Moore Threads Targets Cluster Expansion and Iterative GPU Development

People's Daily reported that domestic enterprise Moore Threads has mass-produced 5 chips and launched 5 generations of GPU architectures across the 6 years since its establishment, reflecting the broader expansion of China's artificial intelligence industry. During its startup phase, the company evaluated developing specialized chips for segmented niches to accelerate commercialization, but ultimately chose the full-function GPU route to build long-term computing infrastructure that incorporates diverse computational tasks into its architecture design. To reinforce computing foundations and address critical technological bottlenecks, the enterprise adopted an iterative development strategy centered on a roadmap delivering one chip iteration each year through continuous product development, user feedback, and market testing upgrades. Following the development of its first full-function GPU chip in 2021, proprietary meta-computing unified architecture, and the construction of a computing cluster, the company continues advancing independent research to deliver computing power for industry digital transformations. Looking ahead, Moore Threads chairman and chief executive officer Zhang Jianzhong stated that the enterprise strives to build an ultra-large-scale intelligent computing center cluster to leverage its model training strengths and support technological self-reliance.

Computing competition has shifted from individual card performance to systematic breakthroughs in large-scale cluster capabilities, leading clients to demand hardware-software synergy, mature ecosystems, higher effective computing utilization rates, and sustained stable training duration.