NEYSA TYTAN

GPU Clusters in India: High-Density Compute for AI Training

Deploy your AI on dedicated GPU clusters built on NVIDIA or AMD silicon, with the network interconnect of your choice. Our engineering team keeps your production runs stable.

Reserve capacity

Zero

Service interruptions or security breaches across 24 months

99.9% Availability

Contractual SLA maintained over six consecutive months

$1.2B Funding

Blackstone-led funding round securing the platform build-out

20,000+ Accelerators

Committed installation path through 2027

Built for:

Built with:

Built for:

Built with:

Why Neysa

Every cluster is engineered around the physics of your workload

We design infrastructure from the silicon up. The full stack is tightly integrated to meet the demands of AI model training and production inference.

Choice at every layer.

Your choice of accelerator, fabric, and storage, with the control plane and SLA built to your requirement.

Infrastructure that holds the line.

High-availability Active/Active or Active/Passive architecture with hot spares. A contracted GPU replacement commitment restores instances promptly.

Observability and deep monitoring.

Full-stack DCGM telemetry from the whole fleet down to each accelerator, across bare metal, Slurm and Kubernetes. See how GPU monitoring and ML observability works.

Operation and service assurance.

An India-based 24x7 NOC and SOC practice with a named technical account manager and a defined four-level escalation path.

Connectivity and network fabric.

Carrier-diverse uplink up to 100 Gbps, private Cloud Connect, and intelligent load balancing with health-checked failover.

Filesystems sized for the run.

Your choice of parallel filesystem over RDMA, WEKA, VAST or DDN, with S3-compatible object storage and persistent snapshots available.

“Their robust platform and continuous support allowed our team to focus squarely on the complexities of LLM development, significantly accelerating our progress in creating advanced AI capabilities for Indian languages.”

Rishi Bal

CEO

BharatGen (TIH, IIT Bombay)

17B Parameters

Param-2 mixture-of-experts architecture deployed in production

22 Trillion Tokens

Cumulative throughput sustained across the entire training run

Zero

Interrupted runs during massive-scale distributed training

"By partnering with Neysa, we have finally moved past the latency and cost barriers of legacy infrastructure to a system that handles our scale with total reliability."

Anil Porter

CIO

ITQ Technologies

141 Billion Tokens

Served every month across two production LLM services

14s to under 2s

P99 latency on live fare-rule and refund workloads

40% Lower TCO

Against the general-purpose cloud design it replaced

Available Compute

High-density 8-accelerator bare-metal deployments, tiered by term duration

Single-tenant hardware with local NVMe and a unified non-blocking fabric. Rates drop as your term grows, from 6 to 36 months. See GPU cloud pricing.

Reserve your cluster

NVIDIA B300 GPU

NVIDIA RTX PRO 6000

AMD Instinct MI350 GPU

Workload Orchestration

Deploy orchestration on your terms

Provision bare metal, initialize managed engines, or accept a turnkey Slurm environment.

Bare Metal For custom engineering stacks

OS-level access allowing your own runtimes, drivers, and orchestration on the silicon
Out-of-band management via BMC and Redfish API
Private, single-tenant deployments with optional MIG or SR-IOV partitioning

Managed Kubernetes Engine For training and inference

Managed control plane provisioned and maintained by Neysa
Integrated Command Center for fleet health and expenditure visibility

Slurm Workload Manager For distributed training

Turnkey cluster provisioning in standard configuration, delivered ready for production
Direct SSH access to the cluster

Regulatory-grade assurance

Infrastructure engineered for trust

Keep full sovereignty over your model weights and production runs, backed by rigorous governance, defined accountability and AI infrastructure security.

CSA Star
Level 2

CSA TRUSTED
CLOUD PROVIDER

SOC 2

ISO
27001:2022

ISO
27017:2015

ISO
27018:2019

DPDPA
READINESS

INDIA AI MISSION
EMPANELMENT

Pricing Calculator

FAQ

GPU cluster FAQs

What GPUs can I get in a Neysa Tytan GPU cluster?

Tytan clusters are high-density, 8-accelerator bare-metal deployments. You can choose NVIDIA B300, NVIDIA RTX PRO 6000 or AMD Instinct MI350 GPUs, each on single-tenant hardware with local NVMe and a non-blocking fabric.

What contract terms are available for GPU clusters?

Clusters are reserved on terms from 6 to 36 months. Rates are tiered by term, so a longer commitment gets a lower rate.

How do I run workloads on the cluster?

You can take bare metal with OS-level access for your own runtimes and drivers, use the Managed Kubernetes Engine run by Neysa, or get a turnkey Slurm environment for distributed training with direct SSH access.

Where are Neysa GPU clusters hosted?

Tytan clusters run in India and are supported by an India-based 24x7 NOC and SOC team, so you keep full sovereignty over your model weights and production runs.

What uptime and support come with a GPU cluster?

Neysa maintains a 99.9% availability SLA, backed by hot spares and a contracted GPU replacement commitment. You get a named technical account manager and a defined four-level escalation path.

Which storage and networking options are supported?

You can choose a parallel filesystem over RDMA (WEKA, VAST or DDN), with S3-compatible object storage and persistent snapshots. Networking includes carrier-diverse uplinks up to 100 Gbps, private Cloud Connect and health-checked load balancing.

Which security and compliance certifications does Neysa hold?

Neysa holds CSA STAR Level 2, SOC 2, ISO 27001:2022, ISO 27017:2015 and ISO 27018:2019. Neysa is also a CSA Trusted Cloud Provider, DPDPA ready and empanelled under the India AI Mission.

Your AI has a home in India

Reserve your cluster