A. Strategic Leap in India’s AI Compute Capacity

In a dynamic push to scale AI compute capabilities, IndiaAI Mission CEO Abhishek Singh confirmed that the third round of the GPU tender will bring in about 3,850 new compute units—a powerful boost to the nation’s AI infrastructure. This milestone ushers total deployed hardware past 38,000 GPUs/TPUs, marking a significant stride in establishing one of the world’s most robust public-sector AI clusters.

B. Google Trillium TPUs Join the Fold

Adding a new dimension to its compute arsenal, IndiaAI has, for the first time, integrated 1,050 Google Trillium TPUs—Google's sixth-generation Tensor Processing Units optimized for high-efficiency AI and machine learning workloads . This inclusion signifies an upgrade in India’s AI training and inferencing capacity.

C. Contributions from Empanelled Bidders

No new vendors were empanelled in this round. Instead, existing partners reinforced their commitments:

1. Locuz is supplying 1,300 NVIDIA H100 GPUs.
2. Ishan Infotech brings in 50 additional Google Trillium TPUs.
3. Sify is delivering 2,500 units, broken down as: 1,000 Trillium TPUs, 800 NVIDIA H200 GPUs, and 700 NVIDIA L4 GPUs.
4. Vensysco reduced pricing but did not add new hardware in this round .

D. Competitive Bidding & Price Adjustments

The financial evaluation of the third round has wrapped up, with minor adjustments to the L1 (lowest bidder) prices. Now, vendors from earlier rounds are being called upon to match these revised rates, reinforcing competitive pricing dynamics .

Previously, price drops between the first and second rounds ranged significantly—from as low as ₹4 per hour to as steep as ₹1,234 per hour for similar GPU models .

E. Scaling GPU Pool and Government’s Vision

IndiaAI surpassed its initial GPU procurement target of 10,000 units long ago. After the May 30 announcement by IT Minister Ashwini Vaishnaw, the cluster reached 34,333 GPUs. With the latest addition of 3,850 units, total compute capacity exceeds 38,000 units, positioning IndiaAI among the leading national AI compute platforms globally .

F. Accelerating AI Innovation and LLM Development

The enhanced infrastructure, now offering one of the lowest subsidized compute rates globally—less than a dollar per hour—is tailored to propel AI research, support startups, academia, and foster indigenous LLMs .

From over 506 proposals received under the foundation model initiative, 43 focus specifically on LLM development, representing a strong wave of interest in building an Indian AI backbone .

G. Expanding Accessibility and Ecosystem Support

Competitive pricing and subsidized access are further bolstered by ease-of-use through the IndiaAI compute platform. This strategic infrastructure investment opens opportunities for startups, researchers, academia, and government agencies to fuel projects—especially in AI and deep learning development .

4 – Why This Matters (SEO-friendly Highlights)

1. AI Infrastructure Growth: IndiaAI Mission’s third tender significantly expands national compute capability.
2. Deep Learning Compute Boost: Integration of Trillium TPUs signals powerful enhancement for AI workloads.
3. Affordable GPU Hours: Subsidized pricing offers startups and institutions cost-effective access.
4. Large Language Models Ecosystem: Intensive support for LLM proposals nudges India toward self-reliant AI leadership.
5. Competitive Tendering: Price adjustments and vendor participation demonstrate market dynamics and scale efficiencies.

Source:economictimesGPT