Oracle is preparing to make one of the largest AI infrastructure investments in history, with plans to spend $40 billion on Nvidia’s latest GB200 chips. The chips will support a colossal new AI data centre in Abilene, Texas, being developed as part of OpenAI’s ambitious Stargate project.
The deal underscores the skyrocketing demand for high-performance computing as companies like OpenAI race to build next-generation artificial intelligence models. With Oracle stepping in to supply the core computing infrastructure, this move signals a dramatic shift in how AI powerhouses scale outside traditional cloud providers.
The Texas site is being called the first “Stargate” data centre, an initiative backed by OpenAI and SoftBank to create some of the world’s most powerful AI facilities. When complete in 2026, the Abilene centre will deliver a staggering 1.2 gigawatts (GW) of computing power — enough to place it among the world’s largest AI hubs.
The facility will house eight buildings across a massive campus, with Crusoe and Blue Owl Capital serving as its owners. The two firms have raised $15 billion through debt and equity, including a $9.6 billion loan package led by JPMorgan, to fund construction.
Oracle has signed a 15-year lease on the site and plans to deploy approximately 400,000 Nvidia GB200 chips — the company’s most advanced processors designed specifically for AI training and inference. Rather than owning the infrastructure, OpenAI will lease computing power from Oracle, making this a strategic alignment rather than a direct infrastructure investment.
Beyond Microsoft: OpenAI’s Push for Independence
This megaproject is more than a hardware upgrade. It reflects OpenAI’s pivot away from Microsoft, its long-time cloud partner. Most of Microsoft’s $13.7 billion investment into OpenAI came via Azure cloud credits, but insiders say OpenAI has grown increasingly frustrated with Microsoft’s inability to meet rising compute demands.
The two firms recently agreed to end their exclusivity arrangement, and discussions are ongoing around the future licensing rights to OpenAI’s models. By partnering with Oracle and others, OpenAI is now creating an independent infrastructure network that allows it to scale at will — and on its own terms.
The Abilene site is just one part of Stargate’s broader global vision. The initiative, launched earlier this year, plans to raise as much as $500 billion over four years to fund large-scale AI infrastructure. So far, OpenAI and SoftBank have each pledged $18 billion, while Oracle and Abu Dhabi’s MGX have committed $7 billion each.
While Stargate hasn’t directly invested in any single data centre project yet, the Texas facility aligns closely with its goals. More significantly, expansion is already underway. Just last week, OpenAI announced a new Stargate site in Abu Dhabi, unveiled during President Trump’s visit to the Gulf region. That project, led by local AI firm G42, will offer 5GW of computing power across 10 square miles — enough to run over 2 million Nvidia GB200 chips.
Rising Competition Among AI Data Giants
Oracle isn’t the only tech titan building AI superclusters. Elon Musk is scaling a separate initiative in Memphis, Tennessee, called “Colossus.” His site is expected to house 1 million Nvidia chips, mainly H100s and H200s, with the goal of becoming the first gigawatt-scale AI training supercluster.
Meanwhile, Amazon is developing a data centre in Virginia that will also surpass 1GW, making it part of the elite club of hyper-scale AI infrastructure projects.
The surge in demand for AI chips is accelerating at breakneck speed. As companies push to develop more powerful models, the bottleneck is no longer talent or data — it’s compute capacity. Nvidia’s chips have become the backbone of this new industrial era, and Oracle’s $40 billion bet confirms that the next war for AI dominance will be won in the data centre trenches.
