Funding of Over $1 Billion. Read More readmore

logo
Contact Sales
Request a Demo

Search Neysa


Rent H100 GPU On-Demand

Unleash extreme performance for training, inference, and HPC workloads

On-demand access. Built for scale. 
Powered by NVIDIA H100 Tensor Core GPUs.

Price Starts At: $0.52/h (Fractional – 10GB)
Reserve Now
rent-h100-top-banners

Why Choose Neysa?

No Lock-in Period

Flexible commitment options 
without long-term contracts

Hourly + Monthly Options

Choose the billing cycle that 
works best for your needs

Scalable on Demand

Easily scale your GPU resources 
up or down as needed

Full-stack Support

Comprehensive technical 
support and guidance

No Lock-in Period

Flexible commitment options 
without long-term contracts

Hourly + Monthly Options

Choose the billing cycle that 
works best for your needs

Scalable on Demand

Easily scale your GPU resources 
up or down as needed

Full-stack Support

Comprehensive technical 
support and guidance

Choose Your 
H100 Configuration

Neysa offers three configurations to match your workload and budget

Pricing Per GPU
GPUs
Hourly
on-demand(per hour)
Monthly
Commitment(Per Hour)
6-month
commitment(per month)
1-year
commitment(per month)
2-year
commitment(per month)
3-year
commitment(per month)
H100 Fractional – 10GB
1g.10GB, 6vCPU, 42GB RAM, 100GB NVME
Hourly
on-demand(per hour)
$0.79
Monthly
Commitment(Per Hour)
$0.52
Save 34%
6-month
commitment(per month)
$0.49
Save 38%
1-year
commitment(per month)
$0.46
Save 42%
2-year
commitment(per month)
$0.40
Save 49%
3-year
commitment(per month)
$0.36
Save 54%
H100 Fractional – 40GB Popular
1g.10GB, 6vCPU, 42GB RAM, 100GB NVME
Hourly
on-demand(per hour)
$2.75
Monthly
Commitment(Per Hour)
$1.45
Save 47%
6-month
commitment(per month)
$1.32
Save 52%
1-year
commitment(per month)
$1.23
Save 55%
2-year
commitment(per month)
$1.14
Save 58%
3-year
commitment(per month)
$1.05
Save 62%
H100 SXM – 80GB
8g.80GB, 48vCPU, 288GB RAM, 1000GB NVME
Hourly
on-demand(per hour)
$4.94
Monthly
Commitment(Per Hour)
$3.25
Save 34%
6-month
commitment(per month)
$3.07
Save 38%
1-year
commitment(per month)
$2.66
Save 46%
2-year
commitment(per month)
$2.58
Save 48%
3-year
commitment(per month)
$2.35
Save 52%
Get in touch with us

Technical Specifications

Category Specification
ArchitectureNVIDIA Hopper
GPU Memory80 GB HBM3 (also available in 40GB & 10GB fractional at Neysa)
Memory Bandwidth2.0 TB/s
Peak FP8Up to 4,000 TFLOPs (Tensor Core)
Peak FP16/BF16Up to 2,000 TFLOPs
Peak TF32Up to 1,000 TFLOPs
CUDA Cores16,896 (per full 80GB H100)
Tensor Cores528 4th Gen Tensor Cores
GPU InterconnectNVLink (SXM5) – 600 GB/s, PCIe Gen5 (PCIe variant) – 128 GB/s
Form FactorSXM5 or PCIe Gen5
Max PowerSXM: 700W / PCIe: 350W
CoolingLiquid Cooling (SXM) / Active/Passive (PCIe)
Multi-Instance GPU (MIG)Up to 7 MIG instances (depending on memory config)
ECC MemorySupported (for error-free HPC workloads)
Software SupportCUDA 12+, cuDNN, TensorRT, PyTorch, TensorFlow, RAPIDS, Triton Inference Server

GPUaaS Cloud Infrastructure

Rich cloud-native & bare-metal offerings

Velocis AI Cloud: A modern, AI-native Cloud designed for AI Workloads. Deploy, Manage and Monitor your GPU-enabled AI Cloud as VMs, Containers, or Bare Metal with Rich Tooling and Insights.

Velocis Orchestration Layer

Create, Manage, and Monitor your AI Cloud environment immediately and 
effortlessly using the Velocis orchestration engine.

Virtual Machines

Deploy GPU VMs as on-demand or reserved Instances on the Velocis.

Bare Metal

Consume GPUs as private clusters for large training and fine-tuning requirements.

Containers

Deploy and manage your K8S GPU clusters using the Velocis Kubernets Engine.

Use Cases for H100

AI Model Training (LLMs, GPT, Diffusion Models)

  • H100 is built to handle the most demanding AI workloads, making it perfect for training large language models like GPT and BERT.
  • Its massive memory and 4th-gen tensor cores significantly reduce training time for billion-parameter models.
  • Diffusion models used in generative AI also benefit from faster iteration cycles and lower compute costs.

High-Performance Inference

  • Deploy large models at scale with real-time responsiveness.
  • NVIDIA H100 delivers lightning-fast inference for applications like chatbots, recommendation engines, and intelligent search systems.
  • It ensures low latency even with massive model sizes, enabling seamless user experiences.

