The PLI Scheme Is Building Factories. Here's Why It Won't Build the Next Infosys.

India announced ₹1.96 lakh crore in PLI incentives to build world-class factories. The same year, every major PLI sector spent less on R&D than a mid-sized Korean firm does in one quarter. The factories are going up. The ideas aren't.

This isn't a critique of industrial policy. It's a critique of this industrial policy, and of the quiet assumption embedded in its design, that if you build enough capacity, capability will follow.

It won't. And the historical evidence is clear about why.

What PLI Actually Is & What It's Optimising For

The Production Linked Incentive scheme is, at its core, an output subsidy. Firms in 14 approved sectors, from semiconductors and mobile phones to pharmaceuticals and textiles, receive cash payouts tagged to incremental production above a defined base year. Exceed the threshold, collect the cheque. The eligibility criteria are structured around committed capital expenditure, revenue scale, and compliance with production targets.

Notice what's absent. There is no PLI payout for a patent filed. No incentive tranche unlocked by an R&D-to-revenue ratio crossing a threshold. No bonus for hiring a PhD. The DPIIT scheme guidelines are explicit, the trigger is production volume, not knowledge creation.

This matters because incentive design is theory of change made concrete. When a government subsidises output, it is betting that more output will solve the problem. That bet is reasonable if the binding constraint is capital, if Indian manufacturers simply lack the money to build the plant. It becomes a bad bet if the constraint is something else: know-how, human capital, patent pipelines, the organisational capability to move up the value chain. PLI assumes the first. The data suggests the second.

Raghuram Rajan and Rohit Lamba, in Breaking the Mould (2023), made a related observation: Indian industrial policy has historically been better at selecting occupants than at creating challengers. PLI's structure, large upfront capex commitments, multi-year compliance windows, complex documentation requirements, is not designed for the firm that doesn't exist yet. It's designed for the firm that can write a ₹1,000-crore cheque.

Indian mobile assembly line at an electronics manufacturing factory

The R&D Gap: What the Numbers Actually Show

India spends approximately 0.65% of GDP on gross research and development. China spends 2.4%. South Korea spends 4.9%. Israel spends 6.35%.

These aren't marginal differences. They represent a structural divergence in what each economy has decided to build.

The gap becomes sharper when you drill into the sectors PLI is targeting. India's pharmaceutical sector is the world's pharmacy by volume, the third-largest producer globally by quantity. Yet when you open the annual reports of PLI beneficiaries and compare the R&D line to the capex line, you find companies that invest heavily in production facilities and comparatively little in the molecular science that creates the next drug. Sun Pharma, one of India's most R&D-intensive pharmaceutical companies, still spends a far smaller share of revenue on research than Pfizer, AstraZeneca, or even mid-tier Korean peers like Hanmi Pharmaceutical. In electronics, the gap is wider still: Dixon Technologies, a PLI beneficiary building mobile phone components, is primarily an assembler. Its value addition is in manufacturing efficiency, not in the design IP that makes a chip worth making in the first place.

The DST's own R&D Statistics & Indicators report reveals another uncomfortable fact: the public-private split in Indian R&D is the reverse of what you'd want. The government does most of the research. The private sector does comparatively little. In South Korea, the corporate sector drives nearly 80% of total R&D. In India, private sector R&D lags far behind public scientific institutions, which means the knowledge being generated is not, by and large, being generated inside the firms that are now receiving PLI payouts.

ICRIER's working paper tracking PLI across sectors reaches a similar conclusion: the scheme has driven capex announcements, and in some sectors real production growth, but there is minimal evidence of corresponding investment in the capabilities that would allow India to move from contract manufacturing to original design.

PLI, in other words, is moving the capex number. It is not moving the R&D number. And the R&D number is the one that determines where India sits in the value chain in twenty years.

Two Models, Two Outcomes: Korea's Factories vs. Taiwan's Ideas

The most instructive comparison isn't India versus the United States. It's South Korea versus Taiwan, two East Asian developmental states that took different bets on how to build an industrial economy, and got very different results.

South Korea chose the chaebol model. State capital, directed credit, and preferential industrial policy flowed to large conglomerates, Samsung, Hyundai, LG, that were expected to scale rapidly, meet export targets, and eventually develop their own technology. The model worked, in the narrow sense that Korea industrialised with remarkable speed. Alice Amsden's Asia's Next Giant (1989) remains the definitive account of how this happened: the Korean state extracted performance in exchange for subsidy, and the chaebols delivered production.

But the chaebol model created its own trap. Large firms that scale on the back of government support develop structural dependencies. They are good at doing more of the same. They are less good at cannibalising themselves, at making the radical bets that create genuinely new categories. The chaebols built factories. They became world-class manufacturers. They did not become the intellectual engine of the global technology economy.

Taiwan took a different path. Dan Breznitz's Innovation and the State (2007) is the essential text here. Taiwan's industrial policy was decentralised, SME-driven, and focused on what Breznitz calls "second-generation innovation", not the breakthrough invention, but the rapid iteration, process improvement, and manufacturing excellence that turns a prototype into a product the world can buy. The result was an ecosystem of competitive, specialised firms rather than a small number of dominant conglomerates. And from that ecosystem came TSMC.

TSMC's origin story is worth dwelling on. Morris Chang deliberately built it outside the vertical integration logic of the chaebol, as a pure-play foundry, it had no product of its own to protect, no downstream business to cannibalise. Its competitive advantage was pure manufacturing science: the ability to fabricate chips designed by others with more precision, at lower defect rates, at scale, than anyone else on earth. That capability was built through decades of accumulated process know-how, not through output subsidies.

