Chinese AI models are gaining ground with U.S. companies as costs surge (www.cnbc.com)

U.S. companies are increasingly adopting Chinese-built AI models, driven by surging costs associated with using leading U.S. systems. Recent advancements from Chinese firms like DeepSeek and Z.ai have narrowed the performance gap with their American counterparts, while offering significant cost savings—some models are reported to be 60% to 90% cheaper than U.S. offerings from organizations like OpenAI and Anthropic. The share of tokens used by U.S. companies on Chinese AI models has dramatically increased; from just 11% before February, it has consistently surpassed 30% this year, indicating a shifting preference towards more economical options. This trend has significant implications for the AI/ML community, especially as the U.S. administration explores regulations on domestic AI models. As firms seek efficiencies and alternatives, engineers are experimenting more with open-source and open-weight models, which Chinese companies are now dominating. The capability of these models is impressive, with recent releases like GLM 5.2 showing performance close to top-tier U.S. models on certain benchmarks. This shift points to a growing reliance on affordable and adaptable AI solutions, even as concerns about the implications of using foreign models increase.