A major customer of NVIDIA's JAX GPUs has reportedly switched to a custom C training framework developed by xAI. This significant shift is attributed to the JAX stack's low utilization, with reports indicating an MFU below 10%. The move suggests a potential re-evaluation of GPU hardware choices for large-scale AI training. AI
IMPACT This shift highlights potential dissatisfaction with current GPU offerings and signals a growing interest in custom, efficient training frameworks.
RANK_REASON A customer's adoption of a new training framework and abandonment of specific hardware is a product-level event.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
