In March 2024, India launched the IndiaAI Mission with an allocation of Rs 10,372 crore—roughly $1.25 billion—and an ambition to compete at the frontier of artificial intelligence. It was, by Indian standards, a serious commitment of resources. It was also, by the standards of every other serious technology nation, missing something fundamental.
From Washington to Seoul, from Tel Aviv to Beijing, governments that have decided to compete in frontier technology have made the same choice: Build institutions where technical judgement lives permanently—held by people who have actually done the work, with the authority to act on what they know. Not administrators. Not advisory committees. Not technical roles outsourced on contract to third parties. Operators with domain expertise, decision authority, and institutional permanence. DARPA was built on this principle. Israel’s National AI Directorate was built on this principle. Korea’s technology institutions were built on this principle. China’s AI governance structure, for all its differences, was built on this principle.
India has not made that choice. Not for the IndiaAI Mission. Not for the Atal Innovation Mission before it. Not for the decade of innovation initiatives in between. While the rest of the world has converged on a model—technical judgement inside the institution, as a permanent organisational property—India remains a conspicuous and costly outlier. In the AI race, the cost of that outlier status is no longer abstract.
Some will argue that a liberal democracy should leave frontier technology to the private sector—that government’s role is to set direction, not to do the work. It is a principled position. It is also one that no serious liberal democracy actually holds in practice when the stakes are high enough.
The argument here is not that government should take over AI—no successful model anywhere suggests that. The goal is more specific: The government needs the technical capacity to drive it. To make credible bets on where the frontier is moving. To evaluate outcomes with informed judgement rather than administrative process. To attract serious talent by signalling institutional seriousness. And critically, to activate the private sector—to provide the technical direction and validated pathways that give private capital the confidence to make serious frontier AI bets rather than safe, incremental, adoption-oriented ones. DARPA didn’t build the internet. It created the conditions under which the private sector could and did. That is the model. What distinguishes India is not its democratic values—it is the institutional gap that those values have been used to justify.
Also read: What South Korea’s KDI gets right that NITI Aayog never did
A pattern, not an exception
The Atal Innovation Mission, launched in 2016 under NITI Aayog, has built tinkering labs across thousands of schools and mentored millions of students. These are real outputs. They are also, by any serious measure of innovation—patents filed, products commercialised, frontier research produced—not what a country of India’s scale should be pointing to after a decade. Now a second iteration has been approved with a fresh budget, building on the same institutional architecture. The IndiaAI Mission followed the same template: Announce the ambition, allocate the budget, appoint a governance structure, and wait.
The National Quantum Mission, approved in April 2023 with a Rs 6,003 crore budget and reviewed by a committee of ministry secretaries chaired by a Minister of State, follows the same institutional template. So does the Atal Innovation Mission 2.0. The pattern runs across every frontier technology initiative India has launched in the past decade.
The numbers tell the same story. India’s parliamentary Standing Committee on Communications and Information Technology recently found that the IndiaAI Mission spent only 32 per cent of its allocated funds in FY 2025-26—and responded by cutting its FY 2026-27 budget by 50 per cent. Meanwhile, AI Curation Units—data management units to be placed across central ministries—were announced in 2025 and announced again in 2026, this time across 50 ministries rather than 20. The pattern is precise: Announce, underdeliver, announce again.
The pattern is a structural feature. And it is worth saying clearly that it is not a reflection on the quality or dedication of the administrators who run these missions. India’s civil service contains some of its finest minds. The problem is not who is in the room. It is who is missing from it. India’s innovation missions have consistently lacked the one organisational property that every comparable nation treats as non-negotiable.
That absence is, in an important sense, by design rather than oversight—and this is not a call to rearchitect everything or dismantle existing organisational frameworks. India’s administrative architecture does what it was built to do: It governs, coordinates, and implements at scale. The argument here is narrower and more specific. Frontier technology missions require a different organisational species—one where technical DNA is built into the institution itself, not imported on contract when needed and returned when the engagement ends. India inherited a governing architecture not designed for that purpose and has not yet built the complement it needs.
What DARPA actually is
The DARPA model is frequently invoked in Indian policy conversations and almost as frequently misunderstood. It is not simply an agency with a large budget and ambitious goals. Its organisational design is specific, deliberately unusual, and inseparable from its results.
