Pricing
GPU server pricing
One flat monthly price per server, with the per-GPU-hour equivalent shown for comparison. No setup fee, no egress fees within the included bandwidth, and you can cancel any month.
- Configurations
- 15
- Available now
- 13
- From
- $449/mo
- Setup & egress fees
- $0
All GPU server prices
Every server includes
- Root access
- Local NVMe
- IPv4 + IPv6
- DDoS protection
- 20–50 TB, no egress fees
- 24/7 engineers
8-GPU servers
-
8× HGX B300 New
NVIDIA Blackwell Ultra · Bare-metal server8× 288 GB HBM3e 8 TB/s per GPUCPU: 2× Xeon 6776P · 128 coresMemory: 4 TB DDR5
Storage: 8× 3.84 TB NVMeNetwork: 8× 800G ConnectX-8
Up to 671Bin FP8 · same in BF16Limited stock$39,900/mo
≈ $6.83/GPU-hr
-
8× HGX B200
NVIDIA Blackwell · Bare-metal server8× 180 GB HBM3e 8 TB/s per GPUCPU: 2× Xeon 8570 · 112 coresMemory: 2 TB DDR5
Storage: 8× 3.84 TB NVMeNetwork: 8× 400G ConnectX-7
Up to 671Bin FP8 · 405B in BF16In stock$34,900/mo
≈ $5.98/GPU-hr
-
8× HGX H200
NVIDIA Hopper · Bare-metal server8× 141 GB HBM3e 4.8 TB/s per GPUCPU: 2× Xeon 8480+ · 112 coresMemory: 2 TB DDR5
Storage: 8× 3.84 TB NVMeNetwork: 8× 400G ConnectX-7
Up to 671Bin FP8 · 405B in BF16In stock$23,900/mo
≈ $4.09/GPU-hr
-
8× HGX H100
NVIDIA Hopper · Bare-metal server8× 80 GB HBM3 3.35 TB/s per GPUCPU: 2× Xeon 8480+ · 112 coresMemory: 2 TB DDR5
Storage: 8× 3.84 TB NVMeNetwork: 8× 400G ConnectX-7
Up to 405Bin FP8 · 235B in BF16In stock$16,900/mo
≈ $2.89/GPU-hr
-
8× Instinct MI355X
AMD CDNA 4 · Bare-metal server8× 288 GB HBM3E 8 TB/s per GPUCPU: 2× EPYC 9575F · 128 coresMemory: 3 TB DDR5
Storage: 8× 3.84 TB NVMeNetwork: 8× 400G RoCE
Up to 671Bin FP8 · same in BF16Limited stock$19,900/mo
≈ $3.41/GPU-hr
-
8× RTX PRO 6000 Blackwell
NVIDIA Blackwell · Bare-metal server8× 96 GB GDDR7 1.6 TB/s per GPUCPU: 2× EPYC 9555 · 128 coresMemory: 1.5 TB DDR5
Storage: 4× 3.84 TB NVMeNetwork: 2× 100 GbE
Up to 405Bin FP8 · 235B in BF16In stock$10,490/mo
≈ $1.80/GPU-hr
Single GPU
-
B300 New
NVIDIA Blackwell Ultra · Dedicated VM · full GPU288 GB HBM3e 8 TB/s per GPUCPU: 28 vCPUMemory: 384 GB
Storage: 3.5 TB NVMeNetwork: 25 Gbps
Up to 120Bin FP8 · 70B in BF16Limited stock$5,190/mo
≈ $7.11/GPU-hr
-
B200
NVIDIA Blackwell · Dedicated VM · full GPU180 GB HBM3e 8 TB/s per GPUCPU: 26 vCPUMemory: 256 GB
Storage: 2.75 TB NVMeNetwork: 25 Gbps
Up to 120Bin FP8 · 70B in BF16In stock$4,590/mo
≈ $6.29/GPU-hr
-
H200 Popular
NVIDIA Hopper · Dedicated VM · full GPU141 GB HBM3e 4.8 TB/s per GPUCPU: 24 vCPUMemory: 240 GB
Storage: 2 TB NVMeNetwork: 25 Gbps
Up to 70Bin FP8 · 32B in BF16In stock$3,090/mo
≈ $4.23/GPU-hr
-
H100 SXM
NVIDIA Hopper · Dedicated VM · full GPU80 GB HBM3 3.35 TB/s per GPUCPU: 24 vCPUMemory: 225 GB
Storage: 2 TB NVMeNetwork: 25 Gbps
