The Machine India didn't build
How a services superpower confronts a world run on compute, not people
Preface
This essay is about a change most people can sense but cannot yet name.
For thirty years, India’s rise has rested on a simple truth: the world had more cognitive work than it had people to do it. India filled that gap. Its programmers, analysts, operators, and support staff became the hidden infrastructure of globalisation. Its digital public platforms proved that the state could execute at continental scale. A services-led nation found a place in a world that rewarded skills more than machines.
The ground under that world is now shifting.
What made India indispensable is becoming the most automatable category of labour. Workflows that once travelled across time zones are being pulled into data centres. The cognitive processes millions of Indians perform every day are being replicated next to GPUs, priced in electricity and latency rather than wages. The export India built its ascent on—skilled human mental work—is no longer scarce once cognition runs on compute, not people.
This is not a story of decline.
It is a story of the global order changing its substrate.
India still has options. But they are no longer the options of the 1990s. They are the options of a machine century, in which leverage comes from energy, compute, and industrial depth—not from cheap cognition or large talent pools.
What follows is not a prediction. It is a sober account of what happens when cognition becomes industrial, and what that means for a civilisation that rose on the old kind.
I. The Half-Body at the Edge of the Machine Century
India’s rise is real, but its architecture is unusual. Most large countries industrialised—steel, power, ports, machines. India scaled through cognition. It grew by processing information the world could not process on its own.
If a corporation had a workflow it could not perform cheaply or quickly enough at home, India became the extension of that workflow. GCCs multiplied. IT services matured into a national capability. A middle class formed around the capacity to execute global mental work at scale.
But this created a structural imbalance.
India became a global head without a global body.
It built the coordination layer on top of an industrial layer it never built.
It exported attention, not machines.
It supplied cognition because the world had more cognitive work than it had humans.
That scarcity is evaporating.
The global system is shifting from a world organised around human cognition to one organised around machine cognition. When the organising principle of an economy changes, every participant’s position changes with it.
India rose when cognition depended on people.
It now stands at the edge of a world where cognition depends on compute.
This is the hinge.
Everything that follows comes from it.
II. Postcolonial Modernity Without Machines
India did not begin as a software nation. Its early ambitions were hardware: analogue computers at TIFR, mainframes at ECIL, indigenous control systems. These were real attempts to build machines.
Two forces stopped them.
First, Cold War controls.
Advanced components—especially semiconductors—were restricted. India could design systems, but it could not source what it needed to produce them reliably. Each technological generation widened the gap.
Second, postcolonial governance logic.
The new state trusted planning more than factories. It valued administrative legitimacy over industrial recursion. Research institutions grew, but the machinery they needed did not. India decolonised its institutions, but not its industrial metabolism. The result was a distinctive kind of modernity: strong laboratories, weak factories; strong imagination, limited machinery.
By the 1970s, the trajectory was set:
a nation of engineers without a nation of machines.
This mattered less when globalisation rewarded services.
It matters now that sovereignty is drifting back toward the physical.
India entered the services era with a missing foundation.
It worked because the world did not require that foundation.
It requires it now.
III. The Services Detour
Liberalisation arrived just as the world had an oversupply of routine cognitive work and a shortage of people to execute it. India had talent, English, wages, and a willingness to reorganise workflows. Telecom costs collapsed. Corporations shifted operations across borders.
A clean development model emerged:
India became the global processor of mental work.
Outsourcing → offshoring → IT services → GCCs.
Millions of Indians performed the cognitive tasks global firms could not perform as cheaply at home: code maintenance, documentation, quality assurance, financial reconciliation, call routing, compliance, operations.
Digital Public Infrastructure reinforced the pattern.
Aadhaar and UPI proved India could run billion-person systems with precision.
The model rested on one premise:
human cognition was the bottleneck in global operations.
For decades, this was true.
Processes could only move as fast as people could interpret and execute them.
Then the bottleneck began to dissolve.
AI systems collapsed the cost of cognition.
Inference began to replace interpretation. AI does not speed up human workflows; it bypasses them.
Latency began to replace wages as the decisive constraint.
Tasks that once required thousands of workers could now execute inside machines.
India didn’t miscalculate.
The substrate it was standing on changed.
IV. The Asian Split
While India optimised for services and coordination, East Asia built a different substrate.
