Oracle plans to expand its multicloud AI database services for the US government, announced on March 31, 2026.
The push centers on deploying NVIDIA B300 GPUs across its US government cloud regions. Those chips are built on NVIDIA’s Blackwell Ultra architecture and aimed at the heavy lifting behind large language models.
What Oracle is actually rolling out
The B300 GPUs are headed to Oracle Cloud Infrastructure (OCI) government regions in US Gov East and US Gov West. They are also slated for the defense-focused regions: US DoD East, US DoD North and US DoD West.
The goal is to support both inference and training workloads for large language models.
Oracle will introduce model options from xAI, specifically Grok, along with NVIDIA’s Nemotron. Both are planned for agency use. Oracle is pitching these models for situational awareness and cybersecurity work.
Oracle also launched a Defense Industrial Base Isolated Cloud Environment designed to give contractors secure, air-gapped spaces to collaborate.
How Oracle got here
Earlier authorizations in February and March 2026 brought OCI Generative AI and other services into Oracle’s government cloud lineup. Oracle already had five US government regions up and running before this expansion. Prior FedRAMP and DISA approvals covered OCI Generative AI and Exadata Cloud@Customer.
The move ties into Oracle’s broader multicloud and Autonomous AI Database strategies. The pitch is that agencies can run production-grade generative AI applications on infrastructure that meets government compliance requirements, including environments handling sensitive and controlled unclassified information.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.