Products
⚡

Dedicated Servers

From $79/mo · 400+ configs
🧠

GPU Metal

From $1,699/mo · H100 / A100 / L40S
⛓

Blockchain Nodes

From $149/mo · 15+ chains
📺

Streaming & Media

From $99/mo · 10Gbps+
🚀

Instant Stock

Deploy in 60 seconds
Quick Links
💰

GPU Inventory
🌍

Benchmarks
⭐

vs. Cloud
Deploy Now — From $79/mo →
Live GPU inventory — deploying now

GPU Metal

Full NVIDIA H100, H200, L40S & A100 servers on bare metal. No hypervisor. No shared tenants. No egress fees. Deployed in minutes.

AD102 NVIDIA L40S
NVIDIA L40S
48 GB GDDR6
18,176 CUDA cores
91.6 FP32 TFLOPS
864 GB/s bandwidth
GH100 814mm² · TSMC 4N HBM3 HBM3 NVLink 4.0 NVLink 4.0 NVIDIA H100 SXM5
NVIDIA H100 SXM5
80 GB HBM3
16,896 CUDA cores · 528 Tensor cores
3,350 GB/s memory bandwidth
NVLink 4.0 · 900 GB/s interconnect
700W TDP · TSMC 4N
REDSWITCHES
2U Rack Server
Dual-socket EPYC
128–512 GB DDR5 ECC
NVMe SSD enterprise storage
IPMI · Full root access
1–25 Gbps uplink
8
GPU Models
7
Countries
4x
Max Multi-GPU
0
Egress Fees
99.99%
Uptime SLA
// gpu inventory
Choose Your GPU. Deploy in Minutes.

Grouped by GPU model. Real specs, real prices. All plans include free setup, DDoS protection, and full root access.

NVIDIA H200
Frontier AI 141 GB HBM3e · 4,800 GB/s · 528 Tensor cores
GPUCPUVRAMRAMStorageNetworkLocationPrice
1x H200 141 GB2x EPYC 9334
64C/128T		141 GB
HBM3e		128 GB2x 960GB SSD1G / 100TB🇨🇦 Montreal€2,485/mo
1x H200 141 GB2x EPYC 9334
64C/128T		141 GB
HBM3e		128 GB2x 960GB SSD1G / 30TB🇬🇧 London€2,556/mo
2x H200 282 GB2x EPYC 9334
64C/128T		282 GB
HBM3e		128 GB2x 960GB SSD1G / 100TB🇨🇦 Montreal€4,299/mo
NVIDIA H100
Enterprise AI 80 GB HBM3 · 3,350 GB/s · 528 Tensor cores
GPUCPUVRAMRAMStorageNetworkLocationPrice
1x H100 80 GB2x EPYC 7413
48C/96T		80 GB
HBM3		128 GB2x 960GB SSD1G / 100TB🇨🇦 Montreal€1,873/mo
1x H100 80 GB2x EPYC 7543
64C/128T		80 GB
HBM3		128 GB2x 960GB SSD1G / 100TB🇨🇦 Montreal€1,947/mo
1x H100 80 GB2x EPYC 7413
48C/96T		80 GB
HBM3		128 GB2x 960GB SSD1G / 100TB🇩🇪 Frankfurt€2,133/mo
1x H100 80 GB2x EPYC 7543
64C/128T		80 GB
HBM3		128 GB2x 960GB SSD1G / 30TB🇸🇬 Singapore€2,518/mo
2x H100 160 GB2x EPYC 9334
64C/128T		160 GB
HBM3		128 GB2x 960GB SSD1G / 30TB🇬🇧 London€4,027/mo
2x H100 160 GB2x EPYC 9334
64C/128T		160 GB
HBM3		128 GB2x 960GB SSD1G / 30TB🇸🇬 Singapore€4,304/mo
4x H100 320 GB2x EPYC 9334
64C/128T		320 GB
HBM3		128 GB2x 960GB SSD1G / 30TB🇬🇧 London€7,300/mo
NVIDIA L40S
Production · Inference 48 GB GDDR6 · 864 GB/s · 91.6 TFLOPS
GPUCPUVRAMRAMStorageNetworkLocationPrice
1x L40S 48 GB2x EPYC 7413
48C/96T		48 GB
GDDR6		128 GB2x 960GB SSD1G / 100TB🇨🇦 Montreal€729/mo⚡ Instant
1x L40S 48 GB2x Xeon 4314
32C/64T		48 GB
GDDR6		128 GB2x 960GB NVMe1G / 100TB🇩🇪 Frankfurt€915/mo
1x L40S 48 GB2x EPYC 7413
48C/96T		48 GB
GDDR6		128 GB2x 960GB SSD1G / 30TB🇸🇬 Singapore€1,348/mo
2x L40S 96 GB2x EPYC 9334
64C/128T		96 GB
GDDR6		128 GB2x 960GB SSD1G / 100TB🇨🇦 Montreal€1,527/mo
3x L40S 144 GB2x EPYC 9224
48C/96T		144 GB
GDDR6		128 GB2x 960GB SSD1G / 100TB🇩🇪 Frankfurt€2,512/mo
NVIDIA A100
Training · Analytics 40/80 GB HBM2e · 1,555 GB/s · 19.5 TFLOPS
GPUCPUVRAMRAMStorageNetworkLocationPrice
1x A100 40 GB2x Xeon 4214
24C/48T		40 GB
HBM2e		128 GB2x 480GB SSD1G / 100TB🇩🇪 Frankfurt€621/mo
1x A100 40 GB2x Xeon 5218
32C/64T		40 GB
HBM2e		128 GB2x 480GB SSD1G / 100TB🇩🇪 Frankfurt€848/mo
2x A100 80 GB2x Xeon 4314
32C/64T		80 GB
HBM2e		128 GB4x 960GB NVMe1G / 100TB🇩🇪 Frankfurt€1,417/mo
L4 · T4 · A30 · MI210
Inference · Entry 16–64 GB VRAM · from €278/mo
GPUCPUVRAMRAMStorageNetworkLocationPrice
1x T4 16 GB2x Xeon E5-2650v4
24C/48T		16 GB
GDDR6		128 GB2x 480GB SSD1G / 100TB🇩🇪 Frankfurt€278/mo⚡ Instant
1x T4 16 GB2x Xeon E5-2630v4
20C/40T		16 GB
GDDR6		128 GB2x 960GB SSD1G / 100TB🇳🇱 Amsterdam€324/mo⚡ Instant
1x L4 24 GB2x EPYC 7413
48C/96T		24 GB
GDDR6		128 GB2x 960GB SSD1G / 100TB🇨🇦 Montreal€423/mo
1x A30 24 GB2x EPYC 7402
48C/96T		24 GB
HBM2		128 GB2x 960GB SSD1G / 100TB🇩🇪 Frankfurt€538/mo⚡ Instant
1x L4 24 GB2x EPYC 7413
48C/96T		24 GB
GDDR6		128 GB2x 960GB SSD1G / 100TB🇩🇪 Frankfurt€650/mo
1x MI210 64 GB2x Xeon 5218
32C/64T		64 GB
HBM2e		128 GB2x 960GB SSD1G / 100TB🇳🇱 Amsterdam€833/mo

