OpenAI is quietly rewriting the map of U.S. compute infrastructure—and Oracle (ORCL, Financial) is riding shotgun. The two companies have agreed to co-develop an additional 4.5 gigawatts of AI-focused data centers under the Stargate project, targeting states like Texas, Michigan, Wisconsin, and Wyoming. This comes on top of OpenAI's current buildout in Abilene, Texas, where parts of the site are already online and running workloads using Nvidia's (NVDA, Financial) newest GB200 chips. Together, these expansions could push OpenAI's total compute capacity past 5 gigawatts, powered by more than 2 million chips.
This isn't just a bigger bet—it's a faster one. OpenAI said the early success in Abilene gave it confidence to scale rapidly. That Texas facility, now partially operational, served as a real-world stress test of what “at-scale and at-speed” deployment looks like. And while the effort is linked to the trio's original $500 billion infrastructure pledge with SoftBank, OpenAI clarified that SoftBank isn't financing this new round—reportedly due to financing delays amid global economic and tariff headwinds. Oracle, meanwhile, has been delivering racks of Nvidia's chips since last month, showing its role isn't just strategic—it's operational.
For investors, this could mark a pivotal shift in Oracle's positioning. OpenAI expects the new capacity rollout to generate over 100,000 U.S. jobs in construction and operations—a headline-grabbing figure, even if long-term staffing remains modest. But the real story could be Oracle's deepening role as a critical infrastructure provider in the AI gold rush. With demand for AI compute infrastructure scaling faster than ever, Oracle may be setting itself up as one of the biggest behind-the-scenes winners.
