud provider and database stalwart has attempted to assuage fears of a default. Speaking on last week’s earnings call, Oracle’s Principal Financial Officer (PFO) Doug Kehring advised investors that Oracle has several avenues to secure capital and was exploring provisions to allow customers to deploy their own chips in its datacenters, reducing the company’s financial liabilities. ®