Finding the sweet spot for India's Sovereign AI

The right balance between economic considerations and strategic imperatives will be key

By Satsheel Shrotriya and Rohan Chinchwadkar

Published: Aug 13, 2024 01:43:56 PM IST

Updated: Aug 14, 2024 02:07:25 PM IST

The impetus for Sovereign AI transcends mere technological supremacy; it embodies a nation’s quest for strategic independence in an increasingly digital world. Image: Shutterstock

As nations grapple with the relentless pace of technological disruption, the Sovereign Artificial Intelligence (AI) framework has emerged as a pivotal element in the strategic calculus of states. Sovereign AI encapsulates the burgeoning trend of countries harnessing AI to bolster their technological autonomy, economic vitality, and security apparatus. It signifies a paradigm shift towards a self-sufficient AI infrastructure, fostering a resilient domestic ecosystem capable of competing on the global stage.

The United States of America (USA) and China are at the forefront of this race, pouring billions into AI research and development. Recently, the Indian government announced that it plans to invest more than Rs10,000 crore in its AI Mission. These commitments underscore a recognition of AI’s transformative potential and its ability to redefine the contours of global governance, particularly in the digital domain.

The impetus for Sovereign AI transcends mere technological supremacy; it embodies a nation’s quest for strategic independence in an increasingly digital world. It embodies a country’s capacity to cultivate, implement, and govern AI systems in harmony with its sovereign interests. Jensen Huang, CEO of NVIDIA, emphasised that “every country needs to own the production of their own intelligence. It codifies your culture, your society’s intelligence, your common sense, your history – you own your own data”.

While the allure of Sovereign AI is manifold, its design and implementation can be highly complex and require a nuanced handling of several considerations: economic, technological, strategic, ethical, political and organisational, to name a few. In this article, we emphasise that developing countries like India need to find the sweet spot in the policy framework that strikes a balance between economic realities and strategic imperatives to get the biggest bang for their buck. Rather than making and owning everything, governments need to create a tailored policy for different layers of the AI stack.

To orchestrate the creation of an effective AI ecosystem, the government can use a combination of four actions at its disposal: make, buy, collaborate and regulate.

‘Make’, which provides the highest level of control and self-sufficiency, should be reserved only for the most strategically critical components and use cases which are unable to attract meaningful investment from the private sector.
‘Buy’ is for components/systems where the private sector (domestic or foreign) has a significant cost advantage over the long term.
‘Collaborate’ focuses on building partnerships between public and private institutions to achieve goals like research and development, technology transfer, large-scale manufacturing, customisation, and so on.
‘Regulate’ is about developing a regulatory framework to protect India’s economic and strategic interests while leveraging the power of its vast data and application estates to attract the best global technology players and enable domestic companies to invest in the Sovereign AI ecosystem.

Also read: What Indian managers should know about Generative AI

To better understand how these choices can play out in the design and implementation of India’s Sovereign AI system, we need to delve deeper into the different layers of the AI stack. A simplified version of the AI stack contains four layers: hardware, cloud, AI models, and applications.

The hardware layer largely refers to high-performance computing hardware like graphics processing units (GPUs) and other specialized AI chips. ‘Make’ is less suitable for areas where technology is progressing at breakneck speed, like high-end chips used for training foundational AI models. However, ‘make’ could be well-suited for end-to-end manufacturing of chips designed to solve India-specific use cases in sectors like defence, agriculture and education. India should ‘buy’ from the top players, like NVIDIA, AMD and Intel, to benefit from the best-in-class innovation, while ‘collaborating’ with these players to secure a stable supply of controlled critical components. Regulation needs to be pragmatic that supports formation of domestic capacity and capital.

The cloud layer involves on-demand scalable infrastructure with advanced capabilities for data storage and computing. For national use cases, the ideal end-state would involve a combination of public, private, hybrid and sovereign cloud - subject to sensitivity of data classification, data residency, privacy and security requirements. Public sector buying from hyperscalers and domestic high-capability players at competitive prices is an efficient solution, which will likely attract the wider global ecosystem to invest in India. Regulation needs to set standards and promote choice, interoperability, and create a competitive market.

The best approach for AI models, also called foundational models, would be to let a thousand flowers bloom! Policy needs to foster an ecosystem that welcomes and rewards top AI innovators to build, and scale their models, both proprietary and open-source, in India. ‘Make’ is expensive, and hence should be only used in selective cases where gaps emerge in the ecosystem, and in smaller models for specific problems; ‘buy’ from the best-in-class model builders. The government can seed innovation in open-source and proprietary model development in specific use cases, while creating regulation that promotes responsible AI governance and enables investments in research and innovation.

Finally, AI applications is a high-potential area where India can invest to accelerate experimentation within local players trying to solve large problems. India is also home to the majority of global system integrators (GSIs) who have access to key decision makers of end-users like banks, healthcare providers and energy companies as well as Big Tech and independent software vendors (ISVs), creating a significant advantage. Here too, ‘make’ should be reserved only for strategic use cases. Public sector should leverage domestic private sector expertise as its primary sourcing strategy, with the intention of positioning India as a global hub for building and scaling AI apps for all, including for the global south. Regulatory framework should continue to enable rapid growth and internationalization of domestic players.

Apart from focusing on its AI sovereignty and economic efficiency, India has the opportunity to lead and partner with Sovereign AI initiatives of other developing nations. The trajectory of Sovereign AI is set to be a defining force in shaping the geopolitical landscape of the 21st century, with implications that will reverberate for generations to come.