GPUBare Metal Server

KoukaryokuPHY

Bare metal servers with NVIDIA B200 Tensor Core GPUs
Designed for complex dataset analysis and large-scale model training

Maximum GPU Performance
from Bare Metal Servers

Koukaryoku PHY comes with eight NVIDIA B200 Tensor Core GPUs, which provide the powerful GPU performance needed for AI development. It also features a high-speed backbone of up to 400 Gbps × 8. This makes it well suited for large language models (LLMs), generative AI, machine learning, and scientific simulations.

Sustainability
Meets High Performance

Our Ishikari Data Center uses renewable energy and has achieved zero annual CO2 emissions. This allows us to run high-performance NVIDIA B200 Tensor Core GPUs for Koukaryoku PHY while keeping the environmental impact to a minimum.

Driving Innovation in Every Field

Koukaryoku PHY accelerates research and business innovation in every field. Use it for training complex models, implementing algorithms, deep learning-based image analysis, natural language processing, scientific simulations, and more.

  • Large Language Models (LLMs)

    For training LLMs used in data analysis, risk prediction, and more

  • Generative AI

    For generating various types of content, including images, text, and audio

  • Scientific Simulation

    For complex scientific calculations and simulations

  • Image Analysis

    For advanced image analysis, such as real-time image processing

Use Case Resources

AI Training Infrastructure
for Autonomous Driving

Autonomous driving is one of the fastest-growing use cases. This paper explains the key concepts and challenges of the GPU infrastructure needed for this field, and introduces solutions using our GPU cloud, Koukaryoku PHY.

Download White Paper

GPU Service Specifications

H100 ModelH200 ModelB200 Model
GPUNVIDIA H100 SXM 80GB x 8NVIDIA H200 SXM 141GB x 8NVIDIA B200 SXM 180GB x 8
CPUIntel Xeon Platinum 8480+ (56 cores) × 2Intel Xeon Platinum 8580 (60 cores) × 2Intel Xeon 6960P (72 cores) × 2
Memory2TB1.5TB1.5TB
Storage7.68TB x 47.68TB x 47.68TB x 4
Global Network «Standard»
*1		«100 Mbps (50 Mbps × 2)»
250 Mbps (125 Mbps × 2)
500 Mbps (250 Mbps × 2)		«100 Mbps (50 Mbps × 2)»
250 Mbps (125 Mbps × 2)
500 Mbps (250 Mbps × 2)		«20 Gbps (10 Gbps × 2)»
Local Network (Sakura Dedicated Server PHY Service Network) «Standard»«4 Gbps (2 Gbps × 2)»
10 Gbps (5 Gbps × 2)
20 Gbps (10 Gbps × 2)		«4 Gbps (2 Gbps × 2)»
10 Gbps (5 Gbps × 2)
20 Gbps (10 Gbps × 2)		«4 Gbps (2 Gbps × 2)»
10 Gbps (5 Gbps × 2)
20 Gbps (10 Gbps × 2)
Local Interconnect (High-Bandwidth Lossless Network)
*2 *3		400 Gbps × 4400 Gbps × 8400 Gbps × 8

  • Bandwidth is provided on a best-effort basis.

  • The interconnect will be set up within three business days, starting from the day after your service begins. We will send you an email when the setup is complete.

  • These are the specifications for the latest model. For specifications of previous models, please click here.

Storage Service Specifications

Koukaryoku Storage SV Series
CPUAMD Turin 9555P UP (64 cores)
Memory768 GB
Storage960 GB × 2 (system area)
30.7 TB × 2 (standard). Optional: you can add drives one at a time, up to 12 drives.
Global Network*120 Gbps (10 Gbps × 2)
Local Network (Sakura Dedicated Server PHY Service Network)
«Standard»		«4 Gbps (2 Gbps × 2)»
10 Gbps (5 Gbps × 2)
20 Gbps (10 Gbps × 2)
Local Interconnect (High-Bandwidth Lossless Network)
*2		400 Gbps × 1

  • Bandwidth is provided on a best-effort basis.

  • The interconnect will be set up within three business days, starting from the day after your service begins.

If you need other options such as storage appliances, please contact us.

Pricing

Please contact us for pricing details.

Service Level Agreement (SLA)

Please see the Koukaryoku PHY Service Level Agreement (SLA) for details.

How to Get Started

  1. Create a New Member Account

    If you do not have a Member ID, please create one first.

    • You will need a phone number to register, since phone number verification is required.

    • If you already have a SAKURA internet Member ID, please go to STEP 2.

    Create a SAKURA internet
    Member ID (Free)

  2. Create a Project
    (formerly called "Account")

    After you log in to the Control Panel, create a Sakura Cloud project.

    • You need to register a credit card to use Sakura Cloud.
      If you have already registered a credit card with another SAKURA internet service, you do not need to register it again.

    • If you have already created a Sakura Cloud project, please go to STEP 3.

    • On April 24, 2025, "Account" was renamed to "Project." The specifications have not changed.

    Create a Project /
    Log In

  3. Payment

    We will send you an invoice. Please check it and complete the payment.

    • The service usually becomes available after your payment is processed.
      If your company has a specific payment schedule, please feel free to contact us.

    • Once we confirm your payment, we will send setup instructions to your registered Member ID.

  4. Start Using the Service

    We will send you setup instructions separately.
    Please follow these instructions to start using the service.

    • You will be redirected to the Sakura Cloud login screen. Enter the account information from STEP 1 and STEP 2 to log in.

    • For detailed instructions, please see the user manual.

    Go to the Server
    Creation

Download Resources

Service Brochure

A complete overview that covers service features, specifications, pricing, and use cases. Download the full Koukaryoku PHY brochure here.

Download Service Brochure

Case Studies

Download case studies of customers who have adopted and are using SAKURA internet's Koukaryoku PHY (GPU Cloud) service.

Download Case Studies

White Paper

A complete guide for organizations planning a GPU computing platform. It explains the key components and considerations to help you choose the right GPU platform for your needs.

Download White Paper

Other GPU Services from SAKURA internet

  • Koukaryoku VRT

    A VM-based GPU cloud service that lets you use high-performance NVIDIA H100 GPUs with the flexibility of cloud-based deployment.

    View DetailsCompare the Koukaryoku Series

  • Koukaryoku DOK

    A container-based GPU cloud service that runs preconfigured Docker images on NVIDIA V100 / H100 GPUs in the cloud.

    View DetailsCompare the Koukaryoku Series

  • SAKURA ONE

    A managed HPC cluster service. You can use the GPU resources of 55 servers at the same time, each with eight NVIDIA H200 GPUs.

    View Details