Details
- Over 10,000 NVIDIA employees across engineering, product, legal, marketing, finance, sales, HR, operations, and developer programs gained early access to OpenAI's GPT-5.5 via the Codex agentic coding app, reporting mind-blowing productivity gains like reducing debugging from days to hours.
- NVIDIA and OpenAI collaborated, with GPT-5.5 running on NVIDIA GB200 NVL72 systems offering 35x lower cost per million tokens and 50x higher token output per megawatt versus prior generations.
- Codex supports secure enterprise use via remote SSH to dedicated cloud VMs with zero-data retention, read-only production access, and auditability, enabling agents to handle real data safely.
- This builds on a decade-long partnership starting in 2016 with DGX-1 delivery; recent milestones include joint GB200 NVL72 cluster bring-up and OpenAI's 10GW NVIDIA deployment commitment.
- Search confirms GPT-5.5 rollout to Codex and ChatGPT paid users, with previews noting fast, effective performance; no direct competitor launches in last 30 days match this enterprise agent scale.
Impact
NVIDIA's internal adoption accelerates AI agent workflows for knowledge work, validating GB200 economics for enterprise inference and pressuring competitors like Microsoft and Google to match agentic coding efficiency. Over 12-24 months, this deepens NVIDIA-OpenAI silicon codesign, steering R&D toward rack-scale systems and boosting GPU demand amid OpenAI's 10GW buildout.
