国 增 科 技 | 人工智能Guozeng Tech | AI
从算力租赁到硬件采购，从芯片级维修到模型部署——国增科技提供一站式AI算力解决方案From GPU rental to hardware procurement, chip-level repair to model deployment — one-stop AI compute solutions
全系列GPU弹性租赁，支持容器、云主机、裸金属三种形态，按需计费Full range elastic rental — container, VM, bare-metal, pay-as-you-go
全系列PCIe加速卡供应，算力+模型一步到位Full range PCIe accelerators — compute + model, one step
国内少数具备SXM形态GPU板卡级修复能力的技术服务商，覆盖全系列GPUOne of few providers with SXM GPU board-level repair — full GPU coverage
模型开发、训练优化、高效部署一站式服务，配套Goflow可视化工作流设计器Model dev, training optimization, deployment — with Goflow visual workflow designer
覆盖内蒙、深圳、广州、北京、上海、香港等数据中心，H100/A100/4090/5090全系列Covering Inner Mongolia, Shenzhen, Guangzhou, Beijing, Shanghai, Hong Kong — H100/A100/4090/5090
支持云主机 · 容器 · 裸金属 三种部署形态 | SSH / VNC / Jupyter 多种连接方式VM · Container · Bare-Metal | SSH / VNC / Jupyter Access
聚焦主流AI模型，提供弹性算力调度与可视化工作流，助你高效开发与部署Focus on mainstream AI models with elastic compute scheduling and visual workflows — empowering efficient development and deployment
芯片级深度维修能力 · 数字化ERP全流程溯源 · 7-30工作日快速交付 · 60天超长质保Chip-Level Repair · ERP Traceability · 7-30 Day Turnaround · 60-Day Warranty
GPU核心植球、板卡级修复，解决H100/A100等高端卡掉卡故障GPU core reballing, board-level repair for H100/A100 drop issues
GPU功耗异常、PCB短路、连接器损坏等深度电路维修Power anomalies, PCB shorts, connector damage — deep circuit repair
GPU互联故障、GPU间通信带宽恢复Interconnect issues, inter-GPU bandwidth recovery
超微/戴尔/华硕/技嘉等品牌服务器主板、电源、散热系统维修Supermicro/Dell/ASUS/Gigabyte server motherboard, PSU, cooling repair
提交故障信息Fault Info
工程师诊断+报价Diagnosis + Quote
芯片级修复+压测Chip Repair + Test
寄回验收+保修Return + Warranty
核心成员毕业于清华、北大、中科大等高校，曾就职微软、谷歌、腾讯、阿里、字节、华为，90%以上研发人员具有研究生学历Tsinghua, PKU, USTC alumni — ex Microsoft, Google, Tencent, Alibaba, ByteDance, Huawei. 90%+ hold graduate degrees
团队源自富士康产线、显卡模组，具备GPU芯片级深度维修、BGA返修、SXM模块测试等核心能力Foxconn module heritage — BGA rework, SXM testing, chip-level repair
在内蒙、深圳、广州、北京、上海、香港等地部署数据中心，支持32-256台H100集群弹性调度Data centers across Inner Mongolia, Shenzhen, Guangzhou, Beijing, Shanghai, Hong Kong — 32-256 H100 clusters
与清华深研院等建立联合实验室，服务清华、北大、中科院等顶级高校及科研机构Joint labs with Tsinghua SIGS — serving Tsinghua, PKU, CAS and top research institutions
自研智算小程序 + GPU维修ERP系统，从算力租用到维修全流程在线操作Self-developed AI compute mini-app + GPU repair ERP — end-to-end online management
数据安全存储，弹性收费，无需硬件采购部署，按需分配按量计费Secure data storage, elastic billing, no hardware procurement — allocate on demand
配备2台H100-80G SXM ×8服务器，不到2个月完成4次效果迭代2× H100 SXM ×8 servers — 4 model iterations in under 2 months
租用4090八卡裸金属服务器，训练"BELLE"和"OpenBuddy"等垂类大模型4090 8-GPU bare-metal — trained BELLE & OpenBuddy vertical LLMs
租用A100 40G服务器半年，部署Stable Diffusion实现文生图、图生图A100 40G for 6 months — deployed Stable Diffusion for text-to-image
租用H100 80G 8卡服务器一年，加速国防AI应用开发H100 8-GPU annual rental — accelerated defense AI development
提供4090、A100等灵活云服务，支持训练vicuna等热门AI模型4090, A100 flexible cloud — training vicuna and popular AI models
修复4090黑屏、H100掉核心、A100掉卡、H800 ECC报错等Fixed 4090 black screen, H100 core drop, A100 drop, H800 ECC errors
国增科技（GZKJ）是一家专注于人工智能算力服务的技术型企业，成立于2015年，总部位于香港。我们一直专注于人工智能领域，致力于为全球的企业和客户提供从底层硬件到上层模型部署的一站式AI算力解决方案。Guozeng Tech (GZKJ) is a technology company specializing in AI compute services, founded in 2015 and headquartered in Hong Kong. We focus on the AI sector, providing global enterprises and clients with one-stop AI compute solutions — from underlying hardware to top-layer model deployment.
具有超10000P算力调度能力、并行推理优化能力，服务于AI算力集群、AIGC模型训练、科学计算、图形图像处理、AI应用落地等场景。With 10,000+ P compute scheduling and parallel inference optimization — serving AI clusters, model training, scientific computing, graphics, and AI application deployment.
助力开发者孵化伟大的AI产品Empower developers to build great AI products
用户体验第一，安全、稳定、实惠为本UX first — secure, stable, affordable
为全球AI企业和开发者提供安全、稳定、实惠的算力服务Provide secure, stable, affordable compute for global AI companies and developers
中国香港 · 0085226524800
energy@verrallc.netHong Kong · 0085226524800
energy@verrallc.net
立即联系我们，获取专属算力方案与最新报价Contact us for a custom compute plan and latest pricing