OpenAI and Dell have teamed up to offer Codex—the company's AI model that turns natural language into code—directly inside enterprise data centers. The partnership, announced this week, lets companies run the tool on Dell hardware and keep everything on-premises, a setup aimed at organizations that need tighter control over their data and workflows.
Why On-Premises Matters
Most enterprise AI tools today live in the cloud. That's fine for many use cases, but not for businesses that handle sensitive code, proprietary algorithms, or regulated data. Putting Codex on local servers means those companies don't have to send prompts or generated code to OpenAI's cloud—data stays inside the building. Dell's infrastructure handles the heavy lifting, from GPUs to network security, so IT teams can deploy the model without changing their existing compliance posture.
The move targets finance, healthcare, and government clients that have long wanted generative AI for coding but couldn't get past the data-sovereignty hurdle. By running Codex on Dell PowerEdge servers, those organizations can now automate code generation, bug fixes, and documentation within their own walls.
Expanding Enterprise Workflows
Codex isn't new—developers have used it in GitHub Copilot, which OpenAI built with Microsoft. But that version runs in the cloud. The Dell partnership shifts the model into a dedicated enterprise environment where it can be integrated with internal tools like DevOps pipelines, version control systems, and security scanners. OpenAI says the deployment is designed for "broader enterprise workflows," meaning it's not just about writing code faster—it's about weaving AI into how teams build, test, and deploy software.
Dell will handle support, updates, and hardware maintenance. The companies haven't disclosed pricing, but on-premises AI typically involves a licensing fee plus hardware costs. For large firms already running Dell gear, the setup could be cheaper than paying per-token for cloud API calls over years of daily use.
The partnership also opens the door to custom fine-tuning. Enterprises can train Codex on their own codebases and internal standards without uploading that data anywhere. That's a big selling point for shops with millions of lines of legacy code or industry-specific syntax.
Neither OpenAI nor Dell has released a timeline for when the on-premises Codex will ship to customers. Beta access is expected later this year, but general availability remains unconfirmed.
