Lenovo's Yang Yuanqing: 70% of Future GPU Servers Will Be Used for Inference, 30% for Training

Yang Yuanqing, Chairman and CEO of Lenovo Group, stated during the earnings briefing that AI infrastructure is experiencing explosive growth, with a very rapid shift from training to inference.Currently, 70% to 80% of AI infrastructure GPU servers are used for training, with only 20% to 30% dedicated to inference. He believes this ratio will reverse in the future, with 70% used for inference and 30% for training. This does not mean that the infrastructure required for future AI training will decrease; on the contrary, it will continue to grow. Relying solely on large cloud service providers will no longer be sufficient to meet the demand.