“Democratising access to AI is key”: Interview with Yotta Data Services’ Sunil Gupta

India’s data centre landscape is undergoing a decisive shift, with sovereign compute, high-density graphics processing unit (GPU) deployments and artificial intelligence (AI)-native cloud platforms emerging as defining priorities for the next phase of digital growth. In an interview with tele.net, Sunil Gupta, Co-founder, Chief Executive officer and Managing Director, Yotta Data Services, discussed the company’s evolution from a hyperscale data centre operator to a full-stack AI infrastructure provider, Yotta’s role in advancing India’s sovereign AI ambitions and the company’s future plans. Edited excerpts…

Can you share a brief background on Yotta and how the company has evolved so far?

We founded Yotta Data Services with a vision to build India’s most comprehensive sovereign digital infrastructure, spanning hyperscale data centres, cloud, AI and managed services. From the very beginning, we focused on owning and operating Tier-3 and Tier-4 certified data centre campuses. This began with NM1 campus in Panvel, Asia’s largest Tier-4 data centre. Since the launch of the facility in 2020, we have scaled to four operational data centres across India: D1 in Greater Noida, G1 in GIFT City, Gandhinagar and TB1 and TB2 in Airoli, Mumbai.

This infrastructure-first approach allowed us to move up the value chain rapidly. Once the physical layer was firmly established, we expanded into infrastructure-as-a-service, offering compute, storage, networking and eventually, GPU-as-a-service (GPUaaS). This led to the launch of Shakti Cloud, India’s first AI cloud platform designed entirely on sovereign infrastructure to support large-scale AI training, inference and high-performance computing (HPC) workloads.

As demand diversified, we extended this infrastructure into various purpose-built platforms. Shakti Studio enables enterprises and developers to build, train and deploy AI models through managed AI workspaces and tools. Urja is a cloud-native Renderfarm-as-a-Service platform that supports GPU-intensive rendering and visual workloads for media, engineering and design use cases. Sudarshan is an end-to-end media asset management and online video platform that allows organisations to store, manage, distribute and monetise digital content. To support this growth at scale, we have further expanded our data centre capacity, completing the core-and-shell development of NM2 in Panvel, Navi Mumbai and laying the foundation for D2 and D3 at our Greater Noida campus.

Today, our evolution is defined by three clear vectors: building hyperscale capacity, delivering sovereign GPU platforms and enabling an ecosystem that drives India’s AI leadership in infrastructure.

How has customer demand evolved over the past one to two years? Which customer groups are you focusing on now and why?

Over the past one to two years, customer demand has shifted sharply from traditional colocation and basic cloud services toward AI-ready, high-performance computing infrastructure. This change has been accelerated by generative AI adoption and the enactment of Digital Personal Data Protection (DPDP) rules across sectors such as banking, financial services and insurance (BFSI), healthcare and manufacturing, among others.

Today, demand is no longer limited to training AI models and has expanded meaningfully into large-scale inferencing. As a result, start-ups and research institutions have emerged as a critical demand segment. Government-backed initiatives and the broader push for indigenous AI models have further intensified this demand. Additionally, NVIDIA has also partnered with Yotta in capacity of a customer and will anchor Asia’s first DGX Cloud supercluster within Yotta’s infrastructure, consuming nearly half of the new GPU capacity under a four-year contract valued at about $1 billion.

Today, Yotta is focusing on three primary customer groups – large enterprises modernising their AI and data platforms, AI-first start-ups building foundational and applied models, and public-sector and academic institutions aligned with national AI priorities.

What are the top two-three design priorities Yotta is building around for the next phase of capacity? How are you preparing your facilities for high-density AI workloads?

Our next phase of capacity is built around three core priorities – high-power density, efficient cooling and scalability.

Cooling is a critical focus area. We deploy a combination of advanced air-based and liquid-assisted cooling solutions, including rear-door heat exchangers and direct liquid cooling readiness, to support GPU-dense environments. We achieved a PUE of <1.4 on air cooled CPU workloads and <1.2 on direct-to-chip liquid cooled GPU workloads.

Scalability is the third pillar. As we fully own our infrastructure, we can quickly retrofit for new GPU generations and commission modular capacity aligned to contracted demand, avoiding stranded capacity, while enabling rapid cluster expansion.

What is Shakti Cloud solving for customers that they do not get from a typical cloud option? How do you track whether your AI offerings are doing well?

Shakti Cloud addresses three critical gaps that many customers face with traditional cloud platforms -sovereignty, AI-scale performance, and ecosystem depth.

First and foremost, sovereignty and trust; Shakti Cloud is built as a sovereign AI platform for India, ensuring data residency, regulatory compliance and operational control—something that global hyperscalers often cannot guarantee end to end.

