HawkInsight

  • Contact Us
  • App
  • English

Oracle Partners with NVIDIA Launches First Zettascale Cloud Cluster

Oracle recently announced the launch of the world's first Zettascale-level cloud computing cluster based on NVIDIA's latest generation Blackwell platform, providing enterprises with unprecedented AI computing power.

Oracle recently announced the launch of the world's first Zettascale cloud computing cluster, based on NVIDIA's latest generation Blackwell platform, to provide enterprises with unprecedented artificial intelligence (AI) computing power.

In addition, Oracle has obtained construction licenses for three small modular nuclear power plants (SMRs), planning to provide carbon-free and stable electricity for its global AI data centers to meet the future needs of the rapid development of AI.

In the announcement on September 11, Oracle said that its cloud infrastructure (OCI) has begun accepting orders and can build a supercomputer cluster with up to 131,072 NVIDIA Blackwell GPUs for enterprises, with a computing power of 2.4 zettaFLOPS, far more than three times the world's most powerful Frontier supercomputer and six times that of other large cloud service providers. This technological breakthrough will help enterprises train and deploy next-generation AI models faster, especially in the field of generative AI.

The OCI super cluster also offers a variety of NVIDIA GPU options, including H100, H200 Tensor Core GPUs and the latest Blackwell GPUs. The H100 GPU version of the super cluster can be expanded to up to 16,384 GPUs, with a computing power of 65 ExaFLOPS. The H200 GPU version can be expanded to 65,536 GPUs, providing a computing power of up to 260 ExaFLOPS, and is expected to be launched at the end of 2024.

In addition to the existing GPU configuration, Oracle plans to launch a super cluster version equipped with NVIDIA GB200 NVL72 liquid-cooled bare metal server in the first half of 2025. Through NVLink and NVLink Switch technology, up to 72 Blackwell GPUs can be connected at high speed in the same network, with a total bandwidth of up to 129.6 TB/s, which will greatly improve the computing efficiency of AI models and provide strong support for AI research and commercial applications.

In the earnings call, Oracle founder and chairman Larry Ellison revealed that the company has obtained construction licenses for three small modular nuclear reactors (SMRs) and plans to provide the huge amount of electricity required for AI data centers.

As a new generation of nuclear energy technology, SMR has the advantages of miniaturization, mass production, rapid deployment and low operating costs. It can provide tens to hundreds of megawatts (MW) of stable electricity for large AI data centers, demonstrating Oracle's emphasis on sustainable energy. In particular, in the context of growing global energy demand, SMR lays the foundation for data centers to provide carbon-free electricity supply.

Although no SMR has been officially put into operation yet, and the trial operation is not going smoothly, Oracle's forward-looking layout shows its emphasis on future power supply security. Ellison also revealed that Oracle has 162 cloud data centers around the world, some of which have been put into operation. The largest center is equipped with a capacity of 800MW and a large number of NVIDIA GPU clusters, which can train the world's largest AI model. Oracle also plans to build a data center of more than 1GW in the future to significantly enhance its cloud computing capabilities.

As Oracle's cooperation with NVIDIA deepens, the global demand for AI computing has risen sharply, and Taiwan's technology supply chain has also benefited greatly. Taiwanese companies including Quanta, Wistron and Foxconn will continue to benefit from the rapid growth of the global AI server market.

Quanta vice president Yang Qiling recently said that demand for AI servers is expected to further heat up in the second half of 2024, and Quanta's AI server revenue is expected to account for more than 50% of its annual server revenue.

In addition, companies such as Wistron, Wiwynn, and Pegatron have also benefited significantly from AI infrastructure orders from Oracle and other cloud service providers. These companies are expected to continue to expand production in the coming year to meet growing global demand. With the overall improvement of the AI ​​server supply chain, the revenue growth of related companies in Taiwan is expected to accelerate further.

Disclaimer: The views in this article are from the original author and do not represent the views or position of Hawk Insight. The content of the article is for reference, communication and learning only, and does not constitute investment advice. If it involves copyright issues, please contact us for deletion.