AI datacenters, especially during training, bring extreme power fluctuations due to the nature of neural networks (gradient descent).
xAI has the largest AI data centers. xAI is using Tesla Megapacks to handle millisecond power fluctuations.
Supermicro provided the racks and systems for the xAI 100,000 GPU data center. They installed 64 H100 GPUs in each rack.
Elon Musk has said that they are adding 50,000 H100s and 50,000 Nvidia H200 GPUs to double the compute power of the data center.
The electrical grid can easily crash with electrical volatility. Think about how power surges can fry your computer or your appliances.
Non-AI datacenters, with steady year-round power usage, were grid-friendly.
By leveraging Megapacks, $xAI effectively reduces grid pressure: “xAI found millisecond power fluctuations when GPUs start training, causing issues with power infrastructure.
The solution was to use generators to charge the Tesla Megapacks, which then discharge to power the training jobs.
Training GPT-5 on 40,000 H100 GPUs over five months would consume ~150 GWh. Fully powering that with Megapacks (each providing 3.9 MWh at $2M per unit) would cost around $50B—obviously impractical.
Megapack are probably being used as a backup solution to buffer those high-power spikes.
Keep reading with a 7-day free trial
Subscribe to next BIG future to keep reading this post and get 7 days of free access to the full post archives.