Up to 32Bin FP8 · same in BF16In stock$2,190/mo
≈ $3.00/GPU-hr
-
RTX PRO 6000 Blackwell
NVIDIA Blackwell · Dedicated VM · full GPU96 GB GDDR7 1.6 TB/s per GPUCPU: 16 vCPUMemory: 188 GB
Storage: 1.92 TB NVMeNetwork: 10 Gbps
Up to 70Bin FP8 · 32B in BF16In stock$1,390/mo
≈ $1.90/GPU-hr
-
L40S
NVIDIA Ada Lovelace · Dedicated VM · full GPU48 GB GDDR6 864 GB/s per GPUCPU: 16 vCPUMemory: 128 GB
Storage: 1 TB NVMeNetwork: 10 Gbps
Up to 32Bin FP8 · 14B in BF16In stock$890/mo
≈ $1.22/GPU-hr
-
GeForce RTX 5090
NVIDIA Blackwell · Dedicated VM · full GPU32 GB GDDR7 1.8 TB/s per GPUCPU: 16 vCPUMemory: 64 GB
Storage: 1 TB NVMeNetwork: 10 Gbps
Up to 14Bin FP8 · 8B in BF16In stock$449/mo
≈ $0.62/GPU-hr
Rack-scale systems
-
Vera Rubin NVL72
NVIDIA Rubin · Rack-scale system72× 288 GB HBM4 19.2 TB/s per GPUCPU: 36× NVIDIA VeraMemory: Up to 54 TB LPDDR5X
Storage: Storage sized per orderNetwork: NVLink 6 · ConnectX-9
Frontier scale72-GPU NVLink domainReservations openCustom quote
Next-gen allocation
-
GB300 NVL72
NVIDIA Blackwell Ultra · Rack-scale system72× 288 GB HBM3e 8 TB/s per GPUCPU: 36× NVIDIA GraceMemory: 17 TB LPDDR5X
Storage: NVMe per tray, sized per orderNetwork: NVLink 5 · 72× 800G
Frontier scale72-GPU NVLink domainOn request$7,990/GPU·mo
From · 72 GPUs · 12–36 mo
Several servers, a longer term or a multi-node cluster? Ask for volume pricing — usually from three servers.
Our GPU server pricing is one flat price per server per month. The table above lists all 15 configurations, from a single GeForce RTX 5090 at $449 a month to an 8-GPU NVIDIA HGX B300 server at $39,900, with rack-scale systems quoted on request. Below we explain how each GPU rental price per month is set, what it includes, how it compares with hourly billing, and how terms, taxes and payments work.
How our GPU server pricing works
You pay for the whole server, monthly and in advance. The price of an 8-GPU server covers all eight GPUs plus the CPUs, RAM, NVMe storage and network cards listed for it. The minimum term is one month. Plans renew automatically, with an email reminder before each renewal, and you can cancel any month with no fee; the server keeps running until the end of the paid month.
To make comparison with hourly offers easier, we also show a per-GPU-hour equivalent: the monthly price divided by the number of GPUs, then by 730, the average number of hours in a month. For example:
- NVIDIA HGX H100, 8 GPUs $16,900 ÷ 8 ÷ 730 ≈ $2.89 per GPU-hour
- NVIDIA H200, 1 GPU $3,090 ÷ 730 ≈ $4.23 per GPU-hour
- GeForce RTX 5090, 1 GPU $449 ÷ 730 ≈ $0.62 per GPU-hour
That figure assumes the server runs around the clock. It is a comparison tool, not how you are billed: you pay the flat monthly price whether the GPUs run for 730 hours or 30.
