DeepSeek Develops In-House AI Inference Chip; Multiple AI Industry Updates

APPSO · wechat · 2026-07-08

According to Reuters citing sources, DeepSeek is developing its own AI chips primarily for inference, aiming to reduce reliance on NVIDIA chips. The project started about a year ago, is still in early stages, and has engaged chip design, foundry, and memory manufacturers, while privately ramping up chip design engineer recruitment in recent months. If successful, DeepSeek will further venture into semiconductor design.

Xiaomi is restructuring Xiao Ai: foundational model capabilities will be provided by Luo Fuli's MiMo base model team, Luan Jian is responsible for cloud capabilities, on-device capabilities are delegated to OS teams for phones, cars, etc., and former head Wang Gang is transferred to robotics. Xiaomi will invest at least 60 billion yuan in AI over the next three years. Additionally, former Huawei "Genius Boy" Li Bojie publicly criticized DeepSeek's recruitment process, saying no interview was scheduled after a written test for half a month, and he was accused of code plagiarism in a second interview, leading to termination.

Other updates: OpenAI's Chief Futurist is leaving; xAI officially renames to SpaceXAI; the world's first AI agent phone to debut; ClaudeCowork launches on mobile; Ant Lingbo releases LingBot-Depth2.0 with training data expanded to 150 million; HKU open-sources AI agent tutor DeepTutor v1.5; former Google exec claims 22-year-old AI-native employees already surpass senior staff in efficiency.

Original post →

DeepSeek Develops In-House AI Inference Chip; Multiple AI Industry Updates

More from Companies & People