01 / SIKUMO CLOUD
Your compute. Your way.
NVIDIA GPU capacity for training, fine-tuning and inference. Choose how much control you need.
On-demandReservedAnnualDedicated
Explore productAI CLOUD SERVICES / TOKYO · UNITED STATESSuperintelligence,
Superintelligence,
at your service.
Advanced compute. Open models. Continuous operations. We turn AI infrastructure into intelligence your business can use.
SILICON × KUMOCOMPUTE → INTELLIGENCE
MANUFACTURING INTELLIGENCE, TOKEN BY TOKEN.GLOBAL TOKEN FACTORY OPERATOR
SIKUMO CloudGPU CloudSIKUMO ModelsOpen Model APISIKUMO FactoryPrivate Token FactorySIKUMO HubModel API Hub
OUR PRODUCTSOne Token Factory.
One Token Factory.
Four ways to make it yours.
Start with the capacity, the API, or a complete private environment. Every product builds on the same operating foundation.
02 / SIKUMO MODELS
Open models. Ready to build.
Leading open models, delivered through a standard API and metered by token usage.
Pay per tokenCommitted usageDedicated endpointEnterprise
Explore product03 / SIKUMO FACTORY
Your factory. Your boundaries.
Compute, models and operations brought together in your dedicated environment.
Dedicated hostingOn-premisesHybridManaged operations
Explore product04 / SIKUMO HUB
Many models. One control plane.
Connect open and authorised closed models through one gateway, with unified governance.
Unified billingTeam budgetsRouting policiesEnterprise terms
Explore productTHE OPERATING FOUNDATION
From silicon to useful intelligence.
A Token Factory combines infrastructure, model serving and continuous operations. Its purpose is to make AI delivery repeatable, measurable and more cost-efficient.
01
Compute
GPU resources and cloud-native orchestration.
02
Models
Open-model deployment and fine-tuning.
03
Inference
Scheduling and serving optimisation.
04
Tokens
Metered delivery through four products.
Compute Platform
GPU IaaS · Orchestration · Multi-tenancy
Inference Platform
Model serving · Fine-tuning · Registry
Private AI Stack
Dedicated environments · Data boundaries
AI Gateway
Routing · Metering · Observability
Efficiency at every layer.
Hardware, model and operating efficiency each influence the cost of delivered tokens. We optimise the system as a whole.
Hardware efficiency×Model efficiency×Operating efficiency
THE MODEL STRATEGYOpen at the core.
Open at the core.
More room to build.
Our model strategy centres on open models. Through SIKUMO Hub, authorised closed-model APIs complement the stack when the workload calls for them.
DeepSeekQwenGLMLlamaKimiMiniMax
Model families in our product portfolio. Available versions and capabilities are confirmed at onboarding.
SIKUMO ModelsGLOBAL FOOTPRINTRooted in Japan.
Rooted in Japan.
Built for a global market.
Japan and the United States are the core markets in our infrastructure strategy, with a distinct role for each factory.
JPJAPAN · TOKYO
INFRASTRUCTURE DIRECTION
SIKUMO Tokyo
Training, fine-tuning and high-throughput inference for Japan and Asia-Pacific.
NVIDIA B300Training + inference
USUNITED STATES
INFRASTRUCTURE DIRECTION
SIKUMO US
Inference-focused capacity and API delivery for North America and global customers.
NVIDIA Vera RubinInference-first
Infrastructure roadmap: B300 in Japan and Vera Rubin in the US. These are portfolio directions, not a statement of current availability. Capacity, procurement, acceptance and commercial status are confirmed per project.
THE COMPANYEngineers by background.
Engineers by background.
Operators by conviction.
SIKUMO is a Tokyo-headquartered Token Factory investor and operator. Our founders and core team bring more than a decade of experience in distributed computing, from cloud infrastructure to large language models.
Learn about SIKUMOOUR MISSION
Make AGI create the greatest value for humanity — and make work simple and effective.
10+years in distributed computing
4ways to access AI
JP / UScore markets
BUILD WITH SIKUMOYour next step in AI,
Your next step in AI,
built on solid ground.
Choose a starting point for your workload. Explore the guide or send an enquiry to our sales team.