Second, AI-native performance at scale. Unlike general-purpose clouds that retrofit AI on top of the existing infrastructure, Shakti Cloud is designed AI-first. It offers access to cutting-edge NVIDIA GPUs such as H100, L40S and upcoming B200, combined with advanced networking such as Quantum InfiniBand and NVLink.

Third, a complete AI ecosystem, not just infrastructure. Shakti Cloud goes beyond raw compute. Through deep engineering collaboration with NVIDIA and partnerships with players such as Microsoft, it provides optimised GPU provisioning, performance tuning and platform-level support for AI/ML, HPC and advanced analytics.

At the adoption level, we look at how quickly customers are moving from pilots to production – GPU utilisation rates, repeat usage and expansion of workloads across training, inference and analytics are strong indicators of real value creation.

From a performance standpoint, we continuously monitor metrics such as model training times, throughput, latency and infrastructure efficiency.

How have partnerships supported Yotta’s growth so far? Which were your most important partnerships in 2025, and what key partnerships are you working on next?

Partnerships have been central to Yotta’s growth strategy. Our partnership with NVIDIA has been particularly foundational. Yotta is one of the few global Reference Platform NVIDIA Cloud Partners, which allows us to build Shakti Cloud on NVIDIA’s advanced GPU infrastructure and AI software stack.

Through this partnership, we deployed large clusters of NVIDIA H100 GPUs to power AI workloads and are continuing to expand capacity as demand grows. We are also scaling our AI compute capacity significantly, with plans to deploy clusters based on NVIDIA’s next generation Blackwell architecture as part of a major AI infrastructure expansion, aimed at positioning India as a global AI computing hub.

Another important dimension of our ecosystem is collaboration with global cloud platforms. Our partnership with Microsoft integrates Azure AI services with Shakti Cloud, enabling developers, enterprises and government institutions to build AI applications using Microsoft’s tools, while running on sovereign infrastructure located in India.

In addition, collaborations with organisations such as Sarvam AI and Bhashini are supporting the development of sovereign large language models trained entirely on India resident infrastructure, strengthening India’s indigenous AI ecosystem.

What role is Yotta playing under the India AI Mission? What is your roadmap to scale up your AI compute and services?

Under the India AI Mission, Yotta plays a foundational role as a compute and platform provider. We contribute 75 per cent of the mission’s advanced GPU capacity, making high-end AI compute accessible to start-ups, researchers, enterprises and government bodies through the Shakti Cloud. This directly supports the mission’s objective of democratising AI infrastructure. Yotta has also committed 10,000 NVIDIA B300 GPUs from the AI supercluster to the India AI Mission, out of the 20,000 we are deploying by August, 2026.

Our roadmap to scale includes phased deployment of advanced GPUs, expansion of AI-optimised data centre capacity and continuous enhancement of AI services such as GPUaaS, AI labs and model APIs.

Over the coming years, where will you invest the most? What kind of funding approach suits your growth best and why?

Over the coming years, our largest investments will continue to go into AI-ready digital infrastructure in India. This includes hyperscale data centre capacity, high-density power and cooling systems and advanced GPU compute. Besides the physical infrastructure, we are investing steadily in higher-layer AI services that make this compute usable for enterprises, start-ups and government programmes. Our focus is firmly on building long-term, sovereign AI capacity within the country, rather than pursuing overseas expansion.

From a funding perspective, our business requires patient, long-duration capital that matches the lifecycle of infrastructure assets. We are planning a public listing in FY2027. Any pre-IPO capital we raise will be focused on strengthening execution and increasing India-centric AI infrastructure.

What are Yotta’s top three priorities for 2026? What are your main expansion and investment plans for the next two to three years?

For 2026, Yotta’s priorities are centred on scaling AI infrastructure, while accelerating the shift toward accessible AI platforms. The first pillar is AI readiness at scale by expanding data centre footprint, power and fibre dense infrastructure, and GPU capacity, while ensuring data centres remain future proof for next-generation workloads through industry leading PUEs and efficient, high-density design. As part of this expansion, Yotta is deploying a large AI supercluster with 20,000 next generation NVIDIA B300 GPUs, strengthening India’s ability to support large-scale AI training and inference workloads.

The second pillar is democratising access to AI. Through GPUaaS, managed orchestration and platforms such as Shakti Cloud and Shakti Studio, Yotta is lowering barriers to AI adoption for enterprises, start-ups, researchers and institutions. These platforms enable customers to focus on building, training and deploying models, particularly inference workloads, without the complexity of owning or managing underlying infrastructure.

Finally, ecosystem development remains critical. By offering affordable, sovereign and globally competitive AI infrastructure and by partnering with ISVs, start-ups, academia and the government, Yotta aims to ensure that more of India’s AI innovation is trained, deployed and governed on Indian infrastructure, aligned with global performance, security and compliance standards.