DARPA is a small, flat organisation where every programme manager draws on deep industry, academic, or government experience. They conceive, plan, and oversee high-risk R&D efforts—defining challenges, setting milestones, and making the calls that determine where resources flow. The critical feature is not the budget or the mandate. It is that technical judgement is not outsourced, not filtered through generalist administrators, and not held by people on two-year rotations who arrive knowing nothing and leave before they have learned enough. It lives inside the organisation, distributed across a corps of people who have done the work, and it accumulates over time because the institutional culture selects for it relentlessly. That accumulation—of judgement, of domain knowledge, of hard-won understanding of what works and what doesn’t—is the asset that no budget line can substitute for.
This is the organisational species India has never built—not for deep tech, not for AI, and not for any frontier technology mission in memory. Not by failure of intention, but because the governing architecture inherited at Independence was designed for administration. The question now is whether India is ready to build something new alongside it.
Also read: India’s think tanks double up as lobbyists, give no disclosures. Hurt policymaking
What the rest of the world did
Israel drew a pointed lesson from this when it established its National AI Directorate under the Prime Minister’s Office. The most important structural choice was not where it was placed but who was put in charge and what authority they were given. Its first director came from military-technical leadership—someone with deep experience deploying AI systems under operational conditions where failure has real consequences, who understands the difference between a system that performs in demos and one that holds under pressure. The directorate was staffed with a small, high-trust team holding real decision rights over budget and partnerships. It was built to act, not to advise. That distinction—between an institution that decides and one that recommends—is the difference that matters. The PMO reporting line reinforced it, but the DNA came first.
Korea built the same principle into its permanent institutions differently. The Ministry of Science and ICT—MSIT—is a domain-specific ministry focused exclusively on science and technology, distinct in character from a generalist administrative ministry handling multiple unrelated portfolios. India has a Ministry of Science and Technology—but a ministry’s name does not determine its institutional DNA. What distinguishes MSIT is not its mandate on paper but the professional formation of its senior leadership, the tenure of its technical staff, and the degree to which domain knowledge is allowed to accumulate rather than rotate out. Beneath it, institutions like IITP and KIAT are staffed by career technologists on long tenures with genuine R&D programme authority. Technical judgement accumulates inside those institutions over years. It does not evaporate on rotation. The government does not need to hire consultants to tell it what it should already know, because what it should know was never allowed to leave.
China, for all the differences in its governance model, made the same structural choice. Its AI governance institutions bring Turing Award winners and frontier researchers into decision-making roles with direct access to the State Council. Technical legitimacy is not parked outside the room. It is seated at the table, with the authority to act across the full scope of the national agenda.
The private sector cannot fill the gap
The standard rebuttal is that India need not replicate any of these models. Why can’t the private sector carry the frontier work while government focuses on enabling conditions?
India’s R&D landscape answers that question directly. Despite years of aspirational speeches, detailed policy frameworks, and genuine political commitment to innovation, India’s gross expenditure on R&D has remained stubbornly below one per cent of GDP—among the lowest of any major economy. Prime Minister Modi has himself repeatedly urged Indian industry to raise the quality of its products, to move up the value chain, to stop competing on cost alone and start competing on capability. That the Prime Minister feels compelled to make this argument publicly is itself a signal: The private sector has not yet internalised innovation as a competitive imperative.
This matters for the AI mission specifically. Frontier AI is not a well-defined implementation problem. UPI—India’s celebrated payments infrastructure—is a genuine achievement and a model for what India’s government can do when the problem is well-structured, the technical requirements are known, and the task is execution at scale. Frontier AI is a categorically different challenge. It requires original research, continuous technical judgement under uncertainty, the ability to recognise what isn’t working and redirect resources before the window closes. The skills and institutional culture that delivered UPI are genuinely valuable. They are not sufficient for what IndiaAI is attempting.
The current structure of the IndiaAI Mission makes the gap concrete. Its technical roles are not held by permanent staff with institutional authority. It is the organisational opposite of what DARPA, Israel, and Korea have built.
