Indian Government Plans 5-Year Procurement of 10,000 AI GPUs

The Indian government is planning a long-term procurement arrangement for around 10,000 artificial intelligence (AI) graphics processing units (GPUs) for a period of five years, as it looks to expand access to high-performance computing and strengthen the country’s AI infrastructure. The move comes as demand for computing power continues to rise among startups, researchers, enterprises and government-backed AI initiatives.

The proposed procurement is linked to the broader IndiaAI Mission, which was approved with an outlay of ₹10,371.92 crore over five years. The mission’s compute pillar originally targeted the creation of a scalable AI infrastructure with at least 10,000 GPUs through public-private partnerships.

Government Looks at Long-Term GPU Procurement

The proposed five-year arrangement would give the government longer-term access to AI computing capacity rather than relying entirely on short-term purchases or individual cloud contracts.

Long-term procurement can also give the government greater negotiating power with technology and cloud infrastructure providers. By committing demand over several years, India could potentially seek more predictable pricing and capacity for AI workloads.

A similar approach has previously been discussed by technology-sector experts. A Moneycontrol analysis noted that the government could approach major cloud providers with a fixed GPU requirement over five years, allowing it to pre-commit spending and negotiate rates based on pooled demand.

The approach is particularly relevant because AI computing infrastructure is expensive and demand for advanced GPUs has increased rapidly.

IndiaAI Mission Already Targets 10,000 GPUs

The 10,000-GPU target is not a new policy objective. The Union Cabinet approved the IndiaAI Mission in March 2024 with a plan to establish public AI compute infrastructure containing 10,000 or more GPUs.

The government said the infrastructure would be made available through a public-private partnership model and would support AI startups, researchers and other users. An AI marketplace was also envisaged to provide access to AI-as-a-service and pre-trained models.

The original IndiaAI framework allocated more than ₹10,300 crore over five years to several areas, including compute capacity, datasets, AI applications, startup financing, future skills and safe and trusted AI.

IndiaAI Mission at a glance

Metric	Details
Mission outlay	₹10,371.92 crore
Mission duration	Five years
Original compute target	10,000+ GPUs
Model	Public-private partnership
Intended users	Startups, researchers, students, academia and other eligible users
Compute access	AI-as-a-service/cloud infrastructure
2026-27 IndiaAI allocation	₹1,000 crore

National GPU Capacity Has Already Expanded

India’s AI compute infrastructure has grown substantially since the IndiaAI Mission was launched.

The government said in March 2026 that more than 38,000 GPUs had been onboarded through the IndiaAI compute portal and were being made available to Indian startups and academia at affordable rates.

This figure is considerably higher than the original 10,000-GPU target, reflecting the government’s strategy of combining capacity from multiple providers rather than relying solely on government-owned hardware.

In February 2026, the government also announced plans to add another 20,000 GPUs to the existing national compute capacity.

The proposed five-year procurement therefore needs to be viewed as part of a larger effort to secure reliable AI computing capacity rather than simply an expansion of the original 10,000-GPU programme.

Why GPUs Are Important for AI

GPUs are particularly important for modern AI because they can perform large numbers of calculations simultaneously.

Training large language models, image-generation systems and other AI models requires enormous amounts of computing power. GPUs are designed to handle the parallel workloads involved in these processes more efficiently than conventional processors.

They are also increasingly important for inference, where trained AI models process requests from users and applications.

As Indian companies build AI applications and develop models for areas such as healthcare, finance, agriculture, education and government services, access to affordable compute can become a significant competitive factor.

Cost and Supply Remain Major Challenges

Securing GPUs for several years is not straightforward.

Advanced AI accelerators are expensive, while demand for high-performance computing has increased sharply worldwide. Supply constraints involving semiconductors and high-bandwidth memory have also contributed to higher infrastructure costs.

PRS India has highlighted the rising cost of AI compute infrastructure, noting that GPU demand has grown rapidly while supply has faced constraints.

The government therefore faces an important procurement question: whether to commit to a specific generation of GPUs for several years or structure contracts that allow newer hardware to be introduced as technology advances.

