Oracle Corp. has postponed the completion timeline for several of the large-scale data centers it is building for OpenAI, shifting projected readiness dates from 2027 to 2028. Individuals familiar with the projects said the delays stem primarily from labor shortages and material constraints affecting multiple construction sites across the United States.
The data centers are part of Oracle’s expansive $300 billion agreement to provide the computational infrastructure needed to train and operate OpenAI’s advanced artificial intelligence models. Despite the adjusted timelines, these U.S. facilities remain among the most ambitious hyperscale data center projects currently underway worldwide.
Oracle emphasized that the company is still positioned to meet its formal obligations. In a statement, a spokesperson said that no delays affect sites required to fulfill contractual commitments and that all key milestones remain on schedule. The spokesperson added that the planning and timing of facility buildouts were developed in close coordination with OpenAI. OpenAI declined to comment on the matter.
The market reaction was swift, with Oracle shares falling more than 4% in afternoon trading on Friday in New York.
During this week’s earnings call, Oracle Co-Chief Executive Clay Magouyrk said the company continues to pursue aggressive capacity expansion goals across global markets. He highlighted that construction of the first dedicated OpenAI data center in Abilene, Texas, is moving forward as planned, with more than 96,000 Nvidia GPUs already delivered to the site.
While the extended timelines add pressure to one of the most closely watched AI infrastructure partnerships in the world, Oracle maintains that its overall strategy and delivery commitments remain intact. The delayed projects underscore the growing challenges faced by hyperscale builders as demand for AI-optimized infrastructure surges and supply chain constraints persist.
About The Author
