Overview

• DeepSeek has postponed a release date for its R2 AI model as CEO Liang Wenfeng remains unsatisfied with its performance.
• The company had targeted May for R2’s debut but its engineering team continues intensive refinements in coding and reasoning benchmarks.
• U.S. export controls enacted in April barred Nvidia from shipping H20 AI processors to China, cutting off a key hardware source.
• Chinese cloud providers report a critical shortage of H20 chips, hampering both ongoing R1 operations and future R2 deployments.
• In preparation for R2’s launch, DeepSeek is providing technical specifications to major Chinese cloud firms to guide infrastructure upgrades.