India Extends Deadline for ₹10,000 Crore GPU Tender Amid Industry Concerns

MeitY Extends GPU Tender Deadline Amid Industry Concerns

The Ministry of Electronics and Information Technology (MeitY) has extended the deadline for submitting bids for its ₹10,000 crore graphics processing unit (GPU) tender to November 12, 2023. This marks the second extension of the deadline, which was originally set for September 6 and later pushed to October 16 1

1

2

Tender Objectives and Specifications

The tender, part of the India AI Mission, aims to establish public AI cloud infrastructure with a capacity of at least 10,000 GPUs through a public-private partnership. Key aspects of the tender include:

1. Bidders are required to have at least 1,000 AI compute units on their cloud platform.
2. The government has lowered performance specifications to allow more participation from smaller companies and startups 1

   1

   .
3. A new clause requires achieving a power usage effectiveness (PUE) of less than 1.35, which has raised concerns among some companies 2

   2

   .

Industry Concerns and Challenges

Several data centre operators have expressed concerns about the tender's requirements:

1. The PUE criterion of less than 1.35 is seen as technically challenging in India due to high ambient temperatures.
2. Achieving this PUE may require increased capital expenditure, which could be passed on to customers.
3. Direct-to-chip cooling technology, while efficient, may not be sufficient to meet the PUE requirements in India's climate 2

   2

   .

Related Stories

Global Tech Giants Compete for India's ₹10,000 Crore AI Tender Amid GPU Ambitions

India's AI Infrastructure Race: Government Selects Finalists for 10,000 GPU Procurement

India's AI Mission Receives 18,000 GPU Bids in Second Round, Boosting Indigenous AI Development

India Accelerates AI Ambitions: Government Assigns GPU Projects Under IndiaAI Mission

Energy Efficiency and AI Infrastructure

Sunil Gupta, CEO of Yotta Data Services, a bidder for the GPU tender, highlighted some key points:

1. AI workloads are extremely high-density, requiring about 100 KW per rack compared to typical 6 KW racks.
2. Globally, data centres consume nearly 3% of total global energy production, which could increase significantly with the introduction of thousands of GPUs.
3. In India, data centres currently consume only 0.2%-0.3% of the country's total energy production, making it less of a concern compared to other countries 2

   2

   .

Future Outlook

Despite the challenges, there is significant interest in AI infrastructure investment in India:

1. Six or seven large conglomerates are willing to invest heavily in AI infrastructure.
2. Yotta Data Services has already placed an order for 16,384 H100 GPUs and deployed the first phase with 4,096 GPUs.
3. Technologies such as direct-to-chip cooling and immersion cooling could help reduce power consumption in the future 2

   2

   .

As of now, the tender's clauses remain unchanged, but the government is reviewing the concerns raised by some companies. The extension of the deadline provides more time for potential bidders to prepare their proposals and for the government to address industry concerns.