India’s most significant AI infrastructure commitment to date combines Yotta’s hyperscale ambition with NVIDIA’s largest APAC DGX Cloud deployment, expected to go live by August 2026 across facilities in Greater Noida and Navi Mumbai.
Yotta Data Services has announced it will deploy 20,736 liquid-cooled NVIDIA Blackwell Ultra GPUs at its hyperscale campuses in Greater Noida and Navi Mumbai, forming one of Asia’s largest AI superclusters. The total investment exceeds $2 billion, with the infrastructure expected to go live by August 2026. Alongside this, NVIDIA will establish one of APAC’s largest DGX Cloud clusters within Yotta’s supercluster under a four-year engagement valued at over $1 billion, significantly expanding a relationship that has been building quietly over the past year.
The scale of the numbers is notable, but the structure of the deal is what gives it strategic weight. NVIDIA is not simply selling hardware to a regional customer. It is anchoring one of its largest Asia-Pacific cloud deployments inside Indian infrastructure, which reflects a meaningful shift in how NVIDIA is thinking about compute distribution across trusted geographies. India, in this framing, is not a secondary market receiving overflow capacity. It is a primary node in a global supply chain that is actively being redistributed.
Jensen Huang, co-founder and CEO of NVIDIA, said: “India is emerging as one of the world’s most important AI markets, driven by extraordinary talent and a bold national vision. Yotta’s deployment of one of the largest NVIDIA Blackwell Ultra superclusters creates advanced AI infrastructure capable of training frontier-scale models and delivering AI at population scale. Expanding AI Factory capacity in India strengthens NVIDIA’s regional footprint while supporting India’s ambition to build secure, sovereign, and globally competitive AI.”
What the infrastructure actually does
The supercluster is built on NVIDIA reference architecture and integrates 800 Gbps NVIDIA Quantum-X800 InfiniBand networking, advanced liquid-cooling systems, and over 40 petabytes of high-performance parallel file-system storage. It is engineered to support trillion-parameter foundation model training and high-throughput inference workloads capable of handling multi-million simultaneous prompts. These are not specifications for a regional demonstration facility. They match what frontier AI development requires.
The primary deployment site is Yotta’s 60 MW D2 data centre at its Greater Noida hyperscale campus, scalable to 250 MW. Additional capacity comes from the Navi Mumbai campus, scalable to two gigawatts. Yotta has outlined a roadmap to scale beyond 80,000 NVIDIA GPUs by FY27, supported by phased infrastructure expansion across both sites. The company’s longer-term trajectory points to over one million GPUs within three to five years, supported by integrated extra-high-voltage substations, dedicated power distribution, and green energy sourcing built into the campus design.
A portion reserved for India’s sovereign AI mission
Not all of the capacity goes to global commercial demand. Yotta is committing over 10,000 NVIDIA B300 GPUs from the supercluster to the IndiaAI Mission, supporting sovereign Indian foundation model development, research institutions, startups, and population-scale public AI platforms. That allocation runs in parallel with the NVIDIA DGX Cloud deployment rather than competing with it, which is a more considered approach than treating domestic and international demand as separate pipelines.
Yotta is also augmenting its Shakti Studio AI platform with NVIDIA Nemotron open models, NVIDIA NIM microservices, and access to the full NVIDIA AI Enterprise software suite. Through Shakti Studio, developers in India gain access to the Nemotron family of open models including weights, training datasets, and recipes, enabling fine-tuning, customisation, and sovereign AI development without requiring offshore compute.
Sunil Gupta, Co-Founder, MD and CEO of Yotta Data Services, said: “India’s AI ambition requires sustained, high-performance compute at scale. By combining Blackwell Ultra infrastructure with open models like NVIDIA Nemotron and the full NVIDIA AI stack, we are enabling developers to build sovereign, globally competitive AI applications from India.”
Capital strategy and what it signals
The combined commitments, more than $2 billion in Blackwell Ultra infrastructure and a $1 billion-plus multi-year DGX Cloud engagement, provide the kind of long-term demand visibility that large infrastructure investments require to make commercial sense. Yotta is not betting on demand arriving. It has contracted demand in place before the hardware goes live.
Darshan Hiranandani, Co-Founder and Chairman of Yotta Data Services, said: “AI infrastructure is becoming foundational economic infrastructure. This NVIDIA Blackwell Ultra supercluster reinforces India’s position in the global AI value chain. Our capital strategy is focused on building scalable infrastructure that serves both national priorities and international AI demand.”
Yotta currently operates over 10,000 NVIDIA GPUs in live production, with another 8,000 going live within the next quarter. The August 2026 Blackwell Ultra deployment adds 20,736 on top of that base. The trajectory is sequential and funded, not aspirational.
Where this fits in India’s broader infrastructure moment
India in early 2026 is receiving AI infrastructure commitments at a pace and scale that would have seemed implausible two years ago. Microsoft, Google, AWS, and now Yotta with NVIDIA have all announced multi-billion dollar compute commitments within months of each other. What distinguishes the Yotta announcement is the combination of frontier-scale hardware, a contracted global cloud anchor tenant, and an explicit domestic allocation running alongside it.
The practical implication for Indian AI developers, research institutions, and enterprises is straightforward. Frontier-scale compute, under Indian governance, with open model access built into the platform, is becoming available domestically. The question of whether India’s AI ecosystem can absorb that capacity productively, and whether the open access commitments hold as commercial demand grows, will be answered over the next two to three years. The infrastructure, at least, is no longer the constraint.
