從芯片到 Token 工廠From Silicon to Tokens
Token 工廠的開發與營運、token 供應及國際業務的平台。The platform for developing and operating token factories, token supply and international business.
算力網絡的建設與營運，提供算力雲與行業模型應用。Builds and operates the compute network, providing cloud compute and industry model applications.
人工智能、芯片與算力領域的投資、產業網絡與專家委員會。Investment, industry network and the Senior Advisory Board across AI, silicon and compute.
芯片產業根基Chip Industry Roots
團隊具 12 英寸晶圓製造的經營經驗，可同時評估芯片供給、製造產能與部署方案。The team has operating experience in 300mm wafer manufacturing and can assess chip supply, manufacturing capacity and deployment together.
在營算力業務Compute in Operation
算力網絡已投入營運，團隊具備液冷集群部署、多集群統一調度與推論優化的實際經驗。The compute network is in operation, giving the team hands-on experience in liquid-cooled cluster deployment, unified scheduling and inference optimisation.
算電協同Power and Compute Together
智算園區選址於電源側，電力與園區同步規劃，並與風電、光伏及儲能發展商合作長期供電，縮短從電力落實到投產的周期。Campuses are sited at the power source and planned together with generation, with wind, solar and storage developers supplying power long term, shortening the path from secured power to live capacity.
全鏈條營運One Operator, End to End
能源、數據中心、算力與 token 供應由同一團隊建設與營運，既持有容量，也對每個 token 的效率負責。Energy, data centres, compute and token supply are built and operated by one team, which holds the capacity and is accountable for the efficiency of every token.
推論效率Inference Efficiency
以每瓦產出的 token 為目標，從加速運算選型、液冷部署到模型量化與並行，逐層優化推論效率。Designed around tokens per watt, with inference efficiency tuned at every layer, from accelerated computing and liquid cooling to quantisation and parallelism.
國際平台International Platform
以香港為平台承接國際客戶合約，並與亞太地區夥伴推進數據中心與算力項目。Hong Kong is the platform for international client contracts and for data-centre and compute projects with partners across Asia-Pacific.
業務板塊Businesses
核心業務是向國際領先的 AI 企業供應 token 與推論產能，方式包括專屬產能與 Token 供應；算力服務與算力營運則把同一能力延伸至模型企業與算力持有方。The core business is supplying tokens and inference capacity to leading AI companies, through dedicated capacity or per-token supply. Compute services and compute operations extend the same capability to model companies and compute owners.
專屬產能Dedicated Capacity
為國際領先的 AI 企業建設並營運專屬智算中心，數據中心、伺服器與營運維護一併交付。Dedicated AI data centres built and operated for leading AI companies, with data centre, servers, operations and maintenance delivered together.
- 客戶CLIENTS
- 國際領先的模型企業與雲服務商Leading model companies and cloud providers
- 計價PRICING
- 五年期合約，按 GPU 小時或按兆瓦Five-year contracts, per GPU-hour or per MW
Token 供應Token Supply
以推論 API 按 token 供應推論產能，吞吐量與延遲指標列入服務水平協議。Inference capacity supplied per token through the inference API, with throughput and latency written into the service level agreement.
- 客戶CLIENTS
- 模型企業與 AI 應用企業Model and AI application companies
- 計價PRICING
- 按 tokenPer token
算力服務Compute Services
液冷算力集群與算力雲，用於訓練、微調與推論，租戶隔離。Liquid-cooled clusters and cloud compute for training, fine-tuning and inference, with tenant isolation.
- 客戶CLIENTS
- 模型企業、機器人與行業應用企業Model, robotics and industry application companies
- 計價PRICING
- 按 GPU 小時或按用量Per GPU-hour or pay as you go
算力營運Compute Operations
為數據中心及算力持有方營運集群，將現有算力轉化為穩定的 token 產能。Cluster operations for data centre and compute owners, turning existing compute into steady token output.
