xAI reportedly has around 550,000 $NVDA H100 and H200 GPUs, but is only using about 11% of that fleet, equal to roughly 60,000 GPUs effectively utilized


According to The Information, the key issue is not hardware availability, but software stack efficiency. At massive scale, idle time increases quickly because distributed training, data pipelines, scheduling, and analysis systems become harder to coordinate
$META and $GOOG are reportedly achieving much better utilization, around 43% and 46%, because their internal software stacks are more mature
xAI’s goal is to reach 50% utilization, but no timeline is given. The main path forward would be better infrastructure orchestration, training software, data pipeline optimization, and workload management
post-image
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pin