Oracle Corp. has delayed completion dates for several of the data centers it is building for OpenAI, shifting targets from 2027 to 2028 due to significant labor and materials shortages, according to a Bloomberg report citing people familiar with the projects. The postponements mark a notable setback in one of the world’s largest and most ambitious AI infrastructure efforts.
The data centers are part of a $300 billion agreement signed earlier this year in which Oracle committed to supplying the computing capacity required to train and deploy OpenAI’s next-generation models. Even with the revised timeline, sources told Bloomberg the schedules remain highly aggressive, as the US sites are designed to be among the largest data centers ever constructed.
Both Oracle and OpenAI declined to comment on the delays. However, Oracle Co-Chief Executive Officer Clay Magouyrk said during an earnings call this week that the company maintains “ambitious achievable goals for capacity delivery worldwide.” He noted that the first OpenAI-dedicated facility in Abilene, Texas, remains on track, with more than 96,000 Nvidia chips already delivered and integrated into the build.
The setback underscores the immense strain global supply chains are under as tech companies race to build AI infrastructure at unprecedented scale. While Oracle works to overcome shortages and logistical bottlenecks, the revised timeline signals just how complex and resource-intensive next-generation AI facilities have become.
