By Inder Gopal & D Manjunath, respectively professors at Indian Institute of Science, Bangalore, and IIT Bombay
Anthropic’s release of Claude Cowork and Claude Code triggered a sharp dip in Indian IT stocks, signalling a vastly changed technology ecosystem. India’s manpower-intensive services companies are clearly under threat from artificial intelligence (AI), and a transition to an intellectual property-driven future leveraging AI is now an existential necessity. When confronted in Davos recently with the label “second-tier AI power”, Union Technology Minister Ashwini Vaishnaw gave a thoughtful riposte, laying out a layered AI taxonomy and arguing that leadership in many of the layers makes India decidedly not second tier.
While this assertion is sound, the underlying anxiety is real—Indians aspire to a indigenous or desi AI that is an unequivocal global leader. The key question about realising this aspiration is not if the government should be involved, but how. We answer this through a careful examination of historical successes and failures in governance of technology development—from Tokyo to Washington, and from Centre For Development Of Telematics (C-DoT) to United Payments Interface (UPI).
Many well-intentioned government projects fail due to a misunderstanding of technological evolution. In the early 1980s, Japan’s manufacturing engine, powered by the ministry of international trade and industry (MITI), was unstoppable. Buoyed by this success, MITI launched the ambitious Fifth Generation Computer Systems (FGCS) project with a massive 10-year budget. They mobilised universities and corporations to develop native hardware for AI. Initial momentum was high and fear of Japanese prowess caused tremors in the US as it contemplated losing its technological lead. However, FGCS ended in failure, set Japan’s computer industry back, and contributed to its “lost decade”.
What went wrong? FGCS was driven by planners and policymakers building for the present not for the future. Borrowing a football metaphor, MITI was “running to where the ball is and not where the ball will be”. Their hubris and a docile community following their lead drove the project down a dead-end path. This provides a stark lesson—non-technologists cannot and should not micro-manage technology innovation.
The development of the internet in the US offers a contrasting blueprint for success. In the 1960s, the Defense Advanced Research Projects Agency (DARPA) sought a resilient communication system and funded a radical proposal for a “packet switched” network from Paul Baran, a technologist at the RAND Corporation. Amid scepticism from many, DARPA handed control to technologists such as David Clark from the Massachusetts Institute of Technology, who collectively established Internet’s core principles. Clark’s guiding mantra—rough consensus and working code—prioritising practical functionality and adaptability over rigid, bureaucratic specifications and standards, is still the principle of Internet evolution.
Taiwan followed a similar trajectory for its semiconductor industry. They recruited Maurice Chang, an Intel executive, to lead their investment in chip foundries. His insight to use bleeding-edge R&D to create a “pure-play foundry” paid off at Taiwan Semiconductor Manufacturing Company Limited (TSMC), transforming global chip manufacturing in a way that no industry outsider could have. Clearly, government must appoint the right experts in decision-making roles. DARPA routinely hires external experts to manage programmes rather than generalist administrators. India must similarly ensure that experts are in charge, not merely advisers without power or risk of failure.
The government’s role is best viewed as the first runner in a relay race. Provide a vision and assume initial risks that may be too high for commercial entities, possibly creating the first version of a solution. Then the baton must be passed to the private sector, unleashing their “animal spirits” to take innovations to the finish line. India Stack and the UPI are testaments to what is achieved when the government provides a vision and puts the right people in charge; here the government provided the infrastructure, policy, and regulatory support, but allowed the ecosystem to flourish.
The baton exchange to the private sector is where many projects fail. A prominent example is C-DoT. Tasked with modernising India’s telecom infrastructure, C-DoT had early successes with microprocessor-based exchanges that worked in India’s harsh conditions of heat, humidity, and dust. But when the “dog caught the car”—after the technology was shown to work—non-technologists stepped in to “manage” the process, licensing only a hardware specification and matching object code to manufacturers. This meant private companies competed on manufacturing efficiency and not on innovation. Animal spirits were unleashed, but the beast was on a leash.
Contrast this with China, where Huawei and ZTE started around the same time as C-DoT. They were encouraged to collaborate, co-develop, and eventually compete with multinationals. Internal competition was nurtured, regulatory help was provided, and today they are global giants. C-DoT lost momentum when control passed to non-technologists.
To ensure India’s position as a top-tier AI power, the government must synthesise these and other lessons into a coherent strategy. We humbly suggest a tech-governance panchshila.
First, enable, don’t direct. Provide broad objectives and let the best minds determine the trajectory. Micro-management causes stagnation, but broad vision leads to innovation.
Second, establish expert leadership. Avoid the trap of administrative generalists running technical programmes. Recruit domain experts and empower them to make decisions and take risks, like TSMC and DARPA did.
Third, provide strategic funding, not corporate charity. Fund innovation and risk-taking and resist the temptation to flood the ecosystem with capital. Funding start-ups just for the sake of it leads to mediocrity and a “dumbed down” start-up ecosystem.
Fourth, embrace open-source. Open-source is a “rising tide” raising all boats and must be preferred over closed, nationalist silos. Regional language large language models are valuable, but such projects alone cannot make India a global leader. We must contribute to, not just benefit from, global commons.
Finally, and most importantly, have patience. Recognise that good things take time. Copycat technologies are often several years too late and will most likely provide only short-term headlines.
The path to Desi AI is to emulate the Internet and the UPI, where the government builds the rails and the private sector runs the trains. Provide a vision, put experts in charge, fund genuine risk, and step back for private sector innovation. The relay race must be executed without tying the runner’s legs.
Disclaimer: The views expressed are the author’s own and do not reflect the official policy or position of Financial Express.