FrançaisFR
enEnglishfrFrançaisdeDeutschesEspañolptPortuguêsitItalianonlNederlandsplPolskitrTürkçeruРусскийukУкраїнськаja日本語ko한국어zh中文arالعربية
Connexion Déployer un GPU
GPU Rent HubGPUs8× H100 SXM

Rent an 8× H100 SXM node by the month, paid in crypto

GPU Rent Hub rents a complete NVIDIA HGX H100 node, 8 x NVIDIA H100 SXM 80GB on one single-tenant machine with NVSwitch · 900 GB/s between the cards, for 12 170 $US a month or 3 420 $US a week, in DFW-1 (Dallas), IAD-1 (Ashburn), PDX-1 (Hillsboro), paid in Bitcoin, Ethereum, USDT, USDC, Monero, Solana or Litecoin with no identity verification.

Updated 2026-09-01
12 170 $USPer month, flat
640 GBTotal GPU memory
2,11 $USPer GPU-hour
2 × 100GNetwork
01 The node

What you get in an 8× H100 SXM node

One physical machine with one tenant on it. No hypervisor between you and the cards, no other customer sharing the PCIe fabric, and root on the host. The interconnect is NVSwitch · 900 GB/s, which is what decides whether splitting a model across the cards is practical or painful.

Ordering 8 cards on one node instead of 8 separate instances saves 1 350 $US a month, about 10%. Prepaying three or six months takes a further 5% or 10% off.

Compare every multi-GPU node we rent

8x H100 SXM NVIDIA HGX H100
GPUs
8 x NVIDIA H100 SXM 80GB
Total VRAM
640 GB HBM3
Interconnect
NVSwitch · 900 GB/s
vCPU
160 dedicated
System RAM
2 TB
Local NVMe
16 TB encrypted
Network
2 × 100G
Traffic
Unmetered in, 20 TB out
Private VLAN
Included
Weekly
3 420 $US
Monthly
12 170 $US
Regions
DFW-1 (Dallas), IAD-1 (Ashburn), PDX-1 (Hillsboro)
02 Alternatives

Other nodes at a similar price

Per-GPU-hour is the monthly node price divided by 720 hours and then by the number of cards, so the figures are directly comparable across platforms.

NodeInterconnect Total VRAMMonthlyPer GPU-hour
8× MI300XAMD Instinct platformInfinity Fabric · 896 GB/s1536 GB11 450 $US/mo1,99 $US
8× H200 SXMNVIDIA HGX H200NVSwitch · 900 GB/s1128 GB16 490 $US/mo2,86 $US
8× A100 SXM 80GNVIDIA HGX A100NVSwitch · 600 GB/s640 GB7 130 $US/mo1,24 $US
4× H100 SXMNVIDIA HGX H100NVLink · 900 GB/s320 GB6 220 $US/mo2,16 $US
8× RTX 5090PCIe Gen5PCIe peer-to-peer256 GB3 890 $US/mo0,68 $US
03 Questions

Renting an 8× H100 SXM node

How much does an 8× H100 SXM node cost per month?
$12,170 a month at GPU Rent Hub, or $3,420 for a one-week term. That works out to $2.11 per GPU-hour across a 720-hour month. The price includes power, cooling, 160 vCPU, 2 TB of RAM, 16 TB of NVMe, 2 × 100G of network and 20 TB of outbound traffic.
Is an 8× H100 SXM node one machine or a cluster?
One physical machine with one tenant on it, and NVSwitch · 900 GB/s between the cards. It is not a virtual cluster stitched over a network, so tensor and pipeline parallelism behave the way the framework expects.
How much cheaper is the node than renting 8 separate cards?
Renting 8 single H100 SXM instances would be $13,520 a month. The node is $12,170, so it saves $1,350, about 10%. The multi-GPU discount also stacks with the 5% and 10% off for prepaying three or six months.
Can I pay for a node in crypto without KYC?
Yes, on the same terms as every other instance: a username and a password for the account, and Bitcoin, Ethereum, USDT, USDC, Monero, Solana or Litecoin for the balance. No identity document is requested or stored.
Where can I get an 8× H100 SXM node?
DFW-1 (Dallas), IAD-1 (Ashburn), PDX-1 (Hillsboro). Larger fabrics of 16 to 64 GPUs with InfiniBand HDR or NDR are quoted per case, with a lead time of two to six weeks.
What total memory does an 8× H100 SXM node have?
640 GB of GPU memory, which is 8 cards of 80 GB HBM3, plus 2 TB of system RAM. A 70B model in FP8 fits on one card; in BF16 it needs two.

Deploy the node, or ask for a bigger fabric

Nodes deploy from the dashboard in minutes. Clusters of 16 to 64 GPUs with InfiniBand are quoted per case.

Ask for a quote →