Oracle is integrating Google’s TPU artificial intelligence accelerators into its public cloud, marking the first time Google provides these chips to another hyperscaler. The deployment is scheduled to begin next year, with plans to gradually expand capacity.
The infrastructure will utilize Google’s OCS and ICI technologies for data exchange, while connecting to Oracle data centers via the proprietary Acceleron networking technology, which features custom SmartNICs for network management and security.
Oracle is expanding its hardware portfolio by adding Cerebras Systems’ wafer-scale accelerators and OpenAI’s Jalapeño inference processors. OpenAI claims the Jalapeño chip runs certain open-source models 1.5 times faster than competing hardware.
The partnership with OpenAI includes new software integrations. Oracle is launching AI Workspace, which connects its business applications to ChatGPT, and App Composer, a tool allowing non-technical users to build custom software with AI assistance.
Oracle is also introducing AI Security Mesh, a tool that uses OpenAI models to identify and prioritize application vulnerabilities. Additionally, the company announced the Oracle AI SMB Bundle for small businesses, which includes $12,000 in annual cloud credits.
