Nvidia rolls out OpenAI Codex internally

Nvidia has rolled out OpenAI’s Codex to all employees after testing it with more than 10,000 staff across engineering and business teams. (nvidia.com) (economictimes.indiatimes.com) Chief executive Jensen Huang told staff that “OpenAI’s Codex, powered by GPT-5.5, has launched and is available to every NVIDIAN,” according to an internal email reported on April 25. Sam Altman said on X that OpenAI and Nvidia had tested rolling out Codex “across a whole company.” (economictimes.indiatimes.com) (storyboard18.com) The tool is not limited to software engineers. Nvidia said early access covered engineering, product, legal, marketing, finance, sales, human resources, operations and developer programs. (nvidia.com) Codex is OpenAI’s coding agent, a system that can write code, review changes, plan tasks and run workflow steps from natural-language prompts. OpenAI said GPT-5.5 became the model behind Codex on April 23. (openai.com) (nvidia.com) Nvidia said engineers using Codex saw debugging cycles shrink from days to hours, while experiments in large codebases moved from weeks to overnight runs. Teams were also shipping end-to-end features from prompts, according to the company. (nvidia.com) To run the system inside a large company, Nvidia IT gave each employee a cloud virtual machine for the agent and used remote Secure Shell connections to approved systems. Nvidia said the setup uses zero-data retention and read-only access to production systems. (nvidia.com) The rollout lands as OpenAI is trying to move Codex from an individual developer tool into a company software platform. On April 21, OpenAI said Codex had grown from more than 3 million to more than 4 million weekly developers in two weeks. (openai.com) OpenAI also launched Codex Labs, a program that sends its specialists into companies, and signed seven global systems integrators including Accenture, Capgemini, Cognizant, Infosys and Tata Consultancy Services. Greg Brockman said OpenAI is exploring similar deployments with other enterprises. (openai.com) (economictimes.indiatimes.com) Nvidia tied the deployment to its own hardware pitch. The company said GPT-5.5 in Codex runs on Nvidia GB200 NVL72 systems, which it says cut cost per million tokens by 35 times and raise token output per second per megawatt by 50 times versus prior-generation systems. (nvidia.com) For Nvidia, the internal rollout doubles as a customer case study: its staff use OpenAI’s agent, but the computing, security controls and infrastructure are Nvidia’s. Huang closed his note with a line that framed the broader bet: “Welcome to the age of AI.” (nvidia.com)