半导体巨头业绩快报释放"芯"动能 国产算力步入价值兑现阶段 有业内人士认为，算力芯片行业具有行业壁垒高、重研发投入、研发周期长等特征，企业成长普遍遵 循"研发攻坚—营收放量—亏损收窄—盈利兑现"的发展路径。最新的业绩快报显示，国产算力在AI需求 的爆发下正走向价值兑现阶段 ◎记者 郑维汉 2月27日晚，寒武纪、摩尔线程、沐曦股份三家半导体"巨头"发布2025年度业绩快报，在人工智能及算 力需求的强劲拉动下，三家公司业绩均表现亮眼。 寒武纪在营收同比大幅增长超4倍的背景下，2025年净利润达到20.59亿元，同比实现扭亏为盈。去年12 月先后登陆科创板的摩尔线程、沐曦股份，交出上市后首份年度业绩快报，在营收实现倍增的同时，两 家公司净亏损均同比明显收窄。 算力需求持续攀升 寒武纪收入同比大幅增长 2月27日晚，国产ASIC龙头寒武纪发布2025年度业绩快报。报告期内，公司实现营业总收入64.97亿 元，同比增长453.21%；归属于母公司所有者的净利润20.59亿元，上年同期净利润亏损4.52亿元，2025 年度实现同比扭亏为盈。 寒武纪表示，受益于人工智能行业算力需求的持续攀升，公司凭借产品的优异竞争力持续拓 ...

摩尔线程表示，得益于人工智能产业蓬勃发展及市场对高性能GPU的强劲需求，公司以AI训推一体智 算卡MTT S5000为代表的产品竞争优势进一步扩大，市场关注与认可度持续提升，推动收入与毛利增 长，整体亏损幅度同比收窄。 北京商报讯(记者王蔓蕾)2月27日晚间，摩尔线程(688795)发布2025年度业绩快报显示，公司全年实现营 业总收入约为15.06亿元，同比增长243.37%；对应实现归属净利润约为-10.24亿元，上年同期为-16.18亿 元。 值得关注的是，摩尔线程旗舰级产品智算卡MTT S5000正加速规模化量产。基于该产品搭建的大规模集 群已上线服务，可高效支持万亿参数大模型训练，计算效率达到同等规模国外同代系GPU集群先进水 平，且具备全精度、全功能的通用计算能力，成为驱动其业绩增长与商业化落地的重要引擎。 ...

Core Viewpoint - The company, Moore Threads, anticipates significant revenue and gross profit growth for 2025, driven by the booming AI industry and strong demand for high-performance GPUs [2][9]. Financial Performance - Moore Threads projects 2025 revenue between 1.45 billion and 1.52 billion yuan, representing a growth of 230.70% to 246.67% compared to 2024 [9]. - The company expects a non-GAAP net loss of 1.04 billion to 1.15 billion yuan for 2025, with a year-on-year reduction in losses of 29.59% to 36.32% [9]. - This marks the fourth consecutive year of revenue growth and narrowing net losses, reflecting the company's enhanced core competitiveness and accelerated commercialization [9]. Product Development and Technology - The GPU industry is characterized by high barriers to entry, significant R&D investment, and long development cycles, requiring comprehensive R&D capabilities [10]. - Moore Threads has developed a complete, sustainable GPU ecosystem through self-designed architecture and tools, establishing a unique "AI + graphics" dual-drive positioning [10]. - The company has successfully mass-produced five chips within five years and completed four generations of GPU architecture, creating a diverse product matrix that supports AI, scientific computing, and graphics rendering [11]. Performance Metrics - The flagship AI training and inference GPU card, MTT S5000, has achieved market-leading performance and has been mass-produced [11]. - A large-scale cluster based on MTT S5000 has been built, achieving a floating-point computing power of 10 Exa-Flops, with a training efficiency of 60% for Dense models and 40% for MoE models [11][12]. - The MUSA architecture and software stack are highly compatible with international GPU application ecosystems, reducing migration costs for developers [12]. Strategic Initiatives - Moore Threads plans to use its own funds to meet the financial needs of its fundraising projects, ensuring the smooth progress of GPU R&D [13]. - The fundraising will focus on three core R&D projects: next-generation AI training and inference chips, next-generation graphics chips, and next-generation AI SoC chips [13]. - This funding strategy reflects the company's confidence and determination to succeed in the new GPU development cycle, reinforcing its technological barriers in the domestic GPU market [13].

