Oracle is making a massive $40 billion bet on AI infrastructure, investing in Nvidia’s powerful GB200 superchips for its new Stargate data center in Abilene, Texas. The state-of-the-art facility will primarily serve OpenAI, which plans to harness this immense computing power for AI training workloads. In total, Oracle will receive around 400,000 GB200 chips, combining Nvidia’s Blackwell GPUs and Grace CPUs in NVL72 systems capable of delivering up to 1.4 exaFLOPS each.
However, energy consumption is a looming challenge. The data center will feature 11,000 racks, each drawing up to 120 kilowatts. Total power demands are projected to exceed 1.5 gigawatts, raising concerns about whether the local grid can handle the load. AI infrastructure’s skyrocketing power needs continue to test energy systems nationwide.
A phased rollout will see just 200 MW online this year, enough for 54,000 superchips, with the remainder due by 2026. The deal hints at OpenAI’s growing independence from Microsoft as it locks in long-term infrastructure commitments with Oracle.
