Odaily reported that Alex Svanevik, CEO of on-chain analytics platform Nansen, said the artificial intelligence industry bubble would break when enterprises begin to use Chinese large models effectively. In his view, the U.S. regulatory environment can limit that process, but the broader direction remains that Chinese models are becoming more efficient and can run on hardware that is not at the absolute cutting edge.
Svanevik also pointed to a rise in global GPU supply, including chips not made by Nvidia. He said recent declines in H100 and H200 GPU rental prices reflect a shift in the supply-and-demand structure for computing power. He raised the market signal of how to interpret falling GPU rental prices, connecting it with the simultaneous improvement in model efficiency and expansion of compute supply.
According to his comments, the AI infrastructure market is entering a repricing phase. As Chinese models become more efficient and more computing resources come into the market, the pricing logic built around high-end GPUs and model infrastructure costs is being reassessed.