That issue matters because AI accelerators are evolving quickly. A GPU that is considered high-end today could be significantly less competitive against newer architectures several years later.

Long-Term Deals Could Improve Price Visibility

A five-year procurement model could provide greater visibility to both the government and suppliers.

For the government, a longer commitment could help secure capacity and potentially reduce the uncertainty associated with volatile GPU prices. For cloud and infrastructure providers, a predictable customer commitment could make it easier to plan data-centre capacity and hardware investments.

The model could also allow the government to negotiate GPU access as a service instead of purchasing every physical accelerator directly.

This distinction is important because owning GPUs is only one part of building AI infrastructure. Data centres also require electricity, cooling systems, networking equipment, storage and technical personnel.

Government Has Previously Used a Cloud-Based Model

India’s initial IndiaAI procurement approach focused substantially on empanelling companies that could provide AI compute and cloud services.

In December 2024, the Ministry of Electronics and Information Technology said more than 50 companies had participated in the pre-bid process for providing AI compute and cloud services. The programme was designed to make 10,000 GPUs available to startups, researchers, students and academics.

A later parliamentary response said empanelled bidders had offered 14,517 GPUs at L1 rates against the 10,000-GPU target, with an average discovered rate of ₹115 per GPU hour. The government also said eligible users would receive support covering 40% of AI compute costs.

This indicates that India’s approach has already moved beyond a simple government hardware-purchasing model.

What It Means for Indian AI Startups

For startups, reliable access to GPUs could reduce one of the biggest barriers to developing AI products.

Early-stage companies often cannot afford to build large computing clusters themselves. Paying for cloud GPUs can also become expensive when companies need to train models or run inference at scale.

Government-supported compute could therefore allow more startups to experiment with AI without making large infrastructure investments upfront.

Researchers and academic institutions could also benefit, particularly when they need significant computing resources for model training, simulations or other AI research.

The Risk of Technology Becoming Outdated

The biggest challenge with a five-year GPU commitment is technological obsolescence.

AI accelerator technology is developing quickly, with successive generations delivering improvements in performance, energy efficiency and AI-specific capabilities.

A long-term contract therefore needs flexibility. Instead of locking the government into one hardware generation, procurement agreements could potentially allow capacity to transition to newer GPUs over the contract period.

Policy researchers have previously highlighted this issue, recommending staggered acquisition and diversification of GPU models and technology providers to reduce the risk of infrastructure becoming outdated.

India Wants to Build Sovereign AI Capacity

The government’s GPU strategy is also part of a broader push to build domestic AI capabilities.

The IndiaAI Mission is intended not only to provide computing power but also to support indigenous AI models, datasets, applications, talent development and startup financing.

The government has also indicated that India wants to expand its overall national AI compute capacity significantly. The IndiaAI Mission CEO said in February 2026 that installed GPU capacity could reach 100,000 by the end of 2026, compared with about 38,000 at the time.

This suggests that the 10,000-GPU procurement plan is one component of a much larger national AI infrastructure strategy.

The Bigger Picture

India’s AI infrastructure strategy is increasingly shifting from simply making GPUs available toward securing sustained and scalable compute capacity. A five-year procurement framework could help the government provide more predictable access to computing resources while potentially improving its ability to negotiate with suppliers.

At the same time, the government’s existing compute capacity has already expanded beyond the original 10,000-GPU target. The key challenge now is ensuring that additional capacity is affordable, technologically current and accessible to the startups, researchers and institutions that the IndiaAI Mission is intended to support.

The procurement model will therefore matter almost as much as the number of GPUs. Flexible contracts, competitive pricing and the ability to upgrade to newer generations could determine whether India’s investment remains useful as AI technology continues to advance.

Looking Ahead

The proposed five-year arrangement could give India greater visibility over AI compute availability at a time when global demand for accelerators remains high. It could also strengthen the country’s ability to support AI startups and research institutions without requiring each organisation to build its own expensive infrastructure.

The longer-term objective will be to create an AI ecosystem in which computing power, datasets, models, talent and financing develop together. If procurement is structured to accommodate rapid changes in GPU technology, the government’s approach could provide a foundation for India’s expanding AI ambitions over the next several years.