Oracle to Build Supercluster of 130,000 NVIDIA GPUs, Plans Nuclear-Powered Data Centers

Oracle announced plans to deploy a supercluster of 130,000 NVIDIA GPUs and build up to 2,000 new data centers, some powered by small nuclear reactors. The initiative aims to meet the rising demand for AI and cloud services, pushing Oracle’s AI capabilities to new heights.

Oracle Reports Surging Demand for AI Cloud Services, Plans Massive Expansion of Data Centers

Oracle's most recent earnings call has been a revelation to those who remain skeptical of artificial intelligence. According to Wccftech, the company has made a concerted effort to demonstrate that the demand for data centers continues to exceed the supply. Furthermore, Oracle's forthcoming facilities are mind-blowing, as they utilize NVIDIA's GPUs on a scale that has never been observed.

Oracle's results for the quarter ending August 31st were satisfactory. In both its top-line and bottom-line metrics, it exceeded the consensus expectations. Additionally, the company's AI-critical cloud division experienced a 21% year-over-year increase.

Oracle impressed investors during the earnings call by announcing that the demand for its products and services, particularly in the cloud sector, was "still outstripping supply." The company addresses this deficit by "laying out a lot of supply."

Oracle is constructing numerous data centers and introducing services like GenAI Agents. These agents combine retrieval-augmented generation technology with large language models (LLMs), enabling clients to interact directly with their cloud-stored data in novel and innovative ways.

Oracle Projects $100 Billion AI Model Development Cost, Announces Zettascale Supercluster and Data Center Expansion

Oracle has estimated that the cost of constructing a foundational model from the ground up, such as ChatGPT or Google's Gemini, will exceed $100 billion in the next four to five years.

Oracle believes that the inference process will be the source of the subsequent significant increase in AI-related expenditures.

"So that goes on, and we'll see more and more applications look at that. So, I wouldn't -- if your horizon is over the next 5 years, maybe even the next 10 years, I wouldn't worry about, hey, we've now trained all the models we need and all we need to do is inferencing. I think this is an ongoing battle for technical supremacy that will be fought by a handful of companies and maybe one nation-state over the next 5 years at least, but probably more like 10. So, this business is just growing larger and larger and larger. There's no slowdown or shift coming."

Oracle's medium-term prospects are, of course, optimistic in light of this expanding demand paradigm.

Additionally, Oracle has recently announced a "zettascale" supercluster system that "can scale up to 131,072 Blackwell GPUs with NVIDIA ConnectX-7 NICs for RoCEv2 or NVIDIA Quantum-2 InfiniBand networking to deliver an astounding 2.4 zettaflops of peak AI compute to the cloud."

This new supercluster is anticipated to be operational by the end of the year and will assist clients in training and deploying next-generation AI models.

Lastly, it is essential to mention that Oracle is preparing to construct up to 2,000 data centers, representing a significant increase from its current capacity of 160. Additionally, a single facility of this nature will require a minimum of one "gigawatt" of electricity and may be propelled by as many as three small-scale nuclear reactors (SMRs).