- 客戶CLIENTS
- 數據中心及算力持有方Data centre and compute owners
- 計價PRICING
- 按營運服務合約Under operating service contracts
Token 工廠The Token Factory
AI 產業可分為能源、芯片、基礎設施、模型與應用五個層級。Token 工廠涵蓋前四層，以電力為輸入、以 token 為產出，供應應用層，智算中心是其物理載體。聯合創芯在前四層均具備能力，以每瓦產出的 token 與每個 token 的成本衡量效率。AI can be described in five layers: energy, chips, infrastructure, models and applications. A token factory spans the first four, taking power in and putting tokens out for the application layer, with the AI data centre as its physical form. UNIC has capabilities across all four and measures efficiency in tokens per watt and cost per token.
智算中心AI Data Centres
聯合創芯在多個市場開發建設大型智算園區。園區的選址、技術標準、規模與能源安排，以主要客戶的實際部署需求與交付時間為依據，在滿足驗收要求的前提下控制單位投資。UNIC develops large AI data-centre campuses in several markets. Siting, technical standards, scale and energy follow the actual deployment needs and timelines of anchor clients, with unit cost controlled within their acceptance requirements.
跟隨客戶集群Near Client Clusters
跟隨主要雲服務商已投運園區的分佈，選擇電力接入有保障、審批環境穩定的市場，落點與客戶共同確定。Sites follow the established campuses of leading cloud providers, in markets with reliable grid access and stable approvals, and are chosen with customers.
按客戶設備設計Designed for Client Equipment
按客戶當期部署的機櫃級系統確定標準，採用冷板液冷與機電預製模組化建造，縮短建設周期，並在驗收要求之內控制單位投資。Standards follow the rack-scale systems customers are deploying, with cold-plate liquid cooling and prefabricated modular MEP that shorten the build, and unit cost kept within acceptance requirements.
按已簽負載分期Phased to Signed Load
大型園區分期建設，每一期規模按已簽負載確定，投產即有負載。Large campuses are built in phases, each sized to signed load, so capacity carries load from the day it opens.
可再生能源長期供電Long-Term Renewable Power
與風電、光伏及儲能發展商簽訂長期購電協議；電力容量與併網安排落實之前，項目不啟動建設。Long-term power purchase agreements with wind, solar and storage developers; no construction starts before grid capacity and connection are secured.
產業網絡Industrial Network
產業網絡以智算中心為主線，分為上游的能源與芯片、中游的智算中心建設與營運，以及下游的模型與客戶。The industrial network follows the AI data centre: energy and chips upstream, building and operating data centres in the middle, and models and clients downstream.
能源與芯片Energy and Chips
- 風電、光伏及儲能發展商以長期購電協議供電。Wind, solar and storage developers supply power under long-term agreements.
- 12 英寸晶圓製造的經營經驗，熟悉芯片供給、產能與採購。Operating experience in 300mm wafer manufacturing, with a working knowledge of chip supply, capacity and procurement.
- 產業網絡覆蓋記憶體、封裝測試、設備與材料企業。The network covers memory, packaging and test, equipment and materials.
智算中心AI Data Centres
- 半導體產業園的規劃、建設與招商經驗，用於智算園區開發。Experience planning, building and attracting tenants to a semiconductor industrial park, applied to AI campus development.
- 冷板液冷與機電預製模組化建造。Cold-plate liquid cooling and prefabricated modular MEP.
- 算力網絡已投入營運，具多集群統一調度能力。A compute network in operation, with unified scheduling across clusters.
模型與客戶Models and Clients
- 主要客戶為國際領先的模型企業與雲服務商，於開工前簽訂長期合約。Anchor clients are leading model companies and cloud providers, signing long-term contracts before construction.
- 智能代理平台與行業模型應用，覆蓋金融、醫療與製造。An agent platform and industry model applications for finance, healthcare and manufacturing.
