The real advantage of Chinese AI companies in this round is that many teams haven’t yet caught the “research above engineering” disease. In many Silicon Valley companies, researchers are seen as royalty, while infrastructure work is considered grunt work—everyone wants to move toward research. But by now, with large models at this stage, research no longer happens solely in algorithms and papers. What sets teams like DeepSeek and Kimi apart is that algorithms, data, and infrastructure haven’t started blaming each other. Everyone works toward the same goal, and whoever encounters a problem steps in to solve it. At this frontier of model development, infrastructure is research.