Make compute capacity standardized, tradable, and priceable, and establish benchmarks for institutional procurement and supplier operations.
Beidou BixinBEIDOU BIXIN SMART COMPUTE GROUP
Beidou BixinSmart Compute Group Ltd.
The Group focuses on AI infrastructure and enterprise model services, developing technology and products across compute standards, supply organization, price discovery, metering, and fulfillment. KAI Compute Capacity Fulfillment Platform is the Group’s core product.
Technical TeamsShanghaiBeijingShenzhenHong KongSingapore
Group UpdateBeidou Bixin Joins the Inaugural Global Unicorn Summit Roundtable in Hong Kong
View Details30+Enterprises Served
7Capacity Suppliers Integrated
24,000+Lots of Capacity Delivered
1,600+Fulfilled Capacity Orders
99.999%+Peak-Period Capacity Assurance
Interim business operating data. Fulfilled capacity orders are enterprise purchases with completed service fulfillment and recorded outcomes.
Company UpdatesNEWS & INSIGHTS
View AllABOUT THE GROUP
Priced, Metered, and Fulfillable Smart Compute Services for Enterprises
Beidou Bixin Smart Compute Group Ltd. focuses on AI infrastructure, intelligent-compute resource organization, and enterprise model services. With technical teams in Shanghai, Beijing, Shenzhen, Hong Kong, and Singapore, the Group continues to build capacity standards, supply organization, metering, delivery, and fulfillment governance.
KAI Compute Capacity Fulfillment Platform is the Group’s core product in this field. It organizes scheduled-capacity products and GPU-hour products, aggregates real bids and trades from capacity suppliers and institutional buyers, and forms capacity prices by product, region, and time window.
Build neutral, transparent, and trusted capacity-fulfillment infrastructure where suppliers and institutional buyers form prices through continuous trading.
Zhang Hongquan张宏泉
2022 Shanghai “Most Beautiful Veteran” honoree · Standing Committee Member of the CPPCC Shanghai Songjiang District Committee
TECHNICAL PRESENCE
Technical Teams Across Five Cities
The Group has technical teams in Shanghai, Beijing, Shenzhen, Hong Kong, and Singapore, working together across platform engineering, compute products, metering, and fulfillment.
PRODUCTS & SOLUTIONS
The Group’s Compute Products and Projects
The Group’s projects span model-service assurance, intelligent workspaces, GPU-hour markets, compute fulfillment, resource operations, index research, and real-time pricing.
PROJECT 01
COMPUTE CAPACITY FULFILLMENT
KAICompute Capacity Fulfillment Platform
Connects institutional buyers and capacity suppliers through common product, metering, and accountability terms, aligning market prices and capacity positions with actual fulfillment.
PROJECT 02
CROSS-PLATFORM GPU SERVICE
TODGPU-Hour & Graphics Service Platform
A cross-platform product for card-hour assets and hourly rental: web supports market decisions and operations, while mobile handles balances, alerts, and on-the-go rental through a shared API and ledger.
PROJECT 03
COMPUTE INDEX & RISK
EODCompute Index Quant Platform
Moves GPU quotations, capacity, and delivery data through collection, analysis, governance, index research, and stress testing to produce reviewable compute indexes, price views, and risk-control workflows.
PROJECT 04
STANDARD CARD-HOUR MARKET
PODGPU-Hour Compute Marketplace
Uses standardized GPU SKUs and one GPU × one hour as the trading unit, taking buyers and suppliers from filtering and listing through delivery, acceptance, metering, and RMB settlement.
PROJECT 05
GPU WEALTH & AI WORKSPACE
CODGPU Asset Operations & AI Workspace
Starts with a traceable GPU-hosting ledger that connects card-hour pricing, devices, workloads, and settlements, then extends the same workspace into chat, code, and multi-model comparison.
PROJECT 06
MODEL SERVICE ASSURANCE
KODModel-Service Capacity Assurance Assistant
Plans, configures, and monitors model-service capacity for critical operating periods, linking provider comparison, usage management, and backup capacity across desktop, mobile, and web.
PROJECT 07
MULTI-PARTY FULFILLMENT
NODCompute Fulfillment Collaboration Platform
Separates customer, agent, and supplier portals while connecting demand, workloads, actual GPU hours, commissions, and supplier settlement under one auditable business reference.
PROJECT 08
GPU RESOURCE OPERATIONS
SODCompute Resource Operations Platform
Brings agency services, GPU hosting, card-hour balances, timed resource offers, and popularity rankings into a mobile-first operating entry for devices, quotations, and asset activity.
