Yotta Plans a 95,000-GPU AI Deployment

Yotta Data Services is preparing a major expansion of its AI computing infrastructure in India. The company is reportedly planning to order 50,000 NVIDIA Vera Rubin GPUs and 45,000 GB300 GPUs. Together, the two platforms would give Yotta a planned deployment of 95,000 NVIDIA accelerators. The equipment is intended for Yotta’s proposed 120 MW D4 data centre in Delhi, which is targeted to go live in 2027. The reported GPU and networking package carries a value of about $7.5 billion. Yotta also expects to spend another $750 million on data-centre construction. That makes the project much larger than a conventional server expansion. It points instead to an infrastructure build designed around the rising power and compute demands of advanced AI workloads.

The scale of the proposed deployment is significant for India’s AI infrastructure market. Large GPU clusters require more than accelerator capacity alone. They need high-density power systems, advanced cooling, high-speed networking and carefully engineered data-centre environments. Yotta’s Delhi project would bring all of those requirements together within a single 120 MW campus.

$7.5 Billion Order Signals a New AI Infrastructure Scale

Yotta Co-founder, Managing Director and Chief Executive Officer Sunil Gupta outlined the company’s reported procurement plans and the investment attached to them. His comments also clarify that the procurement process is still underway. That distinction matters because the reported figures describe an upcoming order rather than equipment already delivered.

“We are in the process of ordering 50,000 Vera Rubin GPUs and 45,000 GB 300s. The overall cost of acquiring the Vera Rubin chips would be around $7.5 billion, which includes all networking and associated equipment. An additional $750 million will also be required for data centre construction.”

The proposed investment would place Yotta among the most aggressive AI infrastructure builders in India. The reported $7.5 billion figure covers the Vera Rubin chips, networking and associated equipment. Construction would require another $750 million, taking the combined reported requirement to roughly $8.25 billion. That level of spending shows how quickly AI infrastructure economics are moving beyond traditional data-centre investment models.

Delhi’s D4 Data Centre Becomes the Key Project

Yotta plans to deploy the reported GPUs at its D4 data centre in Delhi. The facility is planned around 120 MW of capacity and is expected to support large AI computing workloads. Yotta has indicated a targeted go-live window between May and August 2027. GPU deliveries are expected to take place between March and June 2027.

Gupta said, “Our project timelines to go live with this is May-August 2027, for which deliveries of GPUs to our data centre shall be made around March-June 2027,” he said.

That timeline gives the D4 project a defined sequence for hardware delivery and infrastructure readiness. The facility must be ready to receive and integrate a large number of advanced accelerators within a relatively tight window. Power distribution and cooling will be especially important at this scale. Networking will also play a central role because large AI clusters depend on fast communication between GPUs.

Vera Rubin Raises the Stakes for AI Compute

NVIDIA’s Vera Rubin platform represents the next stage in the company’s accelerator roadmap. The platform targets demanding AI training and inference workloads that require substantial compute performance. Its role in Yotta’s planned deployment therefore extends beyond simply increasing GPU numbers. It could also give the company access to a newer generation of AI infrastructure as workloads become more complex.

The proposed combination of Vera Rubin and GB300 systems would create a large accelerator footprint across NVIDIA’s latest platforms. That could support AI training, inference and increasingly sophisticated enterprise workloads. AI systems are also moving toward agentic applications that can perform multiple tasks and interact with external tools. Such applications can create sustained demand for compute rather than relying only on occasional model-training workloads.

For Yotta, the timing is particularly important. AI infrastructure providers are competing to secure advanced accelerators while customers are looking for reliable access to high-performance compute. A large domestic deployment could strengthen Yotta’s ability to serve that demand from infrastructure located in India. It could also support organisations that want greater control over where their AI workloads and data are processed.

Yotta Expands Its AI Infrastructure Strategy

The proposed Delhi deployment follows Yotta’s broader push into high-density AI infrastructure. The company has already announced plans to deploy 20,736 NVIDIA Blackwell Ultra GPUs at its D2 data centre in Greater Noida. That project is based around a 60 MW facility and represents an investment of more than $2 billion. The planned D4 deployment would therefore mark another substantial increase in the scale of Yotta’s AI infrastructure ambitions.

