AI news
1. xAI Activates 'Colossus 2': The World's Largest AI Supercluster
Date: August 22, 2026
xAI has officially operationalized Colossus 2, a massive AI supercomputing cluster located in Memphis, Tennessee. The facility has scaled to 200,000 interconnected GPUs, integrating NVIDIA H200 and Blackwell B200 units. This deployment represents the largest single--site AI training infrastructure currently in operation, designed to accelerate the development of the Grok--4 frontier model.
| Feature | Specification |
|---|---|
| Compute Units | 200,000 GPUs (H200/B200) |
| Power Capacity | 150MW Dedicated Substation |
| Cooling | Advanced Liquid--to--Liquid Cooling |
| Primary Objective | Training Grok--4 Frontier Model |
Key Facts & Business Trends:
- Compute Sovereignty: By building proprietary infrastructure, xAI reduces reliance on third--party cloud providers, allowing for faster iteration cycles and lower long--term marginal costs.
- Infrastructure Lead: The activation positions xAI at the forefront of raw on--site compute density, a critical factor for training next--generation "World Models."
- Energy Infrastructure Integration: The project highlights a growing trend where AI firms must directly partner with utility providers to secure the massive power loads required for frontier--scale clusters.
Sources: Bloomberg, TechCrunch