How China’s New GPU Startup Moore Thread Is Accelerating the AI Race

GPU Shortage and Geopolitical Context

The global chip shortage list highlighted "high‑end GPU" in red, and in October 2022 the United States tightened export controls on Chinese chips, banning Nvidia's A800 and H800 GPUs designed for the Chinese market. This underscored GPUs as the "AI engine" and a focal point of great‑power tech competition.

Why GPUs Are the Heart of Modern Computing

GPUs (graphics processing units) are massively parallel micro‑processors that excel at accelerating compute‑intensive tasks. In AI training and inference, their parallel architecture makes them the dominant choice. An analogy compares a CPU to a learned professor handling complex, varied tasks one at a time, while a GPU resembles a legion of simple soldiers executing many repetitive instructions simultaneously.

This parallel advantage has expanded GPUs from graphics rendering to virtually every frontier technology, including gaming, video editing, AI‑generated content, weather forecasting, drug discovery, autonomous driving, and more. According to Frost & Sullivan, China’s AI compute‑accelerator market is projected to grow from ¥142.5 billion in 2024 to ¥1,336.8 billion in 2029, a CAGR of 53.7%.

Moore Thread: From Startup to IPO Candidate

Founded in June 2020 by former Nvidia global vice‑president Zhang Jianzhong, Moore Thread assembled a team of senior engineers from Nvidia, AMD, and other chip giants, giving it a high "core‑chip" density. Unlike many domestic AI‑focused chip firms, Moore Thread targets a "full‑function GPU" that can handle AI, scientific computing, and traditional graphics rendering.

In just five years, the company’s valuation surpassed ¥255 billion, establishing it as a leader in China’s GPU sector.

Ecosystem Strategy: MUSA Platform and Full‑Stack Integration

GPU performance depends not only on transistor count but also on software ecosystems. Nvidia’s CUDA platform, built over a decade, hosts millions of developers. Moore Thread’s response is the MUSA platform, designed to be CUDA‑compatible, allowing developers to migrate existing applications with minimal cost.

The company is building a full‑stack ecosystem—chip, board, cluster, and software—to bridge the "last mile" for applications. Partnerships include domestic packaging tester Runxin Technology and OS adaptations for UnionTech and Kylin. Industrial software integration with ZWSOFT further reduces migration barriers.

IPO Significance and Funding Roadmap

Moore Thread’s upcoming STAR Market IPO aims to raise ¥80 billion. The proceeds will fund next‑generation AI training/inference chips, graphics chips, AI SoC chips, and working capital. This allocation directly addresses three core challenges for domestic GPUs: AI training compute, graphics for industrial and digital‑twin scenarios, and integrated AI SoC for edge and intelligent terminals.

The IPO has attracted attention across the semiconductor supply chain, from fabs like SMIC and Changjiang to downstream players such as Inspur and Desay SV.

Future Outlook

In the short term, market observers will scrutinize Moore Thread’s next product performance. Mid‑term competition will intensify with domestic rivals like Biren, while long‑term success hinges on breakthroughs in high‑end manufacturing and a robust, self‑sufficient ecosystem. The convergence of capital support, national strategy, and exploding AI compute demand creates a golden window for China’s GPU industry to transition from "domestic substitution" to global competition.