Huawei Pangu Pro Trains 505 Billion Parameters Without Nvidia

Huawei's Pangu Pro model, trained with 20,000 Nvidia chips, demonstrates the potential for large-scale AI training without relying on traditional chip suppliers.

Trains 505 billion parameters

Challenges Nvidia's dominance in AI chip supply

Raises questions about supply chain resilience

Source: https://news.google.com/rss/articles/CBMi5gFBVV95cUxNUzRucnk5VmkxVVdTaGNpUk5UMW1haklZTnoxdllNWF95cTR3aTFTanI1ZzBSZmVyTUJySnBtUS1vQmp3dDFFQVFsVmdJVUFJa01RMi1fTVI4UjdZWEVDcmNjdGI0aGw3R1dVMl83bDRMdjBUWk1ERms0QnhBYUctRHpWd1plWWRYMVpvZnQ4cHd4ZFRoYkJDZEwzUzlxbzI4R0ZqOXpoX2huZUZwZzZmQzlpZHU2dGFia2ZFQ2hoX2wtOGFpanJ4VkVxRVh3bzJHLUFWVTRJVnhaLTVPc2FKVmJBcnZLQQ?ucbcb=1&hl=en-US&gl=US&ceid=US:en