Why Open Networks, DPIs Are Key to Scaling Social Impact
At the India AI Impact Summit 2026, executives argued that open networks, digital public infrastructure and lower AI inference costs are essential to expand access across India and the Global South.
Topics
News
- Altman’s AI Energy Comparison to Human Development Draws Flak
- Why Open Networks, DPIs Are Key to Scaling Social Impact
- Anthropic Expands Claude Into Code Security With AI Reasoning
- IBM Expands GovTech Push in India With Lucknow AI Center
- Tata Communications, RailTel Partner to Upgrade Network for AI Era
- Microsoft Taps Asha Sharma to Lead Gaming as Spencer Retires
Artificial intelligence is advancing quickly, but access remains uneven across regions and income groups, with data concentrated among large technology firms.
At a session, titled “AI and Open Networks Creating Impact at Scale,” leaders from Biocon Group, Google, Networks for Humanity, the World Bank Group and Wadhwani AI discussed how open systems could broaden adoption.
Nandan Nilekani, co-founder and chairman of Infosys Ltd and co-founder of Networks for Humanity, pointed to India’s Unified Payments Interface as a model for AI diffusion. Open, interoperable networks would allow innovators to build AI applications at the edge, he said.
“We keep talking about agents, but the real power of agents is removing complexity for the user,” Nilekani said. Systems that operate in local languages and hide technical complexity could help “get everybody on the system.”
Sangbu Kim, Vice-President for Digital Transformation at the World Bank Group, said open standards are essential as services shift toward user-centric models. “Open standard and open network are crucial to ensure user-centric service,” he said.
Healthcare and agriculture are among the sectors poised for the largest gains.
Kiran Mazumdar-Shaw, Executive Chairperson of Biocon Group, said India’s growing pool of phenotypic, genomic and radiological data could support large-scale AI-driven health delivery.
“The objective becomes universal healthcare delivery at scale in a sustainable way,” she said.
Biological systems, she added, operate with remarkable energy efficiency. AI could complement that intelligence if built on high-quality data.
“Biological intelligence is extraordinary. Cells signal, form circuits, and make decisions through distributed networks that run on sips of energy, not gigawatts. If we can learn from that efficiency and fuse it with artificial intelligence,” Shaw noted.
Digital public infrastructure, or DPI, remains foundational, according to Sunil Wadhwani, Founder and Chairman of Wadhwani AI.
“Without that infrastructure, you simply can’t build effective AI for the social sector,” he said.
Wadhwani AI used India’s tuberculosis data platform, Nikshay, to build models addressing diagnosis delays and treatment drop-offs.
A smartphone-based cough analysis tool helped raise TB detection by 25% nationally within a year of rollout, he said.
Predictive models now help caseworkers focus on patients most likely to discontinue medication.
Lower inference costs will determine how widely such systems spread.
“If you are serving a customer with a query costing Rs500, it won’t work,” Nilekani said.
AI inference costs vary widely based on model complexity and provider, with prices ranging from roughly $0.0001–$0.001 per request for simple tasks (CPU-based) to $0.001–$0.01+ for GPU-accelerated tasks.
“As bigger models are developed and stabilized, the focus will shift to the inference to make it affordable,” he added.
Low-cost inference combined with agentic AI hiding complexity is the key to mass AI diffusion.
For the future Nilekani is hopeful.
“I’d like to see massive diffusion, where applications are rolled out on open networks to reach millions of farmers, students, and patients, and show the world that AI is a force of good. I think we have an obligation to show that otherwise, I think the pushback will be huge.”