PROJECT 09
REAL-TIME GPU PRICING
PricingGPU-Hour Pricing
Presents current GPU-hour quotations by model and region, with filters for performance, location, price, and form factor, plus a real-time JSON API and OpenAPI specification for standardized downstream access.
KAI PRODUCT DETAILS
Fulfillment Product Standards
These two standardized capacity products make up the KAI fulfillment framework in practice.
CAPACITY PRODUCT A
SCHEDULED CAPACITY PRODUCT
Scheduled-Capacity Products
Enterprises do not procure abstract compute. They procure a service commitment with a defined model, time, region, quantity, quality, and accountable party.
Specified Model & VersionSpecified Fulfillment HourDefined Service RegionAgreed Service Standard
STANDARD CAPACITY UNIT1 lot = 1,000,000 TPMw × 1h
CAPACITY PRODUCT B
STANDARDIZED GPU CAPACITY
GPU Hours
Specify GPU type, quantity, region, and fulfillment hour; support listing, quoting, procurement, transfer, and fulfillment.
Specified GPU TypeDefined GPU QuantityAgreed Service RegionFixed Fulfillment Hour
STANDARD METERING UNITN GPU × 1h
MODEL / GPU TYPECAPACITYTIME WINDOWREGIONSERVICE STANDARDACCOUNTABLE PARTY
PLATFORM CAPABILITIES
A Common Product, Pricing, and Fulfillment Framework for Compute
KAI connects capacity suppliers and institutional buyers under shared product and metering terms, providing a common foundation for procurement, supply organization, and real fulfillment.
Standardized Products
Organize capacity by model or GPU type, time window, region, and service standard under comparable product terms.
Price Discovery
Aggregate real quotations and trades from suppliers and institutional buyers to form prices by product, region, and time.
Unified Metering
Measure scheduled model capacity in weighted TPM and underlying compute in GPU hours.
Fulfillment Support
Use positions, cut-off, metering records, and exception handling to establish clear and verifiable fulfillment evidence.
METERING & FULFILLMENT
Weighted TPM for Scheduled Capacity; GPU Compute by the Hour
Unified metering places comparable capacity from different suppliers in one framework and creates a clear link among price, holdings, and actual fulfillment.
SCHEDULED-CAPACITY FORMULA
TPMw = I + W × O
I = Input Tokens · O = Output Tokens
W = Fixed output weight for the same model
ONE SCHEDULED-CAPACITY LOT1,000,000TPMw × 1h
Unified Windows
Inputs and outputs are counted in consecutive, non-overlapping 60-second windows.
Model-Fixed Weight
Output weight is set by model; the same model always uses the same weight.
Versioned Rules
Cache hits and reasoning Tokens count normally; failed retries and cross-window requests follow the rules effective for the relevant position.
GPU-Hour Metering
Meter available compute by specified GPU type, quantity, region, and fulfillment hour.
IMPORTANT BOUNDARY
One lot is not a Token wallet. Capacity cannot accumulate across minutes or carry into another hour; it measures continuously available model-processing load.
PRODUCT LIFECYCLE
From Product Admission to Hourly Fulfillment
The full rule set spans products, quotes, holdings, cut-off, and fulfillment, ensuring clear accountability for every capacity trade.
Product Admission
Define the model service or GPU type and select products with clear demand and standardizable attributes.
Supplier Qualification
Assess legal entity, service capability, continuity, contingency plans, and accountability.
Capacity Registration
Register sustainable future capacity net of self-use, maintenance, and required redundancy.
Product Listing
Standardize model or GPU type, capacity, time window, region, and service standard.
Supplier Quotes
Publish prices and validity periods based on cost, time window, region, and available capacity.
Procurement & Transfer
Institutions procure scheduled-capacity products or GPU hours and may transfer or top up before cut-off.
Fulfillment Preparation
After cut-off, lock final holdings, callable account quotas, and supplier obligations.
Hourly Service
Enterprises use model capacity or specified GPU compute; suppliers fulfill to standard.
Confirmation & Review
Confirm delivery, settle service fees, handle exceptions, and continuously evaluate suppliers.
BUYERFull Prepayment
Pay upon purchase or transfer; funds are managed under platform rules until fulfillment.
SUPPLIERFulfillment Collateral
Register real capacity and remain accountable for all capacity sold.
EXCEPTIONSubstitution First
Coordinate backup suppliers; the defaulting party bears related costs under platform rules.
FULL PROCESSAuditable Records
Orders, usage, metering, settlement, and exceptions are recorded.
STAKEHOLDER VALUE
Capacity Prices Connect Institutional Procurement and Supplier Operations
DEMAND SIDE
Institutional Buyers
- Capacity Assurance
Secure critical-hour capacity in advance for promotions, launches, service peaks, and concentrated workloads.
