In FY24, India's four largest IT services companies — TCS, Infosys, Wipro, HCL — together hired 60,000 freshers.
Two years earlier, that number was 2,25,000.
A 70% collapse. In 24 months. In the single largest absorber of Indian engineering graduates.
This isn't a recession. It's a structural reset. And most of us are still pretending it's not happening.
I've spent the last few months inside 5+ engineering institutes — observing classrooms, reviewing curricula, talking to faculty and students. As a CTO who still codes daily and sits on academic boards, I'm writing this as an insider, not a critic.
What I saw confirmed what the data already screams.
The numbers nobody wants to discuss:
India produces 1.5 million engineering graduates a year. Per TeamLease, only ~10% of the 2024 batch were expected to actually find jobs.
Aspiring Minds: only 4.77% of engineering graduates can write basic correct programming logic.
NASSCOM: fewer than 3% of Indian engineers meet industry needs for AI roles.
AICTE: over 550 engineering colleges have shut down since 2015. The trend is accelerating, not slowing.
Salaries for the freshers who DO get hired: still ₹3.5 LPA — the same as 2014, against doubled cost of living.
Here's the part that will make educators uncomfortable:
Last year, Harvard published a randomized controlled trial in Nature's Scientific Reports. They compared a custom GPT-4 tutor against an active-learning physics classroom run by Harvard faculty.
Result: Students using the AI tutor learned more than TWICE as much. Statistical significance: p < 10⁻⁸. Effect size: 0.73 to 1.3 standard deviations.
In education research, anything above 0.4 is considered meaningful. This was three times that.
Read that again. AI tutors today beat Harvard professors on content delivery.
If AI beats Harvard, what is it doing to the average tier-2/tier-3 college lecturer in India?
The teacher problem nobody is naming honestly:
A significant portion of faculty in our tier-2 and tier-3 engineering institutions hold a PhD on paper but cannot write production-grade code, have never built a real product, have not contributed to open source, and have never spent a day in industry.
I'm not saying this about IITs, IISc, IIITs, top NITs, or strong autonomous private institutions. They're world-class.
I'm talking about the long tail — the several thousand colleges where PhD became a job-ticket, not a marker of expertise. The UGC itself launched a retrospective audit of PhD theses across India's 851 universities. UGC officials told The Print, on record: in tier-2/3 cities, "Doing a PhD is a way of getting employment for many people. Commitment to research is not the point."
That's the regulator saying it. Not me.
Four questions every teacher should ask themselves today:
When did I last build something — anything — outside the classroom?
When did I last read a research paper from this year's NeurIPS, ICML, or arXiv?
Could I get hired at a real product company today on my actual skills?
Do I use AI tools — Claude, Copilot, Cursor — in my own work?
If those questions are uncomfortable, that discomfort is the most important data point in this post.
AI is not coming for the teacher who can answer them confidently. AI has already arrived for the teacher who cannot.
The shift, said plainly:
Content delivery is no longer a teacher's job. AI does it better, in any language, on demand, at near-zero cost.
The teacher's role is changing — not disappearing — from performer to facilitator. From lecturer to guru. From content delivery to judgment, mentorship, and project critique.
This is closer to the Indian guru-shishya tradition than to the British-era classroom we inherited.
The teachers who make this shift will be more valuable than ever. The ones who don't will be replaced — not by AI directly, but by younger faculty who use AI fluently.
To parents reading this:
The ₹15-20 lakh tier-3 engineering bet is no longer the safe one it was. If your child can crack a top-100 institution — IIT, NIT, top BITS, top private — the investment still makes sense. If not, please seriously consider alternatives:
A skilled diploma. Plumbers, electricians, EV technicians, precision-agriculture farmers will out-earn most mid-tier IT graduates by 2030.
A B.Sc. with self-built projects, GitHub, and AI fluency. A 19-year-old with three shipped products will out-earn a tier-3 B.Tech.
Stop banning AI at home. That's like banning calculators in 1995.
Stop optimizing for marks. Start optimizing for what your child has built.
The civilisational opportunity:
For 70 years, we trained our young to serve other people's ideas, in other people's languages, for other people's economies. We were, honestly, a backend of the world.
AI changes this. The cost of building world-class technology has collapsed. Five Indians with strong skills and good AI tools can now build what required 500 a decade ago.
This is the moment to build originally. Products rooted in our own languages, our own problems, our own civilisational genius. To reintegrate Sanskrit, Vedanta, Ayurveda, classical music, indigenous architecture with modern engineering — and create what the West genuinely cannot.
But this requires an education system that produces thinkers and builders, not credential-collectors.
Where this leaves us:
Institutions: you have ~5 years. Adapt or close.
Teachers: your role is more important than ever — IF you make the shift.
Parents: stop measuring your child against a 1995 yardstick.
We are at an inflection point. Possibly the largest in education since the printing press.
India can lead this transformation, or be left behind by it. The choice is being made every day, in classrooms and homes across the country.
Choose carefully.
I've written a detailed paper with all sources, infographics, and stakeholder action checklists.
Download
If this resonated, share it with one teacher, one parent, and one institution head you know. That's how this conversation actually moves.
Disagreement welcome. Pretending the problem doesn't exist is not.
First published on LinkedIn.