Vertiv Opens ATLAS Lab to Test AI Infrastructure Before Chips Arrive

Unite.AI
Read full story· 4 min
Vertiv launched its ATLAS Lab to test AI data center infrastructure as a complete system using emulated NVIDIA workloads before actual chips are available. The lab aims to identify integration issues early by simulating power, cooling, controls, and IT systems together at pod scale.

What happened

Vertiv opened its ATLAS Lab on November 16, 2026, to test AI data center infrastructure as a complete system before production chips are available. The lab uses emulated NVIDIA workloads to assess how power, cooling, controls, and IT systems work together at the scale of an AI factory pod. Vertiv and NVIDIA engineers began technical sessions at the lab on November 6, 2026, focusing on moving from digital simulation to physical integration testing. The facility is designed to find integration issues early, before customer deployments begin.

Key details

  • The lab uses emulated NVIDIA accelerated-computing workloads to mimic power and heat behavior.
  • Testing covers the full system, from grid connection to chip-level liquid cooling and heat rejection.
  • NVIDIA DSX AI Factory Platform and Omniverse libraries are used for simulation and digital-twin workflows.
  • The lab serves as a shared environment for chipmakers, rack-system vendors, power and cooling suppliers, utilities, and operators.

Why it matters

Data center operators, equipment manufacturers, and utilities can now identify and resolve integration problems before deploying new AI infrastructure. This reduces the risk of costly redesigns or delays at customer sites and allows for faster, more reliable scaling of AI compute capacity. The validated pod designs from the lab can be used as reference architectures, streamlining future deployments and reducing the need for site-specific engineering.


More in Chips & Compute