Start free
← Back to Current Affairs
Current Affairs · 27 October 2026

India's AI Firms Have Money but Wait Months for Chips; Rented GPUs vs Token Costs

RPSC RASDainik BhaskarSource: Dainik Bhaskar (Jaipur e-paper, 27 Oct 2026)
1

Why in News

  • Dainik Bhaskar (Jaipur e-paper, 27 October 2026) reports from Mumbai/New Delhi that Indian AI companies have capital but cannot obtain GPU chips.
  • Waiting time for AI model-building and running chips is three to four months, stalling work; fast memory, steady power and data-centre space are also scarce.
  • Juspay co-founder Sheetal Lalwani said the firm wants to add GPUs but was told a three-to-six month timeline.
  • The report contrasts renting GPUs with buying tokens, and notes the IndiaAI Mission has over 38,000 GPUs on subsidised rent.
2

What is the issue

  • Capital versus compute: Indian AI firms reportedly have investment money but face a physical shortage of GPU chips, so projects stall despite funding.
  • Broader bottlenecks: besides chips, the article flags shortages of high-speed memory, continuous electricity supply and space inside data centres.
  • Early-stage cushion: Rahul Agrawal of SenseAI Ventures says new firms get free cloud credits with investment, so spending is not the first problem.
  • Token billing model: AI services are billed per token (a small unit of text), so costs rise automatically as usage grows.
  • Open-source alternative: running free open-source models on own GPUs avoids token billing; Juspay does about 95% of its AI coding this way.
  • Cost comparison: Lalwani says an 8-node GPU setup rents for Rs 25-35 lakh a month, while the same work via tokens may cost five to ten times more.
  • Buy versus rent debate: Ashish Tejwa of ProEx Ventures advises application firms to rent GPUs so capital is not locked in machines that age quickly.
  • Supply chain concentration: only three firms - SK Hynix, Samsung and Micron - make the fast memory used in advanced AI chips.
  • Sovereignty and security: a robotics founder says robots need on-device computers for instant response, and defence-type work cannot use shared servers.
Detailed analysis

The full analysis is in the app

This page is the summary. The detailed analysis of this story covers:

What is the statusWhat are the challengesWhat measuresWhat lies aheadImportant Data PointsKey conceptsPrevious year questions
Open in Quorum IAS