Generative AI

  • H100 GPU dramatically speeds up workflows for text-to-image, image-to-video, and audio generation models.
  • It enables creative professionals and AI labs to generate 
high-resolution outputs faster while supporting advanced frameworks like Stable Diffusion and StyleGAN.

AI Model Training (LLMs, GPT, Diffusion Models)

  • H100 is built to handle the most demanding AI workloads, making it perfect for training large language models like GPT and BERT.
  • Its massive memory and 4th-gen tensor cores significantly reduce training time for billion-parameter models.
  • Diffusion models used in generative AI also benefit from faster iteration cycles and lower compute costs.

High-Performance Inference

  • Deploy large models at scale with real-time responsiveness.
  • NVIDIA H100 delivers lightning-fast inference for applications like chatbots, recommendation engines, and intelligent search systems.
  • It ensures low latency even with massive model sizes, enabling seamless user experiences.

Generative AI

  • H100 GPU dramatically speeds up workflows for text-to-image, image-to-video, and audio generation models.
  • It enables creative professionals and AI labs to generate 
high-resolution outputs faster while supporting advanced frameworks like Stable Diffusion and StyleGAN.

Benefits of renting H100 at Neysa

Extreme Speed

Train LLMs 30x faster 
than A100.

Cost Optimized

Pay only for what 
you use.

Flexibility

On-demand or reserved, scale your way.

Latest Gen Support

FP8, FP16, INT8, 
BF16—all enabled.

HPC-Grade Reliability

Industrial, secure, 
and stable.

Extreme Speed

Train LLMs 30x faster 
than A100.

Cost Optimized

Pay only for what 
you use.

Flexibility

On-demand or reserved, scale your way.

Latest Gen Support

FP8, FP16, INT8, 
BF16—all enabled.

HPC-Grade Reliability

Industrial, secure, 
and stable.

Reserve H100

Real Impact of GPUaaS

Your AI journey deserves a trusted partner who understands both technology and business needs. Here’s how customers have unlocked the value of AI with Neysa Velocis:

Foundational Model Creation

India’s premier educational and research institute.

Challenge

Train a ground-up large language 
model cost-efficiently with access to 
the latest GPUs.

Solution

Velocis GPU Private Clusters

Access to the latest generation of 
Nvidia GPUs as scalable private clusters, with high-speed network and storage.

Impact

↑ 28% training performance.

↑ 22% cost efficiency.

» Accelerated language model 
development.

Our Range of GPUs

Pricing Per GPU
GPUs
Hourly
on-demand(per hour)
Monthly
Commitment(Per Hour)
6-month
commitment(per month)
1-year
commitment(per month)
2-year
commitment(per month)
3-year
commitment(per month)
H100 Fractional – 10GB
1g.10GB, 6vCPU, 42GB RAM, 100GB NVME
Hourly
on-demand(per hour)
$0.79
Monthly
Commitment(Per Hour)
$0.52
6-month
commitment(per month)
$0.49
1-year
commitment(per month)
$0.46
2-year
commitment(per month)
$0.40
3-year
commitment(per month)
$0.36
H100 Fractional – 40GB
1g.10GB, 6vCPU, 42GB RAM, 100GB NVME
Hourly
on-demand(per hour)
$2.75
Monthly
Commitment(Per Hour)
$1.45
6-month
commitment(per month)
$1.32
1-year
commitment(per month)
$1.23
2-year
commitment(per month)
$1.14
3-year
commitment(per month)
$1.05
H100 SXM – 80GB
8g.80GB, 48vCPU, 288GB RAM, 1000GB NVME
Hourly
on-demand(per hour)
$4.94
Monthly
Commitment(Per Hour)
$3.25
6-month
commitment(per month)
$3.07
1-year
commitment(per month)
$2.66
2-year
commitment(per month)
$2.58
3-year
commitment(per month)
$2.35

FAQs

Q: Can I use H100 on-demand or do I need to reserve?

A: You can access it hourly or reserve monthly, quarterly, or annually.

Q: What kind of support is included?

A: We offer prompt support, setup assistance, and necessary recommendations.

Q: Is H100 available with Jupyter or containers?

A: Yes, launch your workload via JupyterLab, Docker, or your favorite environment.

Q:Can I run multi-GPU clusters with H100s?

A: Yes, we support multi-GPU and multi-node orchestration with SLURM, Kubernetes, etc.

Q:How soon will my instance be ready?

A: Provisioning is instant in most cases. For reserved GPUs, setup takes < 30 mins.

Q:Can I rent an H100 GPU without long-term contracts?

A: Yes, Neysa offers complete flexibility and an affordable H100 price in India . You can rent H100 GPUs by the hour or commit monthly, quarterly, or annually—no long-term lock-ins unless you want deeper discounts.

Q:What does H100 on-demand mean at Neysa?

A:  H100 on-demand means instant, pay-as-you-go access to powerful NVIDIA H100 GPUs—no upfront commitment required. Spin up a GPU when you need it, shut it down when you don’t, and only pay for the time you use.

Request for Pricing