Part 6/12:
Unprecedented Scale and Compute Power
Grok 5 is expected to leverage the next-generation Colossus 2 supercomputers, rumored to surpass the gigawatt power barrier. Previous models like Grok 3 trained on approximately 200,000 GPUs; Grok 5’s training is projected to utilize over 200,000 H100 GPUs, enabling it to handle vastly more data and complex reasoning.
This scale could unlock emergent intelligence, extending the model’s coherence over longer interactions and supporting real-time data processing—a true leap toward AGI.