NVIDIA deploys Codex to 10,000

Nvidia has rolled out OpenAI’s Codex to more than 10,000 employees across the company, turning an AI coding tool into a company-wide work app. (blogs.nvidia.com) The deployment reaches engineering, product, legal, marketing, finance, sales, human resources, operations and developer programs, according to Nvidia. The company said employees are already using it for software work and other internal tasks. (blogs.nvidia.com) Codex is OpenAI’s coding agent, a system built to write, edit and reason through programming tasks with less step-by-step prompting from a human. OpenAI said GPT-5.5 became available in Codex on April 23, 2026, for complex coding, computer use, knowledge work and research workflows. (developers.openai.com) (openai.com) Nvidia said this rollout runs on its GB200 NVL72 rack-scale systems, a Blackwell-based machine that packs 72 graphics processors into a single rack. Nvidia describes the system as an exascale computer in one rack with 130 terabytes per second of GPU-to-GPU communication bandwidth. (blogs.nvidia.com) (nvidia.com) The company said engineers had used GPT-5.5 through the Codex app for several weeks before the broader launch. In Nvidia’s account, internal users said results were “mind-blowing” and “life-changing,” and one engineer said debugging jobs that took days were finishing in hours. (blogs.nvidia.com) The move comes as OpenAI is trying to turn Codex from a developer product into an enterprise platform. On April 21, OpenAI said it launched Codex Labs with consulting partners including Accenture, Capgemini, Cognizant, Infosys, PwC and Tata Consultancy Services to help companies move from pilots to production deployments. (openai.com) OpenAI also said Codex had reached 4 million weekly active users, giving Nvidia’s internal rollout a larger market backdrop than a one-off pilot. That places the Nvidia deployment inside a broader push to make AI agents part of everyday software and knowledge work. (openai.com) Nvidia framed the launch as part of a longer partnership with OpenAI that dates to 2016, when Jensen Huang delivered a DGX-1 system to OpenAI’s San Francisco office. The companies have since worked together across model training, inference systems and now internal workplace tools. (blogs.nvidia.com) For Nvidia, the message is concrete: the company that sells AI infrastructure is now using one of the newest AI agents across its own workforce. For OpenAI, it is a live reference customer with 10,000 users running on Blackwell hardware. (blogs.nvidia.com) (openai.com)