India is now courting TSMC and Micron for its semiconductor PLI. The irony is that the PLI structure optimises for outputs, how many chips come out of the factory — rather than for the process knowledge accumulation that makes TSMC, TSMC. You can build TSMC's factory in India. You cannot import TSMC's thirty years of process learning.

Mariana Mazzucato's framing in The Entrepreneurial State (2013) is useful here: the distinction between mission-oriented public investment, funding that builds new capability and pushes the technological frontier, and passive subsidy, which rewards whoever can meet a production threshold. PLI is clearly the second. It is not funding India's equivalent of DARPA. It is funding factories.

The Korea-Taiwan comparison makes the stakes concrete. Korea is still an industrial powerhouse, but its growth in recent decades has come increasingly from a small number of firms (Samsung semiconductor, above all) rather than from a broad, competitive innovation ecosystem. Taiwan built the ecosystem. The question for India isn't which country it resembles more. It's which country it wants to resemble in forty years.

semiconductor manufacturing cleanroom silicon wafer fabrication

Why Infosys Won't Come from a PLI Sector

Infosys was founded with ₹10,000 in seed capital. It did not receive a production subsidy. It operated in a sector the government largely left alone, in a policy environment that was not designed to help it, and in many respects was actively indifferent to it. The same is true of Wipro's technology pivot, of TCS, of the entire ecosystem that made India a global software services powerhouse.

India's IT sector is its most successful industrial policy story, and the government didn't write the policy.

PLI's selection criteria work against the next Infosys emerging from the sectors it covers. To qualify for PLI, a firm must commit large capex, demonstrate it can meet volume targets, and manage compliance over a multi-year window. These are the capabilities of an incumbent or a well-capitalised conglomerate. They are not the capabilities of the firm that is going to invent the semiconductor architecture that displaces today's leading designs.

Dani Rodrik, in "Industrial Policy for the Twenty-First Century" (2004), draws the critical distinction between policies that build capabilities, human capital, knowledge spillovers, organisational learning, and policies that reward incumbents. The latter can be justified on strategic or employment grounds, but it should not be confused with the former. PLI is a reward for incumbents. It is not a capability-building programme.

Here a pushback needs to be pre-empted. The obvious objection is that IT services and semiconductor manufacturing are categorically different industries, one is capital-light, knowledge-intensive, and born global; the other requires billions in capital expenditure before the first chip is made. The comparison isn't quite fair. But the relevant claim isn't that PLI should behave like a software incubator. The claim is narrower and more precise: within manufacturing sectors, PLI's eligibility criteria systematically disadvantage the asset-light, R&D-intensive firm in favour of the firm that can commit capital. An Indian materials science startup with a genuine process innovation that could reduce semiconductor defect rates has no PLI pathway, because it can't meet the capex threshold. The large contract manufacturer that builds to spec does. The incentive structure is selecting for a specific type of firm, and it is not the type that builds new knowledge.

What would ₹2 lakh crore in R&D grants, patent incentives, and university-industry partnerships have produced instead? We'll never know for certain. But we know what the Indian Patent Office data shows: in PLI sectors, resident patent filings (Indian companies filing in India) are a small fraction of total filings. Non-resident filings dominate, meaning the IP in these industries, even as Indian manufacturing scales, is largely owned abroad. You can manufacture a product and not own the idea behind it. PLI is funding the former, while the latter walks out the door.

What a Better Design Looks Like

The argument here is not against industrial policy. It is against this design of industrial policy.

South Korea's own post-1990s reform trajectory offers the corrective. After the chaebol system generated the 1997 Asian financial crisis, a crisis partly caused by over-leveraged conglomerates with weak innovation pipelines, Korea restructured toward R&D tax credits tied to actual research expenditure, university-industry linkage funding, and innovation vouchers for SMEs. The policy still involved state direction, but the incentive was now pointed at knowledge creation rather than output volume.

Philippe Aghion and colleagues, in their AEJ: Macroeconomics paper "Industrial Policy and Competition" (2015), provide the empirical grounding: industrial policy works when it increases competition and capability simultaneously. When it entrenches incumbents and reduces competitive pressure, it crowds out innovation. PLI, by designing for large incumbents with complex compliance requirements, risks exactly the latter.

India already has the ingredients for a different model. ISRO built world-class satellite technology on tight budgets through accumulated mission-by-mission learning, mission-oriented investment of the kind Mazzucato describes. IIT research output has improved markedly in the past decade. NASSCOM data shows the deep-tech startup ecosystem is growing, and Startup India funding is flowing into categories, AI, biotech, space, that will define the next generation of the global economy. These are not PLI sectors.

The plumbing for a different model exists: R&D tax credits calibrated to patents and publications rather than capex, innovation vouchers that let SMEs access university research infrastructure, and a public procurement policy that explicitly rewards firms with resident IP. NITI Aayog's own technology roadmap documents gesture toward these mechanisms. The gap between what the government's vision documents say and what PLI actually delivers is, itself, a form of evidence about where the political incentives lie.

The Uncomfortable Conclusion

Industrial policy is not neutral. Every subsidy selects for something, a type of firm, a type of capability, a theory of where value comes from. PLI's theory is that India's problem is insufficient production capacity, and that if you subsidise enough factories, the rest will follow.

The evidence from a century of industrial development suggests this theory is wrong. The Korea-Taiwan divergence shows it. The India-China R&D gap shows it. The fact that India's most globally competitive industries grew without output subsidies shows it.

What makes an economy durable isn't the hardware, the factories, the capacity, the export volumes. It's the software: the R&D culture, the patent pipelines, the knowledge firms, the human capital that keeps compounding. PLI is building the hardware. Nobody is building the software.

The factories are going up. The ideas aren't.