Summary of the Conference Call for Moores Threads Company Overview - Moores Threads focuses on GPU-related products, established in 2020, targeting AI, digital twin, and scientific computing applications [3][4] Product Lines - The product lines cover three main areas: AI intelligent computing, graphics acceleration, and smart SoC [2] - Key products include: - AI computing products: training/inference cards, super node servers, and cross-edge computing clusters [4] - Graphics rendering products: S3,000 series graphics cards for industrial design and digital twin applications [4] - Smart SoC products: A140 computing box and E300 module for AI PC edge applications [5] Financial Performance - Projected revenue for 2024 is 438 million RMB, a year-on-year increase of approximately 2.5 times [2][10] - AI intelligent computing business contributes 77.63% of total revenue, with expectations to achieve profitability by 2027 [2][11] - Average gross margin for 2024 is 70.71%, with AI intelligent computing business gross margin exceeding 70% and desktop graphics cards at 18.99% [14] Sales and Market Presence - In 2024, the company sold three clusters and 2,058 graphics cards, with average prices of 6.113 million RMB for clusters and 60,000 RMB for computing cards [12] - 99.5% of revenue comes from the domestic market, with a direct sales model accounting for 41% and a distribution model for 59% [13] Supply Chain and Inventory - Following entry into the US Entity List in October 2023, the supply chain shifted from wafer production to raw material procurement, with inventory increasing from 275 million RMB to 647 million RMB [15] - Prepayments rose from 76 million RMB to 567 million RMB, indicating preparation for future growth [15] Research and Development - Cumulative R&D expenses exceeded 3.6 billion RMB from 2020 to 2024, significantly higher than revenue in 2024 [16] - The workforce has seen limited growth, with 1,126 employees by the end of 2024, but R&D personnel constitute over 78% of the total [18] Management Team - The management team primarily consists of former Nvidia executives, including CEO Zhang Jianzhong, who previously held senior positions at Nvidia [8] Competitive Landscape - The company’s Musa architecture competes with Nvidia's CUDA and AMD's CDNA, providing a unified GPU acceleration framework [6] - The latest chip series, Pinghu, achieves FP32 computing power of 32 TFlops, about half of Nvidia's H100, and has overcome challenges in high-efficiency interconnects [7] Industry Insights - The domestic chip industry is expected to recover in Q4 2024, despite current pressures from US restrictions on Southeast Asian AI chip imports [21] - Recommended companies for investment include Haiguang, Cambrian, and Longxin Zhongke, which are positioned to benefit from domestic computing center projects [22] Risks - Key risks include uncertainties from US-China relations, underperformance of fundamentals, and intensified market competition [26]