Now Available: GPU Instances for AI Training
South Africa South African Owned

Infrastructure for Intelligence High-Performance SSD Servers for AI & SSO

Deploy AI models and Single Sign-On infrastructure on South Africa's fastest NVMe SSD servers. Sub-5ms latency from Johannesburg & Cape Town. Optimized for TensorFlow, PyTorch, OAuth2, and SAML.

Get Started View Specifications
99.99%
Uptime SLA
< 5ms
Local Latency
7.5GB/s
NVMe Throughput
SA JHB & CPT Datacenters
AS329206
ZA ai-instance-01.ssdserver.co.za
$ nvidia-smi
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.104.05 Driver Version: 535.104.05 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
|=========================================+======================+======================|
| 0 NVIDIA A100 80GB On | 00000000:00:04.0 Off | 0 |
| N/A 35C P0 45W / 300W | 14789MiB / 81920MiB | 0% Default |
+-----------------------------------------+----------------------+----------------------+
$ python model_inference.py --benchmark
Loading LLaMA-2-70B... [OK]
Inference latency (p50): 23ms
Throughput: 847 tok/s
NVMe I/O: 6.8GB/s read
GPU Ready
A100 & H100 Available
SSO Optimized
Keycloak & Auth0 Ready

AI Model Infrastructure

Purpose-built for machine learning training, inference, and model serving. From LLMs to computer vision.

LLM Hosting

Deploy Llama 2, GPT-J, or custom transformers with optimized CUDA drivers and 80GB A100 GPUs.

  • vLLM & TensorRT-LLM
  • quantized model support
  • API inference endpoints

ML Training Clusters

Distributed training with NVLink-capable nodes. PyTorch, JAX, and TensorFlow pre-configured.

  • Multi-GPU scaling
  • InfiniBand networking
  • Checkpoint to NVMe

Computer Vision

High-throughput inference for video analytics, OCR, and image classification pipelines.

  • OpenCV optimized
  • TensorRT acceleration
  • Batch processing

Optimized Software Stack

Pre-installed ML suite with CUDA 12.2, cuDNN, and optimized Python environments. Start training immediately.

Python 3.11
CUDA 12.2
Docker & K8s
Redis & PostgreSQL
# One-click deployment
$
ssdserver deploy ai-standard
→ Provisioning A100-80GB instance...
→ Mounting NVMe storage (2TB)...
→ Loading ML container stack...
✓ Instance ready in 90 seconds
→ IP: 102.37.128.45 (JHB)
Enterprise Identity

SSO & Identity Infrastructure

Host your identity providers, OAuth services, and SAML endpoints on dedicated high-availability infrastructure. Perfect for Keycloak, Auth0 (self-hosted), and custom IAM solutions.

Keycloak Optimized

Pre-tuned JVM settings, Infinispan clustering, and database optimization for high-throughput authentication flows.

Active Directory Hybrid

Low-latency connectivity for hybrid AD setups. Site-to-site VPN included for secure DC replication.

Zero-Trust Ready

Deploy BeyondCorp-style identity-aware proxies with mTLS termination and device trust verification.

Authentication Flow Live
User Login Request
SAML 2.0 Assertion
2ms
IdP Validation
Keycloak Cluster (JHB)
45ms
Token Issuance
JWT + Refresh Token
12ms
Throughput 12,000 auth/min
AS329206
South Africa

South African Hosting Network

Rackzar's autonomous system operates across South Africa with an open peering policy. Carrier-neutral infrastructure with multi-homed connectivity and 100Gbit peering capacity.

Ultra Low Latency

100% redundant Arista gear optimized for high-performance cloud deployments

Major IX Presence

JINX Johannesburg (190+ peers), NAPAfrica JHB (540+ peers), CINX Cape Town (110+ peers)

Direct Cloud Connect

Private peering with Google Cloud, Microsoft Azure, Amazon AWS, and Cloudflare

100Gbit
Peering Capacity
99.99%
Uptime SLA

Internet Exchange Points

JHB JINX Johannesburg
190+ Peers
JHB NAPAfrica JHB
540+ Peers
CPT CINX Cape Town
110+ Peers
CPT NAPAfrica CPT
270+ Peers

Network Security

DDoS Mitigation
RPKI Validation
BGP Security
24/7 NOC
Redundant upstreams: Workonline & Agile
View Network Details →

Hardware Specifications

Enterprise-grade infrastructure with zero oversubscription

NVMe Gen4

7.5GB/s

Sequential Read

Samsung PM1733 / Intel P5800X

DDR4 ECC

Up to 2TB

Per Instance

3200MHz Registered

Networking

25Gbps

Dedicated

JHB ↔ CT < 25ms

CPUs

AMD EPYC

Milan & Genoa

Up to 96 Cores

Available Instance Types

Instance CPU RAM Storage GPU Network
ai.inference.small 8 Cores 32 GB 500GB NVMe - 10Gbps
ai.training.a100 32 Cores 256 GB 4TB NVMe 1x A100 80GB 25Gbps
ai.cluster.h100 64 Cores 512 GB 8TB NVMe 2x H100 80GB 100Gbps IB
sso.standard 4 Cores 16 GB 200GB NVMe - 5Gbps
sso.enterprise 16 Cores 64 GB 1TB NVMe - 10Gbps

Simple, Transparent Pricing

No egress fees. No API call charges. Just predictable infrastructure costs.

AI Inference

For model serving & APIs

R2,499 /mo
  • 8 vCPU Cores
  • 32GB DDR4
  • 500GB NVMe SSD
  • 10Gbps Uplink
  • 5TB Transfer
  • CUDA Pre-installed
Configure
MOST POPULAR

AI Training

A100 GPU powered

R12,999 /mo
  • 32 vCPU Cores
  • 256GB DDR4
  • 4TB NVMe SSD
  • 1x NVIDIA A100 80GB
  • 25Gbps Uplink
  • Unmetered Transfer
  • NVLink Ready
Deploy Now

SSO Enterprise

Identity infrastructure

R3,899 /mo
  • 16 vCPU Cores
  • 64GB DDR4 ECC
  • 1TB NVMe SSD
  • 10Gbps Uplink
  • High Availability
  • Keycloak Pre-config
Configure
All prices in ZAR (South African Rand) and exclude VAT. Custom configurations available.
South Africa

Ready to Deploy?

Join South Africa's leading AI and SSO infrastructure platform. Deploy in under 90 seconds.

Get Started Contact Us
90s Deployment
DDoS Protected
24/7 Support
SA Data Centers