Commoditising AI as India's Next Digital Public Infrastructure

Commoditising AI as India's Next Digital Public Infrastructure

#GS-3 #Science & Technology #Artificial Intelligence #GS-2 #Governance & Social Justice #E-Governance #Current Events #National #Digital Public Infrastructure #IndiaAI Mission #BHASHINI #Unified Intelligence Interface

Why in News

  • Public policy experts suggest turning Artificial Intelligence into India's next Digital Public Infrastructure (DPI).
  • This concept advocates for lower inference costs and open-source models inspired by Aadhaar, UPI, and DEPA.

Understanding AI as Digital Public Infrastructure

  • Commoditising AI as DPI means treating AI intelligence and inference as an accessible public utility rather than a costly commercial product.

Rationale for AI as Digital Public Infrastructure

  • Adding AI as a foundational layer represents the logical next step after digital identity (Aadhaar), payments (UPI), and data sharing (DEPA).
  • Establishing AI as a public utility helps India stop exporting raw data while importing expensive foreign API tokens.
  • Lowering compute and inference costs provides democratic access to advanced AI tools for rural healthcare, local education, and small businesses.
  • Reducing foundational model costs shifts economic value toward application developers, where India holds its greatest software strength.
  • Utilizing open-source architecture prevents domestic startups from depending completely on foreign technology giants and their pricing policies.

Initiatives Taken So Far

  • The government launched the IndiaAI Mission with a total outlay of ₹10,372 crore to expand affordable compute access for startups.
  • India has onboarded over 38,000 GPUs with plans to reach 1,00,000, offering compute access at roughly ₹65 per GPU hour.
  • Initiatives like BHASHINI build public datasets and translation models across official languages to enable multilingual AI tools.

Challenges

  • India lacks domestic semiconductor manufacturing for high-end AI chips, creating a heavy reliance on imported GPUs.
  • Power grid planning currently fails to classify AI workloads separately, leading to energy shortages for compute infrastructure.
  • Foreign technology companies control most advanced foundation models, creating long-term strategic reliance on proprietary systems.
  • AI training data remains heavily skewed toward Western contexts, resulting in a shortage of structured regional language datasets.
  • Domestic developers face high dollar-denominated token costs when renting foreign models, which restricts large-scale experimentation.

Way Forward

  • Integrating data center compute demands into the National Electricity Plan will fast-track renewable energy supply and lower power costs.
  • Mandating open-weights licenses for models trained with public subsidies or anonymized public datasets will ensure shared access.
  • Developing a Unified Intelligence Interface (UII) will create an open API gateway that connects applications to various AI models.
  • Establishing a national freemium token system will offer state-subsidized monthly API allowances to verified startups and students.
  • Reallocating a small portion of existing industrial and agricultural subsidies into token entitlements will boost AI skilling and research.

Conclusion

  • Converting AI into an open-source public utility transforms India from a raw data provider into an application innovation hub.
  • Combining compute access, Indic language models, and a Unified Intelligence Interface ensures inclusive economic growth through advanced technology.