The launch of DeepSeek's upcoming model, R2, is facing delays due to a shortage of NVIDIA's H20 chips caused by US export restrictions.
R2, the successor to the widely used R1, is still awaiting a release date as CEO Liang Wenfeng is dissatisfied with its performance and engineers are working on improvements.
Cloud providers hosting DeepSeek's models anticipate a shortfall in NVIDIA chips to meet the potential demand of R2, especially if it outperforms current open-source alternatives.
Despite chip shortages, Chinese firms are finding workarounds to access NVIDIA hardware, including sending engineers to Malaysia for training AI models due to export restrictions.