Opinion | India Doesn't Have A Jobs Problem. It Has A Capital-Allocation Problem

Capital and jobs have been inextricably linked; they moved together almost mechanically, so much so that policies assume the link automatically. A rupee, a dollar, a yuan put into a new factory bought a company not just output — but it hired workers to run the machines. This relationship held for so long, across economies, that it is an unstated assumption in every industrial policy: build manufacturing, and employment follows. Hence, incentives, in India, are linked to capital investment; central bank governors talk about capital formation as a crucial indicator of growth. In the last few decades, technology has broken the assumption and relationship. But policymakers still maintain that capital investment will create jobs.

Past waves of automation

were largely complementary to labour — a machine let a worker produce more, but it needed the worker. The current wave, first robotics and now AI robotics, is substitutive. An AI robot replaces the worker, completely. An AI robot is not a linear or mechanical robot; this is an observation alert robot capable of multi-tasking and multiple movements. It can sort the packages, and it can even do packaging, something only dexterous human hands could do. The addition of artificial intelligence capability to robots is changing not just the shop floor but also several industries.

The impact is a structural change in what a capital investment actually creates in an economy. Output rises with investment; capital creates profits for corporations, and even GDP rises—while the number of jobs per rupee invested keeps falling, quietly, underneath it all. The old measures are decoupling, and in some cases, reversing.

This shift is no longer only happening in developed economies. It is visible in hard data across Indian manufacturing too, and it is now spreading into services through AI — reshaping how economies will need to plan for growth and employment as two increasingly separate problems, not one.

The pattern was set in the US: manufacturing employment fell by 2.4 million jobs (17%) between 2002 and 2022; the economy grew, and capital kept flowing into new plants — because the plants being built were automated by design. Nobel-adjacent research (Acemoglu and Restrepo, Journal of Political Economy, 2020) put a number on the mechanism: every additional robot per thousand workers cuts local employment by 0.2 percentage points, independent of trade or offshoring to China.

In India, manufacturing’s share of employment is stuck near 12% since 1991. Capital stock grew by 74% over the last decade, while employment grew by only 36%. Labour’s share of manufacturing value added nearly halved, from 22.2% to 14.3%, between 2000–01 and 2011–12, even as output rose. India’s own Production-Linked Incentive scheme makes the mechanism visible at the scheme level; over ₹2.40 lakh crore invested across 14 sectors has generated 14.15 lakh direct and indirect jobs as of March 2026, roughly ₹17 lakh of capital invested for every job — for work that, on average, pays not much more than ₹4 lakh a year.

The ratio is not uniform: electronics is far more capital-intensive per job than food processing, where a much smaller outlay of ₹9,207 crore still generated 3.35 lakh jobs, beating its target. Same scheme, same incentive structure, opposite capital-to-job ratios — depending entirely on how automatable the sector is.

The reason is visible in hard automation data: global robot installations have more than doubled over the past decade to 542,000 units in 2024, with India as the fastest-growing major market — 9,100 robots installed in 2024, overtaking Germany.

The same is happening in services faster than it hit factories: a Stanford Digital Economy Lab study using actual payroll data in the US found a 13% relative employment decline among 22–25-year-olds in the most AI-exposed occupations. Challenger, Gray & Christmas data shows AI has been the leading stated reason for US job cuts for five straight months through mid-2026. Indian IT firms show early signs of the same entry-level slowdown, though the data isn’t yet conclusive.

The AI and automation waves are converging: Nvidia’s “physical AI” models and SAP’s Joule agent platform are now being built directly into industrial robots, so the old assumption — that services jobs are safer because they aren’t physical — may not hold much longer. Robots are now embedded with AI capabilities allowing them to observe, react and perform in real life, not just preprogrammed movements.

This policy conversation needs to move — to designing where the capital should go. India’s PLI data shows, with more precision than most policy documents manage, exactly which kinds of capital deployment still create jobs and which don’t. That evidence should now drive three concrete shifts in how India allocates incentives, builds AI infrastructure, and measures success.

First, deliberately redirect incentive design towards labour-absorbing categories. PLI’s own data already prove this works: food processing, tourism, restaurants, and other labour-intensive services. A fraction of electronics’ capital outlay into these sectors created proportionally far more jobs.

The next generation of manufacturing and services incentives should be designed around this evidence, not around headline investment totals. A scheme that attracts ₹10,000 crore and creates 3 lakh jobs is doing more for India’s employment target than one that attracts ₹2 lakh crore and creates 1 lakh. Policymakers should start publishing capital-per-job ratios for every incentive scheme, sector by sector. The way PLI’s own data now — and use that number, not just investment totals, as an incentive metric in scheme design.

Second, enable AI-enabled MSMEs and one-to ten-person companies as India’s real jobs multiplier. AI sharply lowers the cost of sales, accounting, design, compliance, and customer service — the overhead that has historically kept small Indian businesses small. Capital aimed at this segment behaves like the food-processing case: it multiplies the number of viable, labour-absorbing units rather than concentrating output inside a handful of automated firms. India has run this multiplier before — NASSCOM-era IT income gains built a new middle class whose spending, in turn, built retail investment, EMI-financed durables, and instalment-financed housing. Reproducing that deliberately means building Indian AI infrastructure — affordable compute, open models, Indian-language datasets — for Indian costs and scale, rather than relying on automation defaults imported from high-wage economies. It also means favouring thousands of small teams iterating in parallel — closer to Ukraine’s distributed drone-innovation model — over two or three national AI champions, since concentration at the top of the stack tends to reproduce concentration in who captures the gains.

Third, change what gets measured. GDP growth is exactly the metric that kept rising over two decades during which the capital-jobs relationship was quietly breaking — it is not a reliable proxy for employment health, and India should stop treating it as one specifically for AI-era policy. District-level data on new businesses formed, jobs created, incomes rising, and — critically — whether those businesses survive and grow rather than merely launch, would tell policymakers and investors far more than aggregate output ever will. This is a measurement-infrastructure investment, not a slogan: it requires the same kind of disclosure discipline the PLI scheme has already shown is possible, extended systematically rather than accidentally.

None of this reverses the underlying decoupling of capital from jobs — no incentive design fully restores a relationship automation economics has structurally broken. What it can do is stop pretending the relationship still holds, and build India’s next round of industrial and AI policy around the sectors and business structures where a rupee of capital still reliably buys a job. The old question was how much capital India could attract. The one that matters now is what kind of jobs, if any, that capital still buys.

Harish Mehta is co-founder of Nasscom and a founder member of the Hundred Million Jobs Mission (HMJM). K Yatish Rajawat is founder of the Centre for Innovation in Public Policy (CIPP) and a founder member of HMJM (www.hundredmillionjobs.org). Views expressed in the above piece are personal and solely those of the writers. They do not necessarily reflect News18’s views.