信子科技2026年中期业绩电话会议

发布时间：2026-08-24 一、AI总结内容 一、核心业绩与核心要点 1. 公司概况：信子科技成立于2016年，是全球领先的实时AI基础设施提供商，提供算力、模型、代币的端到端服务，覆盖资管、高端制造、生物医药、工业机器人、航空航天、能源、城市管理、低空经济等多个领域。 2. 核心财务数据：2026年上半年营收9.7亿元，同比增长389%，创历史新高；调整后净利润6700万元，实现盈利，毛利率稳定在6%；海外营收同比增长267%；客户留存率超90%，单客价值达556万元，同比增长204%；人均效能同比增长379%；总资产35亿元，现金储备6.34亿元，资产负债率8.4%，财务结构稳定。 3. 核心业务进展：形成Token 1基础模型平台、Token OS操作系统、Token Cloud模型训练推理部署平台、Token Router场景与模型结果交换器的全栈AI基础设施能力；Token类业务上半年营收占比达11%，已具备商业可行性，成为新增长动力；跨行业渗透成效显著，上半年非资管行业营收占比提升至87.6%，较去年大幅增长，其中电信、智慧城市、高端制造为核心贡献行业。 二、战略规划与后续布局 1. 阶段划分：2025年及之前为1.0阶段，以订阅、项目制模式验证模型可行性；2026年为2.0阶段，以代币化跃迁驱动，形成新增长曲线；2026年下半年将推进3.0阶段，推出Token交易平台，实现代币跨企业、跨行业、跨场景的安全合规流动，升级代币化模型为平台化模型，深化垂直领域数据积累，助力客户部署私有模型与中小企业模型，提供一站式AI基础设施服务。 2. 长期方向：深化业务模型迭代，提升毛利率；加速跨行业复制，开拓高价值新兴行业；探索前沿AI应用；构建算力、算法上下游战略合作生态；稳步扩大海外布局，提升全球市场覆盖。 3. 资本开支：基于客户实际需求、订单可见性、ROI动态评估，投向新领域拓展、软硬一体算力能力建设、Token相关产品研发、模型训练投入等，不盲目扩张。三、投资者问答核心信息 1. 企业AI需求：处于快速发展的早期阶段，需求来自从通用模型简单调用转向具体场景专用模型，客户群体从金融、电信向工业制造等领域扩散；企业AI预算多为战略级投入，管理层重视，成本敏感度较低，将AI视为未来投资而非运营成本，核心目标是实现业务结果转化。 2. 跨行业扩张：当前已进入智能汽车、低空经济等高壁垒行业，正在拓展能源、高端制造、生物医药等领域，计划每年拓展2-3个高价值、高壁垒行业，长期目标深度渗透20余个此类行业。 3. 代币相关：场景代币基于企业专属数据及决策逻辑，定价由其创造的业务价值决定，不同行业差异大；Token OS负责数据处理全链路审计，Token Cloud侧重硬件协调同步，二者协同降低客户部署成本；2026年上半年代币类业务同比大幅增长，预计全年占比达20%-30%。 4. 海外业务：从战略规划转向实质性增长，上半年海外营收高速增长，已完成日本、新加坡子公司注册，重视产品、部署能力的合规性与可复制性，而非短期营收，将重点推进Token OS与代币化商业模式的海外落地。 5. 生态合作：与国内多个GPU厂商、国家级数据交易平台合作，满足企业算力稳定性、数据主权等需求，提供算力租赁、软硬件集成等灵活服务，解决AI落地中的算力、数据、模型等需求。 6. 竞争壁垒：核心壁垒包括跨领域行业Knowhow积累、端到端全栈AI基础设施能力、与客户长期信任关系及高客户粘性；与上下游伙伴为合作关系，通过联合服务客户，专注于垂直场景落地需求。 二、音频原文（转写） Announcement all participants are in listen onlymode and some housekeeping rules this event isintended exclusively for invited mesters to autorecording a written transcript for participantsinternal reference only and must not be publicly distributed chin tsu has not authorized any media outlets throughto publish the contents of this announcement any unauthorized reproduction redistribution or sharing constitutes an infringement infringement of rights and shinto reserves all rights to pursue legal actions we assume no liability or loss rising from any such unauthorized sharing pleasethese exercise caution in the messment.Let me introduce the management team thec t o mr yang yang and chief scientist tangbo and board secretary chen mansion the floor goes to the board secretary miss chenwan xing.Distinguished investors and annalists good morning welcome to our twenty twenty six interim result announcement.It's my honor to represent the company and give you the presentation for our twenty twenty six h one results we seize the.Rising demandof a i and delivered remarkable results i will walk you through our operations financial resultsand outlookto give you more color on our um performance and then we have another q and a session.First let me give you the basics of our company shin tzu technology was founded in twenty sixteen and we are the leading realtime ai infrastructure provider we provide compute model token the.Ah end to end service we started with asset management and now we're present in high end manufacturing biomedicine robotics aerospace energy urban management and low altitude umeconomy and we are gaining penetration in different sectors and we cover the.呃data and a i closed loop.Products we have now more than four hundred of the mod ules and products we have the mainly second level response and one hundred percent data processing accuracy and we have three core token relatedcapability in first half of the year our revenue and profit has broken.Historic record and we hog the top spot.In the industry our revenue wasnine seventy million r and b up by three eightynine percent and our adjusted net profit was sixty seven million r and band we are turning a profit and we have got proven profitability capability and our customer retention rate is more thanninety percent andi want to highlight that ouroverseas revenue grew by two hundred and sixty seven percent.And我是。 Operational metrics all turning better and showing improvement and means that we have been delivering long term growth potential and giving ourclients value.And in the first half of the yearwe have got the a i compute module and uh.And atthe foundational level we have got the token one first a i d p platform and we have got the token one os platform.Giving our clients professional scenario based token and then we have the.Token cloud that has been tested in the first halfis the.Model inference and training platform tosupport the.Training inference and deployment ofthe models so that it can be better deployed inthe business scenarios of the clients and thenfor the token routers it can connect the uh scenario and our model results it serves as an exchange of token sowe are more than just a high qualitytoken manufacturing and now we're expanding into the high value enterprise based smaller models and in the first half of the year we have gotthe enterprise level a i application full stackarchitecture we have got the compute optimization close loop for data