OpenAI’s GPT-5.5 Powers Codex on NVIDIA Infrastructure, Sparking Productivity Surge

To ensure that AI agents can handle sensitive company data safely, NVIDIA IT implemented a high-security deployment architecture.

RMN Digital AI Desk
New Delhi | April 25, 2026

SANTA CLARA, Calif. — April 23, 2026 — OpenAI has launched its latest frontier model, GPT-5.5, which is now powering the Codex agentic coding application running on NVIDIA GB200 NVL72 rack-scale systems. This new integration is already being utilized by over 10,000 NVIDIA employees across various departments, including engineering, legal, marketing, and finance, with many describing the initial results as “mind-blowing”.

Massive Efficiency Gains

The deployment on NVIDIA’s GB200 infrastructure has reached a new level of economic viability for enterprise-scale AI. The GB200 NVL72 system delivers a 35x lower cost per million tokens and 50x higher token output per second per megawatt compared to previous-generation systems.

These hardware advancements have translated into immediate practical benefits for NVIDIA’s developers:

• Debugging cycles that once took days are now being resolved in hours.
• Complex experimentation across multi-file codebases has been shortened from weeks to overnight progress.
• Teams are successfully shipping end-to-end features generated from natural-language prompts with improved reliability.

In a company-wide email, NVIDIA founder and CEO Jensen Huang urged his staff to embrace the technology, stating, “Let’s jump to lightspeed. Welcome to the age of AI”.

Secure Enterprise Deployment

To ensure that AI agents can handle sensitive company data safely, NVIDIA IT implemented a high-security deployment architecture. Every employee runs their Codex agent in a dedicated cloud virtual machine (VM) via remote Secure Shell (SSH) connections. This provides a secure sandbox for the agent while maintaining a zero-data retention policy and full auditability. These agents interact with production systems using read-only permissions through a specialized toolkit known as “Skills”.

A Decade of Collaboration

The launch of GPT-5.5 is the latest milestone in a ten-year partnership between NVIDIA and OpenAI that began in 2016 when Huang delivered the first DGX-1 AI supercomputer to OpenAI. Looking forward, OpenAI has committed to deploying more than 10 gigawatts of NVIDIA systems for its next-generation infrastructure.

The two companies continue to act as co-design partners; OpenAI provides feedback for NVIDIA’s hardware roadmap, while NVIDIA provides early access to new architectures. This collaboration recently led to the successful bring-up of the first 100,000-GPU cluster using the GB200 NVL72 architecture, setting new benchmarks for system-level reliability.