All plans include: Free setup · DDoS protection · KVM/IPMI · Full root access · 24/7 support

View All 400+ Configurations →
// performance data
GPU Benchmark Comparison

Measured specs across every GPU in our fleet. Compare VRAM, memory bandwidth, and compute power side by side.

GPUVRAMMem BWRelative BWFP32 TFLOPSTDPBest For
● H200141 GB HBM3e4,800 GB/s
51700WFrontier AI, LLM pre-train
● H10080 GB HBM33,350 GB/s
~60700WLLM training, HPC
● MI21064 GB HBM2e1,600 GB/s
22.7300WFP64 HPC, AI training
● A10040/80 GB HBM2e1,555 GB/s
19.5250WTraining & analytics
● A3024 GB HBM2933 GB/s
10.3165WAI training, mixed
● L40S48 GB GDDR6864 GB/s
91.6350WInference & graphics
● L424 GB GDDR6300 GB/s
30.372WVideo AI, inference
● T416 GB GDDR6300 GB/s
8.170WEdge inference, transcode
// bare metal vs. cloud
Same GPU. Half the Price. Zero Overhead.

No hypervisor tax. No egress fees. No noisy neighbours. Predictable monthly billing.

AWS p4d.24xlarge

8x A100 · On-demand
Monthly cost~$23,500
Egress fees$0.09/GB
VirtualisationKVM overhead
Dedicated HWShared infra
Root accessLimited
ContractOn-demand / RI

RedSwitches GPU Metal

H100 bare metal
Monthly costfrom €2,133
Egress fees$0 — zero
VirtualisationNone — bare metal
Dedicated HW100% yours
Root accessFull root + IPMI
ContractMonthly, no lock-in

Lambda Cloud

1x H100 SXM
Monthly cost~$2,160
Egress fees$0
VirtualisationKVM
Dedicated HWShared
Root accessSSH only
ContractHourly / reserved
// why bare metal gpu
This. Machine. Is. Yours.

Not a slice. Not a VM. The entire physical server — from GPU to NIC — is dedicated to your workload.

Zero Virtualisation Overhead

No hypervisor between your code and the GPU. Every CUDA core, every byte of HBM, every PCIe lane is yours. 3–8% measured performance uplift vs. equivalent cloud VMs.

No Egress Fees, Ever

Transfer training data in, pull model weights out. No per-GB charges. Unmetered bandwidth included. Flat monthly pricing, no surprises.

Deploy in Under 10 Minutes

Instant servers ship pre-configured and ready to SSH. Pick your GPU, pick your OS, start training within minutes.

20+ Data Centres, 7 Countries

GPU Metal is available in Frankfurt, Amsterdam, London, Montreal, Los Angeles, Singapore, and Mumbai. More coming Q2 2026.

Full Root + IPMI Access

Custom kernels, proprietary drivers, GPU clock tuning. KVM console for out-of-band management. Your server, your rules.

Pay in Crypto

BTC, ETH, USDC, SOL accepted via Cryptomus. No friction for Web3 teams. Monthly billing in EUR, USD, or INR.

// software stack
✓ PyTorch 2.x
✓ TensorFlow
✓ JAX
✓ CUDA 12.x
✓ cuDNN
✓ Docker
✓ vLLM
✓ Triton Inference
✓ DeepSpeed
✓ Hugging Face
99.99%
Uptime SLA
10 min
Deploy Time
$0
Egress Fees
$0
Setup Fees
24/7
Human Support
20+
Data Centres

Ready to Deploy?

Browse our live GPU inventory and have your server online in minutes. No sales call required.