Databricks co-founder argues US must go open source to beat China in AI (techcrunch.com)

At the Cerebral Valley AI Summit, Databricks co-founder Andy Konwinski warned that the U.S. is losing its edge in AI to China and urged a return to open-source practices to reclaim leadership. Konwinski — who now runs the Laude venture and research institute — said PhD students at top U.S. universities are reading “twice as many” interesting ideas from Chinese companies as American ones. He blamed the trend on large U.S. labs’ proprietary models and high-salary recruiting that drains academic talent, while Chinese labs and the government often favor open-sourcing models (examples include DeepSeek and Alibaba’s Qwen), enabling faster community-driven progress. The technical crux of Konwinski’s argument is that major breakthroughs build on broadly shared methods — he cites the Transformer architecture, published openly, as the foundation for generative AI. If the next “Transformer-level” innovation is developed in an open ecosystem, that nation gains an outsized advantage. The implication is twofold: continued proprietary silos could erode U.S. scientific diffusion, threaten democratic resilience, and eventually undermine even the big U.S. labs’ business positions within a few years. Konwinski’s call is for policies and incentives that restore free exchange between academia and industry to accelerate innovation and maintain strategic leadership in AI.