- 產業網絡覆蓋自動駕駛運算與影像感應企業，連接汽車與機器人場景。The network covers autonomous-driving compute and image-sensor companies, linking to automotive and robotics use.
團隊Team
團隊由管理團隊、算力技術團隊與專家委員會組成，經驗集中在 AI 算力的三個環節：智算中心的建設與營運、算力調度與推論服務，以及 AI 芯片與大模型基礎設施。The team comprises management, compute engineering and the Senior Advisory Board. Its experience is concentrated in three parts of AI compute: building and operating AI data centres, compute scheduling and inference services, and AI silicon and model infrastructure.
董事長 · 專家委員會主任Chairman · Chair of the Senior Advisory Board
長期主持大型半導體企業的經營管理，並在多所大學及人工智能學院擔任兼職教授。主導聯合創芯的戰略方向與產業合作。Has long led large semiconductor enterprises and is an adjunct professor at several universities and an academy of artificial intelligence. Leads UNIC's strategic direction and industry partnerships.
管理合夥人Managing Partner
負責智算中心的開發建設、客戶合作與國際業務。具晶圓製造及新能源企業的高層管理經驗，熟悉芯片供應鏈、產能規劃與電力配套，長期參與 AI 芯片企業的發展。Leads AI data-centre development, client partnerships and international business. Senior management experience at a wafer manufacturer and a new-energy company, covering chip supply chains, capacity planning and power supply; long involved with AI chip companies.
海外合夥人International Partner
常駐矽谷，負責北美項目與技術合作，長期關注人工智能、智能汽車與智能製造；曾在國際網絡通信企業擔任管理職務。電機工程博士。Based in Silicon Valley, responsible for North American projects and technology partnerships, with a long focus on AI, intelligent vehicles and smart manufacturing; previously held management roles at a global networking company. PhD in Electrical Engineering.
集群部署與運維Deployment and Operations
液冷高密度集群的部署、交付驗收與日常運維，覆蓋伺服器、網絡與冷卻系統。Deployment, acceptance and day-to-day operation of high-density liquid-cooled clusters, across servers, networking and cooling.
統一調度與算力雲Scheduling and Cloud Compute
多集群統一調度與多租戶隔離，承載算力雲與專屬產能的交付。Unified scheduling across clusters and multi-tenant isolation, carrying the delivery of cloud compute and dedicated capacity.
推論優化與模型服務Inference and Model Services
按模型進行量化、KV Cache 與並行優化，提供推論 API、智能代理平台與行業模型應用。Quantisation, KV cache and parallelism tuned per model, delivering inference APIs, an agent platform and industry model applications.
合作夥伴Partners
聯合創芯與以下六類夥伴合作，各方在智算園區與算力業務中承擔不同角色。UNIC works with six kinds of partner, each with a distinct role in its campuses and compute business.
國際領先的 AI 企業Leading AI Companies
以五年期合約取得專屬產能，或按 token 採購推論服務；主要客戶於開工前簽約。Take dedicated capacity on five-year contracts or buy inference per token; anchor clients sign before construction.
可再生能源發展商Renewable Energy Developers
以風電、光伏與儲能參與園區建設，並長期供電。Take part in campuses with wind, solar and storage, and supply power long term.
長期持有方Long-Term Holders
家族辦公室與國際金融機構參與園區的長期持有與項目融資。Family offices and international financial institutions take part in long-term ownership and project finance of campuses.
算力持有方Compute Owners
委託聯合創芯營運集群，將現有算力轉化為 token 產能。Entrust cluster operations to UNIC and turn existing compute into token output.
產業與技術夥伴Industrial and Technology Partners
在自有集群上完成部署與基準測試，並對接晶圓、封測產能與供應鏈資源。Run deployment and benchmark tests on UNIC's clusters, with access to wafer and packaging capacity and supply-chain resources.
國際夥伴International Partners
以香港平台合作推進亞太地區的數據中心與算力項目。Advance data-centre and compute projects across Asia-Pacific through the Hong Kong platform.