China has achieved a significant breakthrough in artificial intelligence by successfully training a generative AI model across multiple data centers and GPU architectures. This feat was revealed by Patrick Moorhead, Chief Analyst at Moor Insights & Strategy, who discovered it during a conversation related to an NDA meeting. Training GAIs across different locations and architectures is crucial for China to maintain its AI advancements, especially as American sanctions have restricted the country from acquiring the latest, most powerful chips for research and development.
Nvidia, not wanting to lose the Chinese market, created the less powerful H20 AI chips that fall within Washington’s restrictive performance parameters. However, rumors suggest that even these down-tuned chips might be banned soon. Chinese researchers have been working on combining their limited supply of high-end, high-performance chips with readily available but less powerful GPUs like Huawei’s Ascend 910B or the Nvidia H20.
China’s unique approach to AI training
This technique, though historically inefficient, seems to have found new ground with China’s achievement of training a single GAI model across multiple data centers. While specific details about this GAI model are not yet available, the development highlights the lengths to which Chinese researchers will go to continue driving the country’s AI ambitions forward.
As Huawei stated, necessity is indeed the mother of invention. Some readers have debated the significance of this breakthrough, noting that the use of multiple data centers isn’t a new concept. Others pointed out that one of the primary challenges of consolidating AI training within a single data center is the enormous electricity consumption, which can destabilize entire power grids.
While there are still many questions about the intricacies and authenticity of China’s latest AI feat, the broader implications of this development cannot be ignored. As the global race for AI supremacy continues, such innovations could play a crucial role in shaping the future of technology.
