Neysa scales AI Infrastructure in India

Neysa's founder, Sharad Sanghi, emphasised that India cannot depend on platforms it doesn't control to build its AI future
Sharad Sanghi, CEO, Neysa
Sharad Sanghi, CEO, NeysaLinkedIn
Published on: 

Neysa, an AI cloud provider, has scaled its AI infrastructure in India and deployed a large-scale AMD Instinct MI350X GPU cluster on Neysa's Tytan,  a dedicated AI infrastructure offering. Neysa is an AI acceleration cloud provider helping enterprises and AI builders train, deploy and operate AI at scale.

Founded in 2023 by Sharad Sanghi and Anindya Das, Neysa is backed by Blackstone. Neysa's AI infrastructure footprint is based in India, and it aims to build sovereign AI in the country.

Neysa's founder believes India cannot rely on platforms it doesn't control to build its AI future. Indigenous infrastructure gives customers in India full control over their AI environments, rather than relying on opaque, overseas-hosted platforms with complex compliance requirements and unpredictable costs.

The 8,192-GPU deployment is now operational, expanding dedicated bare-metal AI infrastructure and accelerator choice for global offtakers and Indian enterprises deploying at scale in India. According to Neysa, global customers have already contracted this capacity to support large-scale AI deployments in India.

In February this year, Neysa announced a $1.2 bn fundraise to scale compute capacity. Private equity fund Blackstone and co-investors invested $600 mn in equity. This enabled Neysa to secure an additional $600 mn in debt financing. Other equity investors include Teachers’ Venture Growth, TVS Capital, 360 ONE Assets, and Nexus Ventures. Neysa is using the funding to deploy around 20,000 GPUs in India.

The question for large AI deployments is no longer simply how many GPUs are available. It is whether the infrastructure can be deployed at scale, operated with certainty and configured for the workload.

Sharad Sanghi, co-founder and CEO, Neysa

Neysa Tytan gives AI offtakers control: dedicated bare-metal capacity and the freedom to choose the silicon, architecture and operating model that fit their requirements. "With AMD Instinct MI350X on Tytan, we are expanding that choice while delivering the scale global customers need to deploy AI infrastructure in India,” Sanghi said.

Neysa Tytan is a dedicated bare-metal AI infrastructure offering. It handles deployment and day-to-day operations as one accountable partner. Each customer gets single-tenant, high-density clusters in India and can run them three ways: bare metal with full OS access, Slurm for distributed training, or with Kubernetes.

According to Neysa, in large AI deployments, the expensive mistake is paying for GPUs that sit idle, slow down or fail mid-run. "Neysa Tytan gives customers dedicated capacity on terms they choose, with rates that fall as the commitment grows. Customers also pick the accelerator that fits each workload, so they don’t overpay for silicon they don’t need. Because Neysa runs the cluster, customers don’t have to build their own operations team to keep thousands of GPUs productive,” Sanghi said.

India’s ability to participate meaningfully in the next phase of AI will depend on the availability of AI compute infrastructure that can be delivered at scale, operated reliably and adapted to different requirements.  

The cluster is built for production-scale AI: training and fine-tuning large language models, generative AI, high-performance computing and high-throughput inference. In practice, that means AI developers training foundation and domain models, and enterprises serving models to millions of users where latency and uptime directly affect revenue.

Neysa Tytan competes on cost per useful hour of compute, not on headline GPU price. Dedicated, single-tenant hardware means performance doesn’t vary with other tenants. Term-based pricing rewards longer commitments. A choice of silicon matches price and performance to the workload. And because Neysa runs operations, customers don’t incur the cost of building their own infrastructure team. The result is predictable costs for customers and planning around multi-year AI programmes.

Being silicon-agnostic, the customer’s workload decides the architecture, not the supply chain. Customers choose the accelerator, network fabric, storage platform and service levels. Today that includes AMD Instinct MI350X alongside Nvidia options.

India’s ability to participate meaningfully in the next phase of AI will depend on the availability of AI compute infrastructure that can be delivered at scale, operated reliably and adapted to different requirements.  

Vinay Sinha, India managing director, Sales, AMD, said, "Organisations are looking for flexibility in how they build and scale AI, and open ecosystems play an important role in enabling that choice. The AMD Instinct MI350X GPU is designed to support a broad range of AI workloads while giving customers the ability to align infrastructure decisions with their performance, operational and business requirements."