Request access
Reviewed 2026-10-10

GPU catalog6 GPUs2 racks

NVIDIA GPUs, by the hour or by the rack.

Six GPUs and two NVL72 racks on one price list. On-demand from $1.86 per GPU-hour, spot from $0.87, reserved by quote.

Request capacity Pricing and terms

Preview list prices. They apply when your account opens. Nothing is billed before that.

§01 Catalog Reviewed 2026-10-10

Every listing has a price or a status.

Node CPU, RAM, local storage and fabric vary by pool and are reported at placement.

TAB 01Preview list prices Source · Fantasti price list Reviewed 2026-10-10
Preview list prices. USD per GPU-hour, preview list price, reviewed 2026-10-10.
GPU Memory Interconnect Status On-demand Reserved Spot Action
B300  HGX 8-GPU 270 GB HBM3E NVLink 5 Private Preview $10.45 /GPU·hr Request a quote From $1.09 Spot price moves with supply and demand. Follow it, or set a max price per GPU-hour: you pay the spot price, and nodes stop if it rises above your max.How spot works Request
B200  HGX 8-GPU 180 GB HBM3E NVLink 5 Private Preview $9.35 /GPU·hr Request a quote From $1.09 Spot price moves with supply and demand. Follow it, or set a max price per GPU-hour: you pay the spot price, and nodes stop if it rises above your max.How spot works Request
H200 SXM  HGX 8-GPU 141 GB HBM3E NVLink 4 Private Preview $5.94 /GPU·hr Request a quote From $0.87 Spot price moves with supply and demand. Follow it, or set a max price per GPU-hour: you pay the spot price, and nodes stop if it rises above your max.How spot works Request
H100 SXM5  HGX 8-GPU 80 GB HBM3 NVLink 4 Private Preview $5.40 /GPU·hr Request a quote From $0.87 Spot price moves with supply and demand. Follow it, or set a max price per GPU-hour: you pay the spot price, and nodes stop if it rises above your max.How spot works Request
RTX PRO 6000  PCIe · Server Edition 96 GB GDDR7 PCIe Gen5 Private Preview $2.16 /GPU·hr Request a quote From $0.87 Spot price moves with supply and demand. Follow it, or set a max price per GPU-hour: you pay the spot price, and nodes stop if it rises above your max.How spot works Request
L40S  PCIe 48 GB GDDR6 PCIe Gen4 Private Preview $1.86 /GPU·hr Request a quote Request
GB300 NVL72  Rack-scale · 72 GPUs 279 GB HBM3E NVLink 5 · 72 By request Reserved · contact sales Request a quote Request
GB200 NVL72  Rack-scale · 72 GPUs 186 GB HBM3E NVLink 5 · 72 By request Reserved · contact sales Request a quote Request

USD per GPU-hour · preview list price · applies when your account opens · reviewed 2026-10-10

"From" is the lowest spot price for that GPU. The spot price moves with supply and demand. The spot price always stays below the on-demand rate for the same GPU.

Status

Private Preview

Opens with the first cohort. The price is the preview list price.

B300 · B200 · H200 SXM · H100 SXM5 · RTX PRO 6000 · L40S

By request

Sold on reserved terms through sales. The price is confirmed in a quote.

GB300 NVL72 · GB200 NVL72

Release stages

  • Specifications are NVIDIA reference figures.
  • L40S: list price for the base shape: 1 GPU, 8 vCPU, 32 GiB. Larger shapes are quoted.
  • Reserved rates are quoted per term. Rack-scale systems are reserved capacity. Talk to us about term, delivery and configuration.
§02 Specifications NVIDIA reference figures
TAB 02Specifications per GPU Source · NVIDIA Reviewed 2026-10-10
Specifications per GPU
ListingMemoryBandwidthBetween GPUsFP4, dense / sparseFP8, dense / sparseMax power
Instances and clusters
B300270 GB HBM3E7.7 TB/sNVLink 5 · 8 GPUs per HGX board14 / 18 PFLOPS4.5 / 9 PFLOPS1,100 W
B200180 GB HBM3E7.7 TB/sNVLink 5 · 8 GPUs per HGX board9 / 18 PFLOPS4.5 / 9 PFLOPS1,000 W
H200 SXM141 GB HBM3E4.8 TB/sNVLink 4 · 8 GPUs per HGX boardNot supported1,979 / 3,958 TFLOPS700 W
H100 SXM580 GB HBM33.35 TB/sNVLink 4 · 8 GPUs per HGX boardNot supported1,979 / 3,958 TFLOPS700 W
RTX PRO 600096 GB GDDR7 with ECC1.597 TB/sNo NVLink · PCIe 5.0 x164 PFLOPS (peak, sparse)2 PFLOPS (peak, sparse)600 W
L40S48 GB GDDR6 with ECC864 GB/sNo NVLink · PCIe Gen4 x16Not supported733 / 1,466 TFLOPS350 W
Rack-scale · per GPU
GB300 NVL72279 GB HBM3E8 TB/s72 GPUs in one NVLink domain15 / 20 PFLOPS5 / 10 PFLOPS1,400 W
GB200 NVL72186 GB HBM3E8 TB/s72 GPUs in one NVLink domain10 / 20 PFLOPS5 / 10 PFLOPS1,200 W
Specifications are NVIDIA reference figures. RTX PRO 6000 compute is a sparse peak; NVIDIA publishes no dense figure.
§03 Fit 8 listings

