Why India must rethink semiconductors to win the AI race

If India aspires to global leadership in artificial intelligence, it must first secure its footing in the sector’s most vital building block: semiconductors. The real question is how best to approach this challenge. Success depends less on how much money is spent on new fabrication plants and more on the strategic choices India makes about where to deploy its scarce capital and attention.

One route is to invest in fabrication plants, the massive industrial complexes where chips are physically manufactured. This is the approach most AI commentators in India are strongly vouching for. The other is to pursue a fabless model, in which companies focus on chip design and intellectual property while outsourcing the manufacturing to specialised foundries abroad. This latter approach has powered some of the world’s most successful semiconductor firms, and it offers India a faster and more pragmatic path to influence in the global AI supply chain.

The country already possesses a strong base in design. Tens of thousands of Indian engineers work in chip design, and global leaders such as AMD, Nvidia and Intel operate major research centres in Bengaluru, Hyderabad and Noida. This talent pool covers the full spectrum of expertise, from integrated circuits to system-on-a-chip development. A fabless strategy would allow India to put this capability to work immediately, creating the blueprints for the next generation of AI processors, rather than waiting decades for fabrication capacity to mature.

By contrast, attempting to compete in advanced manufacturing would demand immense resources and still leave India at a disadvantage. A single state-of-the-art foundry can cost more than $20 billion to build, with its technology vulnerable to obsolescence in just a few years. Only a handful of players, such as TSMC and Samsung, dominate the most advanced nodes. Matching them would be like entering a sprint against competitors who already hold a commanding lead. Beyond the financial challenge, there are infrastructure barriers: fabs require uninterrupted electricity and vast quantities of ultra-pure water, both of which remain difficult to guarantee in much of India.

The experience of today’s AI leaders shows the power of the fabless model. Nvidia, which dominates the AI hardware market, focuses on design and outsources manufacturing to TSMC. ARM, another giant, has built a global business by licensing its chip architectures without operating a single fab. These companies demonstrate how design-led strategies can be more profitable, more nimble and less capital-intensive than owning fabrication facilities.

Agility is especially critical in the age of AI. The field is evolving at breakneck speed, with new models and algorithms reshaping demand every few months. Companies tied to their own fabs are often constrained by fixed processes and heavy sunk costs, while fabless firms can adapt quickly, experiment with new architectures and bring products to market faster.

That ability to pivot is a decisive competitive edge in such a dynamic sector.

India’s natural strengths in software provide another reason to lean into design. A thriving fabless ecosystem would allow the country to pair hardware innovation with advanced AI frameworks and applications optimised for India-designed chips. This would create an end-to-end stack, from silicon to software, without the financial and infrastructural burden of maintaining foundries.

Choosing this approach recognises India’s comparative advantage, and channels resources toward areas where the country can lead globally. By concentrating on design and intellectual property while relying on established manufacturing specialists, India can accelerate its rise as a central player in the AI economy. The fabless path aligns with national strengths, avoids costly distractions and offers a realistic chance to shape the chips that will define the future.

(Views are personal; the author is an AI researcher and investor, spent 10 years in JP Morgan leading AI research)