This is precisely why the technocratic institution matters as a private sector catalyst, not just a government function. India’s conglomerates have capital. Its startups have energy. What they lack is the technical signal—the credible, authoritative direction that tells private investors where the frontier actually is, which bets are worth making, and what partnerships with global AI players are worth pursuing. Without that anchor, private capital will continue to flow toward safe, incremental, adoption-oriented AI rather than the frontier R&D that would move India’s position in the global technology order. The technocratic institution’s ultimate measure of success is not what government builds. It is whether India’s private sector begins making serious frontier AI bets—and a technically credible institution, driving rather than following, is what makes those bets rational.
The institution India needs
India does not need another inter-ministerial committee or a renamed advisory panel. And it does not need a handful of technocrats grafted onto the existing IndiaAI Mission structure. What India needs is a new institutional species entirely: A body where technical judgement is not outsourced, not advisory, and not rotational—but permanent, empowered, and built into the organisational DNA from the first day.
The IndiaAI Mission and MeitY’s existing architecture can remain intact. What India needs is something new built alongside it.
The question of where this institution should sit has a clear answer—one that field research for a forthcoming book confirmed independently across multiple conversations. The Prime Minister’s Office maintains a level of project discipline—reviewing progress, holding timelines, and creating the institutional expectation of delivery—that does not consistently exist elsewhere in the system.
On 13 February 2026, Prime Minister Modi inaugurated Seva Teertha—the new integrated complex housing the PMO, the Cabinet Secretariat, and the National Security Council Secretariat. In his remarks, he drew a deliberate contrast between the colonial administrative architecture of South Block, built to reflect the thinking of Britain’s monarch, and the new complex built to fulfil the aspirations of 140 crore citizens. It was a statement about institutional purpose.
If Seva Teertha represents a genuine break from the governance architecture India inherited, then the institution of technocratic innovation leadership should be born there. In the corridors of Seva Teertha, with the proximity to power and the cross-ministry reach that Israel’s directorate has demonstrated and that DARPA’s programme managers have always understood to be non-negotiable.
In practice: No more than 100 to 150 people, with the majority of senior staff recruited from frontier AI labs, research universities, and deep-tech environments—both within India and internationally—on fixed multi-year terms. The number is deliberate—large enough to cover the breadth of India’s frontier technology agenda, small enough to maintain the culture of technical judgement that makes the institution worth building. DARPA operates with approximately 220 people total. Israel’s AI Directorate was built with a fraction of that. Scale is not the point. DNA is.
The institution should carry two explicit mandates in equal standing. The first is technical direction: Functioning as the government’s internal conscience on AI—evaluating what compute investments are actually buying, assessing whether foundational model programs are on credible trajectories, and having the authority to redirect spending when the evidence demands it.
The second mandate is talent: Serving as the visible organisational signal that draws India’s diaspora researchers back—not primarily through financial incentives, but through the credibility of the mission and the quality of the people already in the room. A senior Indian-origin researcher at a frontier lab is asking one question when they consider returning: Is there an organisation here whose technical judgement I respect, and that would know what to do with mine? For most, today, the honest answer is no. The right institution, built on the right DNA, changes that answer immediately.
Also read: Why India’s science and research ecosystem is failing its global ambitions
The call that needs to be made
In the 1960s, as Korea embarked on its transformation into a technology-driven economy, President Park Chung-hee did something that surprised the world: He wrote personal letters to Korean scientists and engineers abroad, urging them to return and build the nation. It was not a policy announcement. It was a direct, personal signal from the apex of power that Korea was serious—that the people who came back would find an institution worthy of their return, and a mission worthy of their talent. They came. And Korea built what it built.
India has the scale, the diaspora, and the intellectual capital. It has a new seat of power in Seva Teertha that bears the inscription Nagrik Devo Bhava—the citizen is akin to god. It has a Prime Minister who has shown he understands the language of national ambition.
What India needs now is the institution that makes the ambition real—and the call that tells its best minds, scattered across the world’s frontier labs, that there is finally something worth coming home for. That call would be a fitting first act for India’s next chapter in innovation.
Vasu Eda is the author of Get Job Ready (Penguin Random House) and founder of the Eda Institute. He is working on a forthcoming book, Innovation and the Future of India. He is on X @VasuEda.
(Edited by Theres Sudeep)