Which GPU for which work.

What each listing is good at, where another one fits better, and whether you can run it on demand, on spot or reserved.

TAB 03Workload fit and capacity Source · Fantasti catalog Reviewed 2026-10-10
Workload fit and capacity: what each listing is best for, what it runs as, and whether it is offered on demand, on spot and reserved.
Listing Best for Look elsewhere for Runs as On-demandSpotReserved
Instances and clusters
B300
  • Reasoning and test-time-scaling inference
  • Large-model training
  • Video generation
  • FP64 HPC
  • INT8 workloads
 Instances and GPU Clusters On-demand: yes Spot: yes Reserved: yes
B200
  • Large-model training
  • High-throughput inference
  • HPC
 Nothing listed Instances and GPU Clusters On-demand: yes Spot: yes Reserved: yes
H200 SXM
  • Memory-bound LLM inference
  • Training and fine-tuning
  • HPC
 Nothing listed Instances and GPU Clusters On-demand: yes Spot: yes Reserved: yes
H100 SXM5
  • Training and fine-tuning
  • Inference
  • HPC (FP64 34 TFLOPS, FP64 Tensor 67 TFLOPS)
 Nothing listed Instances and GPU Clusters On-demand: yes Spot: yes Reserved: yes
RTX PRO 6000
  • Inference and agentic AI
  • Physical AI and simulation
  • Rendering and video (4 NVENC, 4 NVDEC)
  • Multi-node training that needs InfiniBand
 Single instances On-demand: yes Spot: yes Reserved: yes
L40S
  • Inference
  • Fine-tuning smaller models
  • 3D graphics, rendering and video (3 NVENC, 3 NVDEC, AV1)
  • Multi-node training that needs InfiniBand
 Single instances On-demand: yes Spot: no Reserved: yes
Rack-scale
GB300 NVL72
  • Reasoning and test-time-scaling inference
  • Large-scale training
  • Video generation
  • FP64 HPC
 Reserved racks On-demand: no Spot: no Reserved: yes
GB200 NVL72
  • Real-time trillion-parameter inference
  • Large-scale training
  • Data processing
 Nothing listed Reserved racks On-demand: no Spot: no Reserved: yes
A filled mark means the listing is offered that way. Reserved capacity never runs as spot.
Sheet
01 / 01
Title
GPU catalog
Unit
USD per GPU-hour
Reviewed
2026-10-10
FILM 03Heatsink fins Illustrative
§04 Datasheets 8 drawings

Every listing has a datasheet.

A shop drawing tied to its specification rows, the node shapes, the price and the terms. Each one prints on paper.

§05 Node fabric Platform configuration

Node fabric, stated in Tb/s.

An 8-GPU B200, H200 or H100 node has eight 400 Gb/s InfiniBand ports: 3.2 Tb/s of scale-out bandwidth, which is 400 GB/s. The fabric depends on the pool and is listed in every placement record.

  • B200, H200 and H100400 Gb/s per GPU, 3.2 Tb/s per 8-GPU node
  • B300800 Gb/s per GPU
  • RTX PRO 6000 and L40Sno InfiniBand; they run as single instances

GPU Clusters

8-GPU node, fabric ports Not to scale
DWG 018-GPU node, 8 × 400 Gb/s InfiniBand Source · Fantasti platform configuration
8-GPU node with eight 400 Gb/s InfiniBand ports, dimetric drawing
  1. 1. GPU under its heatsink · 8 per node, on one baseboard
  2. 2. InfiniBand adapter · 8 per node, single port
  3. 3. 400 Gb/s port · 8 × 400 Gb/s = 3.2 Tb/s per node, where the pool provides it
  4. 4. Chassis · cover and near walls removed
§06 Request Private Preview

Tell us the GPU, the count and the term.

Reserved rates are quoted per term. Rack-scale systems are reserved capacity. Talk to us about term, delivery and configuration.

Request capacity Pricing and terms

Next