Core Viewpoint - The article highlights the significant growth forecast for Moer Thread's revenue and gross profit by 2025, driven by the booming AI industry and strong demand for high-performance GPUs [2]. Group 1: Financial Performance - Moer Thread anticipates revenue between 1.45 billion to 1.52 billion yuan in 2025, representing a growth of 230.70% to 246.67% compared to 2024 [2]. - The company expects a narrowed net loss of 1.04 billion to 1.15 billion yuan in 2025, with a year-on-year reduction of 29.59% to 36.32% [2]. - This marks the fourth consecutive year of revenue growth and reduced net losses, reflecting the company's improving core competitiveness and accelerated commercialization [2]. Group 2: Technological Advancements - Moer Thread has established a complete, sustainable GPU system through self-designed processes from architecture to software ecosystem, positioning itself as a rare "AI + graphics" dual-driven GPU manufacturer in China [4]. - The company has successfully mass-produced five chips within five years and completed four generations of GPU architecture, creating a diverse product matrix that supports AI, scientific computing, and graphics rendering [4]. - The flagship AI training and inference GPU card, MTT S5000, has achieved market-leading performance and large-scale production, with a floating-point computing capability of 10 Exa-Flops [5]. Group 3: Ecosystem and Collaborations - Moer Thread's MUSA architecture and software stack are highly compatible with international mainstream GPU application ecosystems, significantly reducing developers' migration costs [5]. - The company has validated its capabilities through key collaborations, achieving record performance in large-scale models and demonstrating the effectiveness of its domestic AI infrastructure [6]. - The company has invested over 4.3 billion yuan from 2022 to mid-2025 to support continuous innovation and ecosystem expansion [6]. Group 4: Funding and Future Projects - Moer Thread announced it will use its own funds to meet the financial needs of its fundraising projects, ensuring the smooth progress of GPU R&D [7]. - The funding will focus on three core R&D projects: next-generation AI training and inference chips, graphics chips, and AI SoC chips [7]. - This decision reflects the company's confidence and determination to succeed in the new GPU development cycle, reinforcing its technological barriers in the domestic GPU market [7].

摩尔线程表示，得益于人工智能产业蓬勃发展及市场对高性能GPU的强劲需求，公司以AI训推一体智 算卡MTT S5000为代表的产品竞争优势进一步扩大，市场关注与认可度持续提升，推动收入与毛利增 长，整体亏损幅度同比收窄。 值得一提的是，在发布业绩预告的同时，摩尔线程还于同日发布公告，宣布将使用自有资金补足募投项 目的资金需求，保障GPU研发项目顺利推进。根据公告，摩尔线程的募投资金将投向三大核心研发项 目：新一代自主可控AI训推一体芯片、新一代自主可控图形芯片、新一代自主可控AI SoC芯片。 北京商报讯（记者 王蔓蕾）1月21日晚间，摩尔线程（688795）发布2025年业绩预告显示，公司全年归 属净利润预计亏损9.5亿元到10.6亿元，与上年同期相比，亏损收窄幅度为34.5%—41.3%。 业绩预告显示，经财务部门初步测算，摩尔线程预计2025年实现营业收入14.5亿—15.2亿元，与上年同 期相比，收入增长幅度为230.7%—246.67%；预计实现归属净利润亏损9.5亿—10.6亿元，与上年同期相 比，亏损收窄幅度为34.5%—41.3%；预计实现扣非后归属净利润亏损10.4亿—11.5亿元，与上年同期 ...