WORLDTECHOracle plans to expand AI database services for US government clouds
Oracle has announced plans to widen its multicloud AI database services for the US government, bringing NVIDIA B300 GPUs into its government cloud regions. The company presents the step as a way for agencies to run generative AI applications that meet government compliance requirements.
Chips destined for government regions
The B300 processors, which use NVIDIA's Blackwell Ultra architecture, are to be installed in Oracle Cloud Infrastructure government regions US Gov East and US Gov West. They are also intended for the defense-oriented regions US DoD East, US DoD North and US DoD West.
According to the report, the chips are meant to handle both inference and training work for large language models, the computation these systems depend on.
Isolated environment for defense suppliers
Alongside the chip rollout, Oracle has launched a Defense Industrial Base Isolated Cloud Environment. It is designed to give contractors secure, air-gapped spaces in which to work together, meaning spaces physically cut off from other networks.
Oracle is marketing the AI models that run on this infrastructure for situational awareness and cybersecurity tasks.
Built on earlier approvals
Oracle's government cloud offering has grown over recent months. Authorizations in February and March 2026 added OCI Generative AI and other services to the lineup, and Oracle already operated five US government regions before this expansion. Earlier approvals from FedRAMP and DISA covered OCI Generative AI and Exadata Cloud@Customer.
The expansion ties into Oracle's wider multicloud and Autonomous AI Database strategies. The company argues that agencies can run production-grade generative AI applications in environments that handle sensitive and controlled unclassified information.
Open questions
- When will the B300 GPUs become available in the government regions?
- Will the planned Grok and Nemotron models be deployed in these regions?