New Delhi’s Counter‑Narrative: Jobs, Built from Silicon Up
On Monday in a packed hall in New Delhi, India’s top tech policymaker did something unfashionable: he refused to treat jobs as the inevitable casualty of artificial intelligence. S. Krishnan, Secretary at the Ministry of Electronics and IT, didn’t merely reassure the audience at the India AI Impact Summit that work would survive. He argued that India intends to manufacture the conditions for more work—by lowering the cost of the inputs that make AI possible and by building the hardware that the rest of the world is scrambling to source.
From PowerPoints to Wafers
The optimism wasn’t abstract. Krishnan tied it to a concrete industrial arc that begins not in software, but in a cleanroom. India’s first commercial‑scale semiconductor production, he said, is set to begin shortly, with Micron’s facility expected to start operations later this month under the India Semiconductor Mission’s slate of approved projects. He went further, signaling a national ambition to pull high‑bandwidth memory into the mix—HBM, the narrow throat through which the entire AI world is currently trying to breathe.
If that happens, it changes the jobs math in a way spreadsheets rarely capture. A domestic HBM and packaging base doesn’t just employ chip engineers; it spins up layers of work across chemicals, gases, ultrapure water systems, tool installation, logistics, compliance, and maintenance. Construction teams break ground years before yield curves asymptote. Technicians and operators follow, then suppliers who feed the line, then design houses and testing labs that co‑locate because the cycle time is shorter when you can walk to the fab. It’s a hierarchy of employment that begins with wafer starts and ends in service exports, and India is trying to insert itself at the start of that chain rather than just at the end.
Compute, Priced Like Infrastructure
Chips are only half the bet. The other half is time on those chips. Krishnan outlined a model that prices AI compute at roughly ₹65 per GPU‑hour for researchers, students, startups, and small businesses. That reads less like a grant and more like the early stages of a utility—the way a country decides that electricity and bandwidth should be cheap enough that new firms appear because the fixed costs are no longer fatal. If you’ve watched what happened when India made payments rails radically accessible, you’ve seen the playbook: compress a foundational cost, flood the market with new entrants, and let unexpected categories of work coalesce around the lowered barrier.
Applied to AI, subsidized compute doesn’t just produce more fine‑tuned models. It produces hiring. Labs need data engineers. Startups need product people who speak to customers in Odia and Tamil, not API docs in English. SMEs with access to experimentation time can reconfigure processes they previously outsourced, turning “users” of AI into employers of people who can wield it. The headline claim—AI as a net job creator—becomes plausible only if the per‑hour price of trying ideas collapses. India is attempting to force that collapse.
Sovereign Stack, Domestic Demand
The throughline is a sovereign AI posture that runs from silicon to services. Build memory and packaging capacity at home; stand up affordable compute; seed a developer base that doesn’t need permission from distant incumbents. That stack does more than cushion against automation in routine tasks. It tilts the global cost curve so that the marginal AI company, the marginal data center, the marginal tooling startup, and the marginal services firm are more likely to be Indian—or at least to hire in India—because the cheapest reliable place to iterate is onshore.
Displacement Will Happen—But So Will Recomposition
None of this erases the reality that routine customer support and back‑office tasks are under pressure. The wager is that recomposition outpaces displacement: semiconductor ecosystems absorb industrial labor; data centers and connectivity projects pull in electricians, HVAC specialists, and grid engineers; AI‑enabled services knit local language, regulation, and domain knowledge into products that don’t travel well if they’re designed an ocean away. It isn’t a one‑for‑one swap. It’s a shift in the topology of work, from task execution to system building and system stewardship.
Execution Is the Entire Story
Ambition does not equal yield. “Shortly” has to turn into wafers that pass test, and HBM aspirations have to survive the unforgiving math of supply chains, geopolitics, water, and power. Subsidized compute must be accessible in practice, not captured by a handful of well‑connected players. If the ₹65 promise surfaces on login screens in tier‑2 and tier‑3 cities, you’ll know the policy is working; if it lives only in brochures, it becomes another footnote in the long literature of industrial policy misfires. And while India navigates export controls and alliances, it must ensure that domestic energy and environmental constraints don’t become the new bottlenecks that keep factories quiet and GPUs idle.
Why This Was the Day’s Defining Line
Plenty of leaders say AI won’t take your job. Few attach that optimism to specific levers that move markets. By yoking the jobs debate to semiconductors and cheap compute on the opening day of a flagship summit, India’s government chose a side and named its instruments. Watch two gauges to see if the claim holds: the moment Micron’s facility spins up and the effective price of a GPU‑hour for a student founder in Bhubaneswar or Coimbatore. If those needles move, employment follows—not because AI is benevolent, but because the inputs that make AI productive got domesticated and discounted. In that world, “AI replaced me” starts to sound less like a eulogy and more like the beginning of a different kind of job description.
