Moore Threads

Moore Threads (摩尔线程) was founded in Beijing in October 2020 by Zhang Jianzhong, who spent nearly fifteen years at NVIDIA, rising to general manager of NVIDIA's China business from 2006 until his departure in 2020. Zhang's move from running NVIDIA's China operation to building a domestic competitor is one of the more pointed "poacher becomes competitor" stories in Chinese semiconductors, and Moore Threads has been explicitly marketed — by press and investors alike — as "China's Nvidia." The company drew early backing from Tencent, ByteDance, and a roster of state-linked investment funds, reflecting Beijing's interest in cultivating a full-function domestic GPU champion rather than a narrower AI-accelerator specialist. Unlike Cambricon or Biren, which built dedicated AI training and inference accelerators, Moore Threads designs the MTT S-series as full-function GPUs intended to serve graphics rendering, gaming, and general-purpose AI compute simultaneously — mirroring NVIDIA's own product philosophy rather than carving out a narrower niche. Its consumer-facing MTT S-series graphics cards target desktop gaming and workstation graphics, while its data center-oriented chips are pitched at AI training and inference clusters for Chinese cloud customers. The company has continued to iterate its architecture toward higher AI throughput as its data center ambitions have grown, but its chips remain a generation or more behind NVIDIA's leading data center GPUs in raw performance-per-die, a gap attributable in large part to its fabrication constraints. Moore Threads' access to leading-edge fabrication has been directly shaped by US export policy. On October 17, 2023, the Bureau of Industry and Security added Moore Threads entities spanning its Beijing headquarters and subsidiaries in Chengdu and Shanghai to the Entity List, restricting the supply of US-origin technology, software, and components without an individual export license that is presumptively denied. The listing effectively cut Moore Threads off from TSMC, whose leading-edge fabrication depends on equipment and IP under US technology jurisdiction, and pushed the company onto SMIC's mature-node processes for its GPU dies. SMIC's process limitations constrain the density and power efficiency Moore Threads can achieve relative to NVIDIA's TSMC-fabbed products, and its EDA design flow depends on Cadence and Synopsys tools that carry the same export-license exposure affecting other Entity-Listed Chinese fabless designers. Moore Threads went public on Shanghai's STAR Market in one of the defining capital-markets events of China's 2025-2026 domestic chip boom. The offering priced at 114.28 yuan per share for 70 million shares (14.89% of total equity), raising roughly 8 billion yuan (about $1.1 billion) and implying an IPO-price valuation of approximately 53.7 billion yuan. Retail demand was extraordinary: online investor subscription orders exceeded the shares on offer by more than 4,000 times, and the final retail allotment rate was just 0.036%. When trading opened on December 5, 2025, shares surged as much as 502% intraday to roughly 688 yuan, pushing Moore Threads' market capitalization past 300 billion yuan and, within days, past 400 billion yuan as the rally continued — one of the largest and most oversubscribed STAR Market debuts on record. That enthusiasm ran well ahead of Moore Threads' financial fundamentals. 2025 revenue rose approximately 243% year-on-year to about 1.5 billion yuan, continuing four consecutive years of triple-digit growth, but the company remained deeply unprofitable, with cumulative net losses exceeding 5 billion yuan since 2022, and management has guided that Moore Threads does not expect its first profitable year before 2027. On the software side, Moore Threads has built its own MUSA (Moore Threads Unified System Architecture) compute platform and a CUDA-compatibility layer to lower porting costs for developers, but the breadth and maturity of NVIDIA's CUDA ecosystem — its libraries, frameworks, and years of accumulated developer tooling — remains a structural disadvantage that no amount of raw silicon performance fully offsets. Moore Threads' path forward depends on converting its IPO capital and STAR Market visibility into a durable customer base among Chinese cloud providers, gaming hardware partners, and enterprise buyers substituting away from NVIDIA under export-control pressure and government procurement guidance. Its dual graphics-plus-AI strategy gives it a broader addressable market than pure AI-accelerator peers, but also means it competes simultaneously against NVIDIA, AMD, and domestic AI-specialist rivals such as Cambricon, Biren, MetaX, and Enflame across more fronts at once. With profitability still years away and its process technology capped by SMIC's mature-node limits, Moore Threads' near-term story is less about matching NVIDIA on raw performance and more about proving it can scale sustainable revenue from a captive but rapidly expanding domestic market.