Yotta’s strategy increasingly centres on building infrastructure specifically for accelerated computing. Traditional data centres can accommodate a wide range of enterprise workloads. AI facilities, however, place much greater demands on power density, cooling and network performance. Yotta’s expansion suggests that it expects these requirements to become a defining part of India’s next data-centre cycle.

India’s AI Compute Market Gains Momentum

India’s demand for AI compute is growing as businesses, technology companies and developers adopt increasingly capable models. That demand creates an opportunity for domestic infrastructure providers to build large-scale GPU capacity within the country. Yotta’s proposed investment would add another major project to that expanding market. The scale of the planned deployment also reflects the growing connection between AI adoption and data-centre infrastructure.

However, installed GPU capacity does not automatically translate into profitable AI infrastructure. Operators need strong utilisation rates and long-term customer commitments to support investments of this magnitude. They must also manage electricity costs, cooling requirements, network performance and hardware availability. These factors will determine whether large AI campuses can deliver sustainable returns as competition increases.

The proposed D4 campus could therefore become an important benchmark for India’s AI infrastructure sector. Its success will depend on more than completing construction and receiving the GPUs. Yotta will need to turn those accelerators into consistently utilised compute capacity. That requires customers, software infrastructure and reliable operations working together.

Execution Will Determine the Value of the GPU Bet

The reported order puts execution at the centre of Yotta’s next growth phase. A 95,000-GPU deployment requires coordination across semiconductor supply, networking, construction, power infrastructure and cooling. The D4 project must bring these elements together before the targeted 2027 operational window. Any delay in one part of the chain could affect the wider deployment schedule. The construction requirement adds another layer to the project. Yotta has indicated that $750 million will be required for the data-centre build itself. That investment sits alongside the much larger reported hardware and networking commitment. The combined spending highlights the infrastructure intensity behind modern AI compute.

The March-to-June 2027 GPU delivery window will therefore be closely tied to the progress of the Delhi facility. Yotta will need the supporting infrastructure ready when the accelerator systems arrive. It will also need to ensure that the systems can operate at the density expected from a large AI cluster. The project’s execution could offer a useful measure of India’s readiness for hyperscale AI infrastructure.

What 95,000 GPUs Could Mean for India

If Yotta completes the reported procurement and brings D4 online as planned, the project could materially increase India’s available AI compute capacity. The proposed combination includes 50,000 Vera Rubin GPUs and 45,000 GB300 GPUs. That would create one of the country’s largest planned deployments of advanced NVIDIA accelerators. It would also strengthen Yotta’s position in a market where access to high-performance compute is becoming strategically important. The significance goes beyond the headline GPU number. Large accelerator clusters require substantial power, cooling and networking infrastructure to remain productive. They also require customers that can use the capacity at scale. Yotta’s investment therefore reflects both the opportunity and the financial risk emerging around India’s AI infrastructure build-out.

The project could also influence how other data-centre operators approach AI infrastructure. Larger deployments can create economies of scale when demand remains strong. They can also increase exposure to hardware cycles and rapidly changing AI architectures. Yotta’s decision to commit to the next generation of NVIDIA systems shows how infrastructure providers are increasingly planning around the pace of AI hardware development.

Yotta’s Bet Could Test India’s AI Infrastructure Ambitions

Yotta’s reported $7.5 billion GPU and networking commitment represents a major escalation in India’s AI infrastructure race. The additional $750 million construction requirement makes the proposed investment even more significant. Together, the figures point to an infrastructure strategy built around large-scale accelerated computing rather than conventional enterprise capacity. The 120 MW D4 data centre would become the physical foundation for that strategy.

The bigger question is whether demand can keep pace with the infrastructure being planned. AI developers and enterprises need access to powerful compute, but providers must maintain high utilisation to justify enormous capital commitments. Yotta is betting that India’s AI market will require that capacity as model development, inference and enterprise adoption accelerate. If the company delivers the project on schedule, its Delhi campus could become a defining piece of India’s next-generation AI infrastructure landscape.