Oracle to spend $40B on Nvidia chips to power OpenAI's Texas megacenter
The cloud provider will buy roughly 400,000 GB200 GPUs for the Abilene data center under a 15-year deal.
Oracle plans to spend roughly $40 billion on Nvidia's flagship GB200 GPUs — about 400,000 chips — to power OpenAI's new data center in Abilene, Texas, under a 15-year arrangement. The facility, part of the Stargate effort, is slated to deliver 1.2 gigawatts of power when it comes online by mid-2026.
The chip buy is one piece of a far larger Oracle infrastructure push. The company has told investors it intends to raise $45 billion to $50 billion in gross cash during 2026 to meet contracted demand from a customer list that includes AMD, Meta, Nvidia, OpenAI, TikTok and xAI. Oracle is also hedging its silicon supply, agreeing to deploy 50,000 AMD Instinct MI450 GPUs starting in the third quarter.
Why it matters
Texas has become the physical home of the AI buildout, and Abilene is its flagship. A single tenant committing $40 billion to one site's silicon shows how AI infrastructure now gets financed at the scale of national pipelines and ports. It also reshapes the regional power and water debate: gigawatt-class campuses are landing in West Texas faster than the grid conversation can keep up.
Oracle is expanding its cloud infrastructure business to meet contracted demand from customers including AMD, Meta, Nvidia, OpenAI, TikTok and xAI.— Oracle, investor disclosure
Reported by Next in Austin. Based on reporting from Network World.
