NVIDIA rolls out OpenAI Codex

Nvidia has made OpenAI’s Codex available company-wide after first giving the tool to more than 10,000 employees across departments. (economictimes.indiatimes.com) (blogs.nvidia.com) Chief executive Jensen Huang wrote in an internal email that Codex, powered by GPT-5.5, had “launched and is available to every NVIDIAN.” OpenAI chief executive Sam Altman then posted that email on X and said the company-wide rollout “was awesome to see it work.” (economictimes.indiatimes.com) (storyboard18.com) Nvidia said the earlier deployment covered engineering, product, legal, marketing, finance, sales, human resources, operations and developer programs. The company described Codex as an “agentic” system, meaning software that can carry out multi-step tasks instead of only answering prompts. (blogs.nvidia.com) (finance.yahoo.com) The company tied the rollout to measurable internal gains. Nvidia said debugging jobs that once took days were finished in hours, multi-file experiments that took weeks ran overnight, and some teams shipped end-to-end features from natural-language instructions. (blogs.nvidia.com) (letsdatascience.com) The underlying pitch is broader than coding. Nvidia’s own list of users includes legal, finance, marketing and human resources, which turns Codex into a general workplace tool inside a company best known for chips and engineering. (blogs.nvidia.com) (storyboard18.com)) Nvidia also used the announcement to promote the hardware behind the software. The company said GPT-5.5 is running on Nvidia GB200 NVL72 systems and claimed those racks deliver 35 times lower cost per million tokens and 50 times higher token output per second per megawatt than the prior generation. (blogs.nvidia.com) (techspot.com) That makes the rollout part product demo, part customer case study. Nvidia is showing that the same company selling artificial-intelligence infrastructure is also using OpenAI’s software internally at scale. (blogs.nvidia.com) (finance.yahoo.com) The backdrop is a decade-long relationship between the two companies. Nvidia said its work with OpenAI dates to 2016, when Huang delivered the first DGX-1 system to OpenAI’s San Francisco office, and the new Codex rollout extends that partnership from training models to daily office work. (blogs.nvidia.com) With Altman publicly inviting other companies to try the same model, Nvidia’s internal deployment now reads as a reference account for selling AI agents beyond software teams. (storyboard18.com) (kingy.ai)