OpenAI is reevaluating its artificial intelligence infrastructure strategy. Reportedly, the company plans to abandon parts of its expansion project for the "Stargate" data center with Oracle in Abilene, Texas, and instead seek to deploy more advanced next-generation AI chips.
According to informed sources, the data center was originally planned to use Nvidia's Blackwell GPU architecture, but the related power infrastructure is expected to take about a year to connect. OpenAI believes that by the time the project is completed, more powerful next-generation GPUs might already be available on the market, so it is reevaluating the deployment plan.
The report indicates a clear mismatch between the data center construction cycle and the GPU technology update pace. As the demand for AI computing power grows rapidly, the performance iteration cycle of new-generation chips may outpace the progress of large-scale infrastructure construction.
Meanwhile, Oracle is supporting its AI infrastructure expansion through debt financing. Market insiders say the company now has over $1 billion in booked debt, while its free cash flow is currently negative.
