Oracle and NVIDIA Expand Vera Rubin Infrastructure to Multi-Gigawatt Scale on OCI
Key Facts
Oracle and NVIDIA expanded their plans for shared cloud computing infrastructure to a multi-gigawatt scale, announcing the general availability of Vera Rubin NVL72 systems and Vera CPUs on Oracle Cloud Infrastructure. The expansion is intended to meet customer demand for AI training, inference and agentic applications, according to an announcement by NVIDIA and Oracle. The companies said customers can scale CPU and GPU resources independently to suit their workloads. Adding Vera CPUs to OCI gives customers the option to expand CPU capacity alongside GPU resources. The announcement came as Oracle said demand for AI training and inference services is growing faster than supply.
OCI Superclusters combine rack-scale NVL72 systems with Oracle Acceleron and Spectrum-X Ethernet networking, as well as Spectrum-6 switches and BlueField-4 data processing units. According to the announcement, each rack connects 72 GPUs through an NVLink network, while OCI’s multilevel network is designed to scale across hundreds of thousands of GPUs. The plan is built on gigawatt-scale power and cooling infrastructure, enabling large computing clusters to operate in data centers. The companies said the configuration targets training, inference and agentic workloads that require coordination across computing, connectivity and data transfer resources. The ability to scale CPU and GPU resources independently allows capacity to be matched to each workload’s requirements.
Oracle also plans to add HGX Rubin NVL8 systems, x86 servers with 8 Rubin GPUs connected through NVLink. This configuration is part of the announced expansion plans, while Vera Rubin NVL72 systems and Vera CPUs are now generally available on OCI. The announcement did not specify a separate date for the introduction of HGX Rubin NVL8 systems. It marks a shift in the partnership from a broad framework for infrastructure expansion to offering specific systems to customers on the cloud. The company is thus providing more than one way to configure computing resources, with CPU and GPU capacity available to scale independently.
The move follows strong growth in Oracle’s cloud business, with the company reporting revenue of $19.3 billion in the first quarter of fiscal 2027, up 30% year over year. Cloud revenue rose 62% to $11.6 billion, while cloud infrastructure revenue increased 121% to $7.4 billion. In the same quarter, Oracle signed additional contracts for cloud AI services worth more than $30 billion and said demand for training and inference exceeded supply. Remaining performance obligations also rose $209 billion year over year to $664 billion. They had reached $638 billion in the fourth quarter of fiscal 2026, after a quarterly increase of $85 billion.
The expansion provides context for Oracle’s spending to build OCI capacity, after the company said it raised $43 billion in debt financing and $5 billion in equity financing during fiscal 2026. Oracle expects to raise about $40 billion through a mix of debt and equity in fiscal 2027. Markets are watching Oracle’s ability to convert its accumulated obligations into revenue amid data center delays and the possibility of additional financing, according to a Reuters report citing JPMorgan analysts. In extended trading following September results, Oracle shares rose nearly 8%, despite being down more than 20% since the start of the year at the time of the report. The announced expansion reflects the company’s bet on sustained demand, while execution and the conversion of contracts into revenue remain key areas of investor focus.
Oracle expects total revenue growth of between 30% and 34% in the second quarter of fiscal 2027, and dollar cloud revenue growth of between 65% and 71%. The company also raised its fiscal 2027 revenue forecast to at least $90 billion, with non-GAAP earnings per share of $8.10. In the AI infrastructure market, Google said general availability of A5X instances would begin in the month following its announcement, and that Virgo Network would support expansion to 80,000 Rubin GPUs at a single site and 960,000 across multiple sites. Those figures relate to Google’s announced plan, not Oracle’s expansion. Cloud providers continue to compete to make Vera Rubin resources available and expand the capacity needed for AI workloads.