The Next Fifteen Years

A forecast built from first principles
Section future / 03-domains / contested / geopolitics / india.md

India - the exposed sector is the growth model#


Contents

India is the clearest case in the document of a country whose development strategy and AI exposure are the same thing, and it is worth treating separately because the arithmetic does not resolve itself.

The arithmetic#

The BPO/IT-services rung was the ladder. Game 5's signal collapse and Game 3's competitive dissipation both apply directly: the work is cognitive, remote, English-language, process-defined, and delivered against a service-level agreement - which is to say it has cheap, immediate ground truth and is therefore in the fastest-compressing category in Part I.

The industry's own defense is that it moves up the value chain into consulting and systems integration. That is the standard answer and it has worked before. The problem this time is arithmetic rather than capability: the higher-value tiers employ a fraction of the people the lower tiers do. Moving up the chain is a strategy for the firms and not for the cohort.

What actually happens to the sector#

The distinction that matters is between the firms and the employment.

Indian IT services revenue continues growing through 2032 while sector headcount flattens or declines. The decoupling of revenue from employment is the thing to watch, and it is publicly reported quarterly by listed firms - one of the cleanest available tests of Game 4 anywhere.

Failure mode for that prediction, and it is a real one: it assumes the productivity gain is captured by the vendor rather than competed away to the client. Game 3 argues that a capability everyone can buy dissipates into consumer surplus, and the buyers here are sophisticated procurement organizations who know exactly what a delivery model costs and will reprice fixed-price contracts at renewal. If that repricing runs faster than volume growth, revenue flattens alongside headcount and the firms are not fine either - the sector then looks less like margin expansion and more like the classic outcome for an industry whose core input has been commoditized. The observable that separates the two worlds is revenue per employee against contract value per delivered outcome; the first rising while the second falls is the dissipation case.

Where the cohort actually goes#

The prediction above is about a sector. The macroeconomic question is about the ~10M annual entrants, and the honest answer is that the sectors with capacity to absorb them are the low-productivity ones: construction, retail, domestic and care services, informal logistics, and agriculture that the growth model was supposed to be moving people out of. Absorption in that sense is not reassuring, because employment and productivity come apart exactly the way welfare and growth come apart in the Global South page. A cohort that finds work at half the productivity of the job it was trained for is fully employed and the country's convergence path is still broken.

Two second-order effects follow, and neither is priced in the usual discussion.

The consumption channel. Formal-sector IT wages anchor urban middle-class demand in a small number of metros, and they anchor a great deal else: mortgage underwriting, private schooling, discretionary retail, and the aspiration that keeps engineering enrollment high. A pyramid that stops hiring at the base does not just remove jobs, it removes the credible expectation of those jobs, and the expectation is what the education spending and household borrowing were priced against. The adjustment shows up in enrollment and in urban credit quality before it shows up in employment statistics.

The fiscal channel. The sector is a disproportionate contributor of formal-sector tax and foreign exchange, and India has no unemployment-insurance system capable of carrying a cohort-scale shock; the existing rural employment guarantee is a floor for subsistence, not a bridge for displaced graduates. So the state faces rising claims on a narrowing base at the same moment it would need to fund the energy and skills build-out that the pivot requires. This is the general shape of the problem in fiscal form, arriving earlier here than in rich countries because the safety net is thinner and the exposure is more concentrated.

The three real assets#

India's position is not weak. It is mismatched - strong on inputs that matter later, exposed on the one that matters now.

  1. Domestic market scale. A market large enough to sustain domestic AI companies serving domestic demand is the pivot the hub page prescribes, and India is one of very few countries that can attempt it. Digital public infrastructure - identity, payments, data exchange - is genuine, unusual, and directly relevant: it is state-owned distribution, an inelastic complement most countries lack entirely.
  2. Talent depth. The engineering pipeline is real and large. But note what Inference economics implies: talent that builds on frontier models holds a wasting asset. Talent that builds the inelastic complements does not.
  3. Non-alignment as a position. Able to buy from both blocs, and increasingly courted by both. That is worth more in a partially-bifurcated world than in a fully decoupled one.

The binding constraints#

Why this case generalizes#

India is the largest instance of a pattern covering the Philippines, Bangladesh, Kenya, Vietnam, and much of Eastern Europe: countries whose growth strategy was to sell cognitive labor across borders at a discount.

That trade was the single most effective poverty-reduction mechanism of the last thirty years outside Chinese manufacturing. It is being compressed by a technology none of these countries control, on a timescale shorter than the one their institutions move on.

This is the largest single welfare consequence in the document and it receives almost no attention - the discourse on AI and labor is conducted almost entirely about workers in rich countries, who are better insured, better capitalized, and better represented than the workers who are actually most exposed. → The Global South

What would change the path#

Three escapes, none free. (1) Domestic demand for cognitive services grows fast enough that export-ladder compression is a terms-of-trade hit rather than an employment crisis - requires middle-class income growth that demography and politics may not deliver on this timeline. (2) Physical-economy upgrading (logistics, manufacturing with robots, energy) absorbs labor the export services sector sheds - runs into the power and manufacturing gaps above, and into robotics supply chain geography that does not favor late entrants. (3) Digital public infrastructure as an export - India selling DPI stacks and identity rails the way it once sold IT services - a real niche, too small alone to replace the BPO wage bill. The base case remains: compression of the cognitive export ladder is the central Indian AI story, and it is a welfare story before it is a geopolitics story.

Failure mode of this page: treating every IT layoff in Bangalore as confirmation. Cyclical US tech demand and structural AI substitution are different; the discriminating test is whether junior and mid-tier export services employment recovers with the US cycle or decouples from it. → B1 logic applied to trade data.

Revenue/employment decoupling in the export sector is the early-warning series governments should watch and usually don't - still growing invoices with thinning junior headcount is the ladder failing quietly. → geopolitics hub


Related: Game 4 - Labor · The Global South · Software · Uncertainty 3 · Energy

View markdown source

select · Enter open · Esc close