B300 in Production: What 18 Benchmark Runs Actually Showed
Predictable pricing that rewards a longer commitment.
Neysa Velocis delivers performance with cost predictability. We offer better unit economics and ~70% lesser TCO due to lean infrastructure structure, purpose-built for AI.
- AI Cloud
(Managed VM Instances) - Bare Metal GPUs
(Single-Tenant Servers) - Managed Kubernetes Clusters
(Master Nodes / Worker Nodes)
Rates are per GPU, in 1x, 2x and 4x sizes. Every instance ships with 100 GB of disk, 50 GB OS and 50 GB data, scalable in 50 GB steps.
H200 SXM (141GB)
Starts at $4.72 / hour
From $1,865.12/month
(36-month reserved)
NVIDIA
H100 SXM (80GB)
Starts at $4.39 / hour
From $1,779.96/month
(36-month reserved)
NVIDIA
RTX Pro 6000 (96GB)
NVIDIA
L40S (48GB)
Starts at $1.95 / hour
From $713.96/month
(36-month reserved)
NVIDIA
L4 (24GB)
Starts at $1.17 / hour
From $428.37/month
(36-month reserved)
NVIDIA
Every node includes local NVMe and sits on a non-blocking fabric. Managed WEKA performance storage attaches for workloads that need extreme throughput and IOPS.
8 x B300 (2,304GB)
NVIDIA
8 x RTX Pro 6000 (768GB)
NVIDIA
8 x MI300X (1,536GB)
$12,981.96 / month
36-month reserved
AMD
8 x H200 SXM (1,128GB)
$13,810.58 / month
36-month reserved
NVIDIA
8 x H100 SXM (640GB)
$12,422.59 / month
36-month reserved
AMD
8 x L40S (384GB)
$4,302.79 / month
36-month reserved
AMD
Flat across every term from 1 to 36 months. Worker nodes are billed as GPU instances at the rates above, on demand or reserved.
VKE Master Node (Non-HA)
$154.44/ month
NVIDIA
3-Master Node (HA)
$463.33/ month
NVIDIA
Startup program
See a lower GPU compute bill as you grow.
Ship your AI on Neysa AI Cloud, and earn deeper discounts at every milestone you hit.
Pay for only what you use
One idea runs our whole cost model: you pay for the compute you use, and see it before you commit.
Predictable pricing.
Every rate is published and billed to the minute, so the number you plan for is the number you pay.
Transparent cost for the rest.
Setup, platform, provisioning, storage I/O, API calls, your first VPC, an IP and uplink.
Flexible configuration.
Size storage and bandwidth to the workload, and change them when it changes.
Included Add on
- vCPU, system RAM and GPU memory
- 100 GB instance disk (OS and data)
- Local NVMe on every bare-metal node
- First VPC on the platform, plus an IP
- 1 Gbps uplink
- Jupyter notebooks and ML lifecycle tools
- Command Center for health, usage and spend
- Inference endpoints, at the GPU rate
| Add it when you need it | Cost |
|---|---|
| Storage: block, file or object | $0.10 / GB / month |
| Performance storage on WEKA | $0.11 / GB / month |
| Dedicated bandwidth for your VPC | $1.89 / Mbps |
| Bandwidth beyond 10 TB | $0.04 / GB |
| Kubernetes cluster foundation (VKE) | from $154.44 / month |
| Load balancer, single or fault tolerant | $341.07 / $682.13 per month |
| Extra public IP | $5.56 / month |
| First VPC on bare metal | $345.86 / month |
ESTIMATE YOUR SPEND
Know your GPU compute cost before you commit.
Tell us four things, which GPU, how many, how many hours, and for how long, and we work out the price for you.
Here is what you get
- On-demand and reserved prices, side by side, so you can see what committing saves.
- A cost for your actual workload, whether it trains or serves models.
- The monthly bill, and the total across the term.
- Your bill in rupees or dollars, whichever you plan in.
- Kubernetes clusters priced in too, not left out.
Run your AI in production with Neysa.
Speak to our experts for custom solutions. Build your AI without compromises.