India AI Impact Summit 2026

building-public-interest-ai-catalytic-funding-for-equitable-access-to-compute-resources

5 min read

TRACK:Democratizing AI Resources
SPEAKERS
  • Abhishek Singh, Ministry of Electronics & IT, GoI
  • Andrew Sweet, The Rockefeller Foundation
  • Deepali Khanna, The Rockefeller Foundation
  • Dr. Saurabh Garg, Ministry of Statistics and Programme Implementation, GoI
  • Martin Tisné, Government of France
  • Shaun Seow, Philanthropy Asia Alliance
  • Shikoh Gitau, Qhala
  • Sushant Kumar, Kalpa Impact
  • Vilas Dhar, Patrick J. McGovern Foundation
SESSION OVERVIEW

This event will mark the launch of the working report on "Opening Up Computational Resources for New AI Futures" by Secretary, MOSPI in the presence of Additional Secretary, MeitY. The event includes a keynote and a panel discussion that brings together senior govt officials and global AI experts to discuss pathways for democratising AI. How will catalytic funding for public interest and South–South partnerships democratise access and affordability of compute in the Global South?

VIDEO RECORDING

Building Public Interest AI: Catalytic Funding for Equitable Access to Compute Resources

Detailed Summary

  • Purpose – The moderator frames the session as a move from “diagnosis to action” on the compute divide that is turning the digital divide into a compute divide.
  • Key Message – Democratisation of AI is not merely “catch‑up” but “expanding who gets to lead”. India’s AI Mission and its Compute Capacity Plan (≈ 38,000 GPUs) are presented as a flagship public‑interest model that can be replicated across the Global South.
  • Three‑Act Structure – The discussion will explore (i) What is being democratised?, (ii) How South‑South partnerships & catalytic financing accelerate progress?, and (iii) Concrete commitments for the year.

2. Launch of the Working Report (Moderator)

  • The working report titled “Opening up Computational Resources for New AI Futures” is released.
  • The report was drafted by the Democratising AI Resources Working Group, co‑led by Dr Saurabh Garg.
  • Attendees are invited to give feedback until 31 March.

3. Keynote Speech – Dr Saurabh Garg

3.1 Framing the Challenge

  • Compute as the Core Barrier – Access to GPUs, accelerators, and high‑performance clusters is the primary bottleneck for AI development worldwide.
  • Four Pillars – Compute, Collaboration, Connectivity, Compliance & Context are identified as essential levers.

3.2 Vision of a Shared Compute Infrastructure

  • Public‑Good Model – Proposes treating compute as a shared, digital public good that can be modular, voluntary, and non‑binding.
  • METRI Platform – Introduced as a Multi‑Stakeholder AI for Trusted & Resilient Infrastructure digital public good that countries could adopt, customize, and extend.
  • Beyond Hardware – Emphasises the need for open‑source models, data governance, talent development, and institutional capacity.

3.3 Future‑Facing Questions

  • Raises the possibility of domain‑specific, lightweight models that could reduce compute demand, contrasting the current trend toward ever‑larger models.
  • Calls for rethinking compute‑heavy paradigms to make AI more inclusive.

4. Panel Discussion (Moderated by Andrew Sweet)

4.1 Governance of Compute as a Public Utility – Dr Saurabh Garg

  • Intelligent Prioritisation – Instead of outright rationing, compute should be allocated based on public‑interest priorities.
  • Philanthropic Role – Philanthropy can help de‑risk projects, unlock capital, and facilitate collaborations to make affordable compute widely available.

4.2 Moving Nations from Consumers to Co‑Creators – Martin Tisné

  • Current Concern – Risk of “white‑elephant” data centres that remain under‑utilised.
  • Two‑Fold Focus – (i) Data – Need for local, multilingual datasets; (ii) Open‑Source Ecosystem – Critical dependencies are under‑funded, and philanthropic funding is needed to sustain them.

4.3 Data Bottleneck & Stewardship – Martin Tisné (continued)

  • Highlights the lack of scalable data‑trust mechanisms and the necessity for privacy‑preserving data sharing.
  • Calls for significant resources to develop technical safeguards and governance frameworks.

4.4 South‑South Collaboration & Compute Demand Index – Dr Shikoh Gitau

  • Compute Demand Index – Quantifies GPU‑hour needs for Africa (≈ 2.5 M GPU‑hours/year; 7.5 M GPU‑hours over three years).
  • Investment‑Readiness Index – Assesses whether a country has the power, talent, data, and use‑cases to effectively consume compute.
  • Practical Example – Africa currently can provide only 5 % of its required GPU‑hours; partnerships with India (or other donors) must be needs‑based and use‑case‑driven.

4.5 Institutional Architecture for Catalytic Funding – Vilas Dhar

  • Argues for new institutional intermediaries (e.g., “Culpa Impact”) that can bridge governments, philanthropies, and private sector to align resources with public‑interest AI goals.
  • Stresses the need for co‑creation of AI stacks that are open, resilient, and locally adaptable.

4.6 Philanthropy Asia Alliance – Sean Xiao

  • Network of 80 members can coordinate shared compute resources across Asia (India, Indonesia, etc.).
  • Suggests aggregating demand to negotiate better cloud pricing and subsidising compute for impact‑driven startups.

4.7 Critical Reflections on Sovereignty & Interdependence – Andrew Sweet

  • Challenges the Westphalian notion of sovereignty (pure territorial control of hardware) and proposes a relational sovereignty focused on data stewardship and agency.
  • Calls for an open, collaborative global stack that respects privacy while providing choice and agency.

4.8 Final Remarks – Dr Saurabh Garg

  • Emphasises the need to extend frameworks beyond compute – encompassing models, talent, data, and governance for a comprehensive public‑interest AI ecosystem.

5. Closing & Announcements

  • The working report is now publicly available; feedback is requested by 31 March.
  • Kalpa Impact and other partners thanked for logistical support.
  • The session concludes with acknowledgements to all panelists and a reminder of the broader AI Summit agenda.

Key Takeaways

  • Compute is the new frontier of the digital divide; equitable access is essential for inclusive AI development.
  • India’s AI Mission (38 k GPUs) exemplifies a public‑interest compute ecosystem that can be replicated across the Global South.
  • The METRI platform is proposed as a modular, voluntary digital public good for shared compute infrastructure.
  • Catalytic philanthropy should focus on risk reduction, capital unlocking, and partnership convening to make affordable compute widely available.
  • South‑South collaborations must be use‑case‑driven; the Compute Demand Index (e.g., 2.5 M GPU‑hours/year for Africa) offers a concrete metric for partnership planning.
  • Open‑source dependencies are under‑funded; targeted philanthropic support is required to sustain critical components.
  • Data stewardship and privacy remain major bottlenecks; scalable data‑trust mechanisms are needed.
  • Institutional intermediaries (e.g., “Culpa Impact”) are crucial to connect governments, donors, and innovators for public‑interest AI outcomes.
  • Philanthropy Asia Alliance can aggregate demand to secure cheaper cloud compute for impact‑focused organisations.
  • A shift from territorial sovereignty to relational agency is needed to create an open, resilient global AI stack that balances privacy with collaborative innovation.

See Also:

Graph View

Backlinks