Yotta to deploy 20,000+ NVIDIA Blackwell Ultra GPUs in $2 billion AI supercluster, DGX Cloud cluster coming to India
Indian data center and AI infrastructure provider Yotta Data Services said it will deploy more than 20,000 NVIDIA Blackwell Ultra GPUs to build one of Asia’s largest AI superclusters, alongside hosting one of the region’s largest DGX Cloud clusters from NVIDIA, in a combined multi-billion-dollar expansion of AI compute capacity in India.
The company announced plans to install 20,736 liquid-cooled NVIDIA Blackwell Ultra GPUs as part of a supercluster expected to go live by August 2026, with total infrastructure investment exceeding $2 billion. The deployment will be anchored at Yotta’s 60 MW D2 data center at its Greater Noida hyperscale campus, which is scalable to 250 MW, with additional capacity drawn from its Navi Mumbai campus, designed to scale up to 2 GW.
As part of the same buildout, NVIDIA will establish one of the largest DGX Cloud clusters in the Asia-Pacific region within Yotta’s Blackwell-based supercluster under a four-year engagement valued at more than $1 billion. NVIDIA’s DGX Cloud platform has already been using Yotta’s GPU infrastructure over the past year, with the new agreement significantly expanding that footprint.
The companies said the move reflects a broader shift toward geographically distributed AI compute infrastructure across trusted regions, as demand for large-scale model training and inference capacity accelerates worldwide.
The supercluster will be built on NVIDIA reference architecture and will incorporate 800 Gbps Quantum-X800 InfiniBand networking, advanced liquid cooling and more than 40 petabytes of high-performance parallel storage. Yotta said the platform is designed to support trillion-parameter foundation model training and high-throughput inference workloads handling millions of simultaneous prompts.
Alongside global capacity, Yotta said it will allocate more than 10,000 of the Blackwell Ultra GPUs from the cluster to India’s government-backed AI initiatives, including the IndiaAI Mission, to support domestic foundation model development, startups and research institutions. The company said this split model is intended to balance sovereign AI priorities with international AI compute demand.
Yotta is also integrating NVIDIA’s software stack into its Shakti Studio AI platform, including Nemotron open models, NIM microservices and the NVIDIA AI Enterprise suite, to support model development and deployment for enterprises and developers in India. The company said access to open models and optimized inference services will allow local developers to build and fine-tune production AI systems on domestic infrastructure.
Darshan Hiranandani, chairman and co-founder of Yotta Data Services, said AI infrastructure is becoming a foundational layer of economic capacity and that the Blackwell Ultra supercluster strengthens India’s role in the global AI value chain. CEO Sunil Gupta added that large-scale domestic compute is essential for India to build globally competitive and sovereign AI systems.
NVIDIA CEO Jensen Huang said the deployment creates frontier-scale AI infrastructure in India capable of supporting advanced model training and population-scale AI services, while expanding NVIDIA’s AI factory footprint in the region.
Yotta said it currently operates more than 10,000 NVIDIA GPUs in production, with another 8,000 scheduled to go live in the next quarter. With the addition of the Blackwell Ultra deployment, the company plans to scale beyond 80,000 GPUs by FY27 and has outlined a longer-term roadmap to support more than one million GPUs over the next three to five years through phased data center and power capacity expansion.