China, Korea, Japan, Singapore, parts of SE Asia and now the Gulf invested in:
grids
ports
robotics
batteries
compute clusters
semiconductor fabs
tightly integrated supply chains
They built environments where machines reproduce machines.
Industrial density compounded.
Capacity created more capacity.
India built digital coherence on top of hardware it did not control.
The contrast is blunt:
East Asia built the substrate. The countries building industrial depth are the ones building the homes where machine cognition will actually run.
India built the interface.
Interfaces coordinate; substrates compound.
Interfaces make a country legible; substrates make a country sovereign.
This divergence is not moral or cultural.
It is metabolic.
Both strategies worked in their era.
Only one aligns with a century organised around machines.
V. India in the Reordered World-System
The global system is reorganising around two physical blocs:
A services–financial bloc anchored by the United States,
where software and human talent once dominated.
An industrial–material bloc anchored by China,
where energy, machines, and supply chains concentrate.
India is stretched between them.
It earns in the first—through exports of cognition and code.
It spends in the second—importing energy, electronics, components, materials.
This is the currency–metabolism trap:
India’s income structure and dependency structure belong to different systems.
What changes now is not ideology but physics.
As machine labour rises, companies optimise for compute adjacency, not labour arbitrage. The decisive question becomes:
Where is the power? Where are the GPUs? Where can workflows run at the lowest latency?
The Satya Nadella interview makes this explicit. The future firm provisions virtual computers for agents, not employees. AI workers need identity, storage, audit trails, and observability pipelines. They run inside VMs next to data centres because every millisecond matters. Work moves to where compute resides—not to where humans live. The moment workflows run on electricity rather than wages, every advantage India built for the old world begins to leak out.
Dwarkesh and Dylan interview Satya – How Microsoft thinks about AGI … go here when they talk about agents, hybrid human-agent teams, then full agents.
The Fairwater campuses—multi-gigawatt, continuously refactored industrial plants—are built for machine workers. The product is cognition.
India’s position becomes exposed not because services vanish, but because the determinants of power move down the stack—to grids, substations, fabs, cooling, and compute.
India remains important.
But importance and leverage diverge.
VI. After the Detour: India and the Rise of Machine Labour
India still has a future—just not the one it planned for. The services boom created mobility and embedded Indian talent inside the world’s operations. DPI remains a major strategic asset.
But a civilisation that rose on exporting cognition now faces a world where cognition is industrial output.
AI systems do the work India exports: testing, documentation, compliance, triage, support, analysis. They do it near compute, not near labour. They execute workflows inside agent VMs with their own identity, storage, security, and logs. They run at the speed of electricity.
The future of Indian labour will be decided inside data centres India did not build.
When labour becomes a function of power and GPUs, the geography of work collapses into the geography of compute. The labour market becomes a power market.
This produces a structural reversal.
India’s most valuable export becomes the easiest to automate.
The GCC model hollows not because companies leave India but because tasks stop needing to migrate to humans at all. Workflows move inward, toward AI campuses, not outward to low-wage labour pools.
A civilisation that mastered the export of cognition now confronts the limits of that strategy:
the head can no longer rent itself out.
And the body—power, factories, fabs, logistics—is still incomplete.
But this is not an ending.
Machine labour still needs governance, supervision, integration, and physical infrastructure. If India builds the substrate it deferred—grids, semiconductors, automation, sovereign compute—it can become a major deployment hub for machine labour.
That requires breaking with seventy years of habit.
Infrastructure must become strategy, not background.
Energy must become political power, not utility.
Machines must become central, not peripheral.
India can remain the world’s nervous system—skilled, global, weightless.
Or it can build the body it never built.
The window is open.
But it is not widening.
Conclusion
India is not running out of talent.
It is running into a world where talent is no longer the organising principle.
That is the shift.
Whether India remains powerful in the next era depends on what it builds, not what it already knows. Countries that anchor intelligence in physical infrastructure will set the terms of the machine century. Countries that depend on the infrastructure of others will adapt to those terms.
India’s past explains how it arrived here.
Its future depends on whether it chooses to build what it deferred.
The window has not closed.
But it has stopped expanding.
FAQ + Objections
1. “Isn’t this just another automation scare?”
No.
This is not about machines replacing workers.
It is about where cognition runs in the global system.
For thirty years, cognition ran on people, which meant it ran where people were.
