OpenAI and Dell: Codex Now Runs on Companies' Own Servers

OpenAI and Dell have announced a strategic partnership to deploy Codex in the hybrid and on-premises clouds of corporate customers. It is OpenAI's first serious step into private, local AI, a market that has long been waiting for such a solution.

How it works

Codex is a large language model trained specifically on thousands of examples of source code. The model can understand a task, write code and fix bugs. Dell provides the infrastructure, integration technology and support. OpenAI provides the model itself, fine-tuning for the customer's tasks, and technical support.

The result is a hybrid cloud in which data stays inside the corporate network. The company's source code does not reach OpenAI's servers; all processing happens locally, on Dell hardware.

Who needs this most

Banks, government agencies, pharmaceutical companies, manufacturers of strategic goods: in short, everyone who works with secret or highly confidential source code. GDPR, HIPAA, SOC 2 and other privacy requirements prohibit sending data to third parties.

Until now, that meant a choice: either a cloud AI from OpenAI or Google (which the law ruled out), or local models (which were weak and required the company's own investment in training). Now there is a third option: a model of cloud-grade quality with a fully private deployment.

• Financial institutions (banks, investment funds, insurance companies)
• Government, defense ministries and intelligence services
• Pharmaceuticals and biomedicine
• Large-scale manufacturing (defense, aerospace, energy)
• Critical infrastructure (telecom, power grids)

A competitive move

Microsoft through Azure and Google through Vertex have offered local deployment of generative models for several years. OpenAI, however, stayed cloud-only for a long time: all its services run through an API on OpenAI's servers. The enterprise market, which accounts for effectively half of global AI demand, requires private deployment. OpenAI fell behind.

This partnership is a direct response to growing demand from large corporations that would have chosen OpenAI for model quality but could not use a cloud API because of regulation. Dell is a strategic partner here, with deep existing relationships with the IT departments of large companies. For OpenAI, it is a faster way into the enterprise market. For Dell, it is a new service it can sell to top corporate clients.

What it means

On-premises deployment of AI models is becoming the enterprise standard, not the exception. For OpenAI, it is a path to billion-dollar contracts with large companies. For everyone else, it is a clear signal: if you are not ready to offer private deployment of your model, you are losing half of the enterprise market.