Oracle throws full weight into AI cloud race with data center build outs for OpenAI

Oracle is spending tens of billions of dollars to build massive data centers for OpenAI, as part of a huge deal announced under President Donald Trump’s administration at the White House in January.

This is Oracle’s biggest push yet into artificial intelligence infrastructure, and it’s reshaping the company’s future in the global cloud market.

The contract with OpenAI includes over 5 gigawatts of compute power, enough to run millions of homes, and the facilities are set to be online as early as summer 2026, with full completion targeted by early 2027, according to Bloomberg.

Oracle’s site in Shackleford County, Texas will run entirely on gas generators, burning more than $1 billion per year to stay operational because local power infrastructure can’t deliver what’s needed.

The site is being developed by Vantage Data Centers, owned by DigitalBridge, and will hit 1.4 gigawatts in capacity, making it one of the largest in the world. These builds are reportedly optimized for training and running advanced AI models, a process that requires thousands of Nvidia GPUs, extreme cooling, and nonstop power.

Oracle bets everything on cloud after years of internal battles

Oracle’s push into AI cloud started after a long stretch of confusion and failed projects. Back in 2008, Oracle’s co-founder and current chairman Larry Ellison called cloud computing “complete gibberish.”

For years, he resisted the change because Oracle’s traditional database business, which sold on-premise licenses, was highly profitable. But as competitors like Amazon and Salesforce started eating into Oracle’s base, Ellison changed course.

Through the 2010s, Oracle ran two competing internal projects to get into cloud. One was led by Thomas Kurian, who launched a product in 2012 that flopped.

Another group pushed a “bare metal” approach, where customers don’t share servers. That idea won Ellison’s support and became Oracle Cloud Infrastructure (OCI). Kurian left in 2018 and now runs Google Cloud.

OCI’s design helped Oracle grow in countries where big players hadn’t expanded yet. It also appealed to customers who didn’t want to share infrastructure.

Oracle’s salespeople promoted lower prices and simpler product offerings while mocking AWS’s sprawling catalog and side projects. AWS pushed back, saying more customers choose them because of “unmatched choice, pace of innovation, security, and reliability.”

Inside Oracle, getting CEO Safra Catz on board was tough. She worried about cloud’s low margins and huge upfront costs. But once TikTok came knocking, everything changed. The ByteDance-owned company, under scrutiny in the U.S. over ties to China, needed a domestic cloud partner.

Oracle was seen as a politically safe pick due to its national security background and Republican ties. In 2022, TikTok announced that all U.S. traffic would flow through Oracle servers.

Oracle’s growth hinges on Clay Magouyrk’s division and OpenAI’s demand

Clay Magouyrk, a blunt and aggressive executive who once told another Oracle leader their decision was “f—king stupid” back in 2021, is now leading OCI.

Magouyrk joined OCI early after a stint at Amazon and moved Oracle’s cloud operations to Seattle, outside the traditional Silicon Valley bubble. The division, nicknamed “Sparta,” offered better pay and copied Amazon’s practice of firing low performers often. Today, 23,000 employees report to him.

Magouyrk was promoted to president in June 2025, and many inside the company believe he could eventually take over from Ellison, who is now 81. Magouyrk himself said on the Modern CTO podcast, “Sometimes you look back and say — I’m not sure that I was entirely qualified at each step of the way. But it seems to have worked out so far.”

Oracle has poached over 600 Amazon workers in the last two years, helped by Amazon’s strict return-to-office policy. Oracle still allows hybrid and remote setups, which is appealing to cloud engineers.

Oracle has also signed cloud deals with Zoom and Uber, while serving Nvidia, which uses Oracle for both internal projects and its own cloud services.

Nvidia rents capacity from an Oracle GPU cluster in Japan and another site being built on Batam Island in Indonesia. Oracle is also in talks with Meta and Elon Musk’s xAI for future capacity deals.

Join Bybit now and claim a $50 bonus in minutes