Now cognition runs on compute, which means it runs where compute is.
This is not automation.
This is a change in the substrate of labour.
2. “Won’t India benefit because it has so many engineers?”
India will always have large talent pools.
The issue is not talent; it is adjacency.
Machine labour runs:
next to GPUs
next to power
next to high-bandwidth data
next to data-centre-grade cooling
next to sovereign AI infrastructure
India’s engineering talent matters only if it is co-located with these assets.
If not, the work will run elsewhere.
3. “Isn’t India becoming a manufacturing hub anyway?”
India is expanding assembly, batteries, and some component manufacturing.
But the machine century requires:
continuous industrial recursion
reliable grid abundance
fabrication capacity
automation-heavy supply chains
mid-tier robotics
proximity to compute corridors
India is not yet building at that depth.
The gap is not direction — it’s tempo and scale.
4. “But India has DPI. Isn’t that an advantage?”
It is.
But DPI is an interface, not a substrate.
DPI improves coordination, identity, payments, and service delivery.
It does not:
produce energy
produce compute
produce machines
shorten industrial supply chains
increase physical redundancy
reduce latency for AI workloads
DPI matters — but it sits on top of hardware India does not control.
5. “Why can’t India simply skill up for higher-value cognitive work?”
Because the bottleneck is no longer human cognition.
AI systems perform higher-value work faster and cheaper near compute.
The world’s limiting factor is:
power
cooling
chips
rack density
data-centre geography
Not skill formation.
Training more workers does not solve a constraint that has become physical, not educational.
6. “Can’t AI just augment Indian workers rather than replace them?”
There will be augmentation.
But the firm’s core question changes from:
“Where can I hire skilled people?”
to:
“Where can I run this workflow with the lowest latency and cost?”
The substrate changes.
The optimisation strategy changes with it.
Augmentation does not prevent recentralisation of labour around compute.
7. “Isn’t India too big to fail?”
Size guarantees relevance.
It does not guarantee leverage.
China’s leverage comes from industrial capacity.
The US’s leverage comes from capital and trust.
India’s leverage came from cognition.
When the basis of leverage changes,
scale is not enough.
8. “What is the fastest way for India to adapt?”
There are three:
1. Build energy abundance
Machine labour runs on power, not wages.
2. Build compute adjacency
Machine labour stays where the GPUs are.
3. Build industrial depth
A country cannot govern machine labour if it cannot host machine cognition.
Everything else — skills, policy, incentives — is secondary to these three.
9. “Will machine labour fully replace India’s services exports?”
No.
But it will compress them:
fewer people for each workflow
narrower categories of human-heavy work
increased sensitivity to latency and reliability
more demand for machine-adjacent operations
India keeps services.
It loses the strategic leverage of services.
10. “Could India become a machine-labour hub?”
Yes — but only if it builds:
reliable substations
sovereign compute cores
AI-ready industrial parks
local agent-governance infrastructure
power corridors
chip capacity (front-end or back-end)
Machine labour needs a physical home.
Services-led India never needed one.
Machine-labour India will.
11. “Is this essay anti-India?”
No.
It is anti-nostalgia.
It is anti-1990s mental models.
It is anti-services exceptionalism.
It is pro-reality.
India’s strength was built in a world configured around people.
The world is now configuring itself around machines.
Seeing that clearly is not pessimism.
It is sovereignty.
Dwaipayan Banerjee, Computing in the Age of Decolonization. (2026)
Arun Mohan Sukumar, Midnight’s Machines.
Raghuram Rajan & Rohit Lamba, Breaking the Mould.
Asianometry, “India’s Tata Group and Semiconductors.”
Satya Nadella in conversation with Dwarkesh Patel, 2025.
If you are going to have an LLM write the article, ask it to be less repetitive. This could have been 6 paragraphs and would have been much better.
It’s a solid topic for sure
You must understand that although AGI is possible, it needs much more energy than an average human.
Yes, it is possible to build more computers regardless, temporarily. But this kind of thing introduces a lot of redundancy. Economics follows supply and demand, not 'everything AI'. You might be getting chatgpt for free, and may even use it for every task. But this is not scalable. AI companies need to make a profit somewhere.
And no, this is not the reason for 'hatred towards Indians'. The people making AI and the haters are on the opposite sides politically.