- Budget Certainty
Fix future service prices at procurement and incorporate model cost into budgets and client quotes.
- Cross-Supplier Procurement
Use standardized scheduled-capacity and GPU-hour parameters to reduce comparison and coordination costs.
- Procurement Flexibility
Transfer surplus capacity or add time windows before cut-off as demand changes.
KAIPRICE
DISCOVERY
DISCOVERY
SUPPLY SIDE
Capacity Suppliers
- Advance Capacity Sales
Sell future time windows in advance to plan revenue and utilization.
- Revenue Management
Price base, peak, and near-fulfillment windows for model capacity and GPU hours.
- Expansion Signals
Use capacity price curves to guide resource investment and product mix.
- Broader Customer Reach
Standardization reduces repeated work across products, metering, delivery, and fulfillment systems.
Buyers use scheduled-capacity and GPU-hour prices to plan budgets and procurement. Suppliers use time-specific prices to manage revenue and utilization. Every quote, trade, and fulfillment outcome informs the next pricing cycle.
ECOSYSTEM
Industry Collaboration Network
The Group works across model services, compute supply, institutional procurement, and professional services to create clear product, metering, and fulfillment interfaces for enterprise compute needs.
Priority Scenarios
E-Commerce & Livestream PeaksCustomer-Service SurgesConcentrated Coding AgentsBatch Content & ResearchEducation & ExamsFinancial Research & RiskEnterprise End-of-DayGame Launches & Events
GOVERNANCE
Beyond Price Discovery, Trusted Fulfillment Governance
Admission, collateral, evaluation, and exception handling clarify accountability, so prices rest on defined standards and verifiable fulfillment records.
Delivery Risk
Real registration, collateral, backup suppliers, and compensation for default.
Quality Risk
Unified type, version, and delivery standards with independent review.
Credit Risk
Buyer prepayment, supplier collateral, and settlement after fulfillment.
Liquidity Risk
Limit the number of core products and introduce anchor participants.
Concentration & Product Risk
Limits, position controls, related-account checks, and transparent type and version management.
Conflicts of Interest
Public rules, operational separation, and external review.
Service FirstCoordinate substitute capacity first
Clear AccountabilityPrevent unfulfilled capacity from being sold twice
Bounded LossPlatform rules define compensation and liability caps
Layered ResolutionApply service fees and collateral under the rules
Continuous EvaluationFulfillment records affect limits and admission
NEWSROOM
Newsroom
Updates from KAI across industry engagement, ecosystem dialogue, and technology development—grounded in real compute supply and application environments.
Beidou Bixin Joins the Inaugural Global Unicorn Summit Roundtable in Hong Kong
COO Lucas Shen joined industry leaders to discuss AI compute infrastructure and global innovation, and called for cross-supplier compute fulfillment standards.
Read the full report ↗In-House H100 Systems Arrive
The systems expand internal validation for GPU-hour products, scheduling, performance metering, and fulfillment workflows.
Expanding the In-House Mac mini Cluster
KAI uses its own cluster and network environment for experiments, operational validation, and platform iteration.
Inside Intelligent Computing Infrastructure
KAI and Beidou Spacetime discussed intelligent computing center development, compute supply organization, and platform fulfillment.
Connecting Energy, Data Infrastructure, and Compute
KAI continues to visit operating sites to understand how energy, data, and compute environments work together.
A Working Dialogue on Compute Services and Applications
KAI and Shanghai Shuoyao Technology exchanged views on service organization, technical applications, and enterprise needs.
BUSINESS COOPERATION
Work with the Group
We welcome institutional buyers, capacity suppliers, and industry partners.
BUSINESS EMAILbeidou@kai.com
BUSINESS COOPERATION HOTLINE400 108 2026
OFFICE LOCATIONSSongjiang Smart Compute Center Office Senlan Marketing Office Minhang Technology Office Jinqiao Administration Office Beijing Guomao Financing Office Singapore Investor Relations Office Hong Kong Listing Advisory Office
Beidou Boyang Spatial Information Industrial Park
3450 Shengang Road, Songjiang District, Shanghai
6F, Tower A, South Building, Senlan Meihuan
515 Qifan Road, Pudong New Area, Shanghai
Room 601-8, Building 2
Lane 953 Jianchuan Road, Minhang District, Shanghai
Rooms 408, 502, 503 and 507, Building 4, Financial Data Port
56 Fanchang Road, Pudong New Area, Shanghai
Rooms 424–425, Tower 1, Guomao Office Building
Chaoyang District, Beijing
543 Jelapang Road, #08-66
Singapore 670543
1012, Tsim Sha Tsui Centre
66 Mody Road, Tsim Sha Tsui, Kowloon, Hong Kong