What the monthly price includes
Every price covers the full server as listed, plus:
- bandwidth with no egress fees: 20 TB per month on single-GPU plans and 50 TB on 8-GPU servers, under fair use, with inbound traffic not counted;
- a public IPv4 address and an IPv6 /64, with DDoS mitigation on public IPs;
- root access over SSH, with OS images that come with drivers, CUDA or ROCm, and Docker preinstalled;
- the 99.9% availability SLA with service credits, and 24/7 support from engineers.
There is no setup fee, and because the price is flat, an idle server costs nothing extra. Two things are not included: VAT and other taxes where applicable, and backups, since local storage is not backed up unless agreed in writing. If you consistently exceed your bandwidth allowance, we contact you before any change to agree on a suitable option; our bandwidth and egress guide explains the allowance.
Monthly vs hourly GPU rental prices
Hourly billing is the better deal when you need a GPU for short, irregular bursts. A monthly price wins when the GPU stays busy. The break-even point is easy to find: divide our monthly price by the hourly rate you would otherwise pay. The result is the number of hours per month at which both cost the same; if you would use the GPU for more hours than that, the monthly plan comes out ahead.
Hourly prices also move. FullGPU's October 2026 survey of public price lists found that over the past year, the median on-demand price per GPU-hour rose by roughly 13% for H100, 27% for H200 and 20% for B200. Hourly bills can carry extras too: hyperscalers typically bill egress separately and require quota approvals for GPU instances. Model your own case with the GPU cost calculator, or read our full monthly vs hourly comparison.
Volume and term pricing
If you plan to rent for 6 to 12 months, or need several servers, usually from three, ask us for volume and term pricing; the terms are agreed in writing before you pay. Rack-scale systems are priced separately. NVIDIA GB300 NVL72 starts from $7,990 per GPU per month on 12- to 36-month terms, and Vera Rubin NVL72 is quoted individually, with reservations open. Tell us what you need through our contact form, or read about our rack-scale systems.
Taxes, invoices and payment methods
Prices are in US dollars and exclude VAT and other taxes, which are added where applicable. Every payment comes with an invoice; our guide to invoices and VAT covers the details.
You can pay by Visa, Mastercard or American Express, or in crypto: Bitcoin (on-chain or Lightning), Ethereum, USDT on TRON or Ethereum, USDC on Ethereum, Base or Solana, and Litecoin. The price and the invoice are the same whichever method you choose, and for crypto the exchange rate is locked at checkout. See how it works on our crypto payments page.
Price changes
Your price is fixed for each paid month. We may change prices for future months, but only with at least 30 days' notice, and you can cancel before the change takes effect. Our guide to renewals and cancellation explains how renewals work.
Compare with an hourly quote
Enter any hourly price and your usage to see the break-even and the monthly difference.
Open the cost calculator Ask for volume pricingHow much does it cost to rent a GPU server per month?
Our prices range from $449 a month for a single GeForce RTX 5090 to $39,900 a month for an 8-GPU NVIDIA HGX B300 server. Rack-scale systems are quoted on request.
How is the per-GPU-hour price calculated?
It is the monthly price divided by the number of GPUs and by 730, the average number of hours in a month. It is a comparison figure only: you are billed the flat monthly price.
Are there setup fees or egress fees?
There is no setup fee, and no egress fees within the included bandwidth of 20 TB per month on single-GPU plans and 50 TB on 8-GPU servers, under fair use. Inbound traffic is not counted.
Do prices include VAT?
No. Prices are in US dollars and exclude VAT and other taxes, which are added where applicable.
Can I get better pricing for a longer term or several servers?
Yes. Volume and term pricing for 6 to 12 months or for several servers, usually from three, is available on request.
Can my price change during a paid month?
No. Your price is fixed for each paid month, and we give at least 30 days' notice before any change for future months, so you can cancel before it applies.
Talk to an engineer
Get your GPU server this month.
Tell us what you need. A GPU engineer replies with availability, a confirmed delivery date and a payment link.
- Delivery date confirmed before you pay
- Cancel any month
- Card or crypto, same price