PhysicsWallah Says AI Writes 90 Per Cent Of Its Code, And Its Technology Bill Rose By Half

In this storyPhysicsWallah

PhysicsWallah told analysts on its second-quarter earnings call that its total employee count and faculty strength both declined sequentially, attributing the reduction in part to efficiencies from artificial intelligence.

Management said headcount stood at around 19,100 and faculty at approximately 7,100 at the end of September. It said AI had created efficiencies across content creation, quality checks, test paper preparation and evaluation. Employee expenses, excluding ESOP costs, declined 2.3 per cent year on year during the quarter.

The faculty figure moved in two directions. The company had 84 more salaried teachers on payroll at the end of September and 235 fewer faculty consultants, a net fall of 151. Management did not provide a breakdown of how many people were laid off, resigned or reached the end of a contract, or how much of the change was attributable to AI.

On software, management said AI now writes around 90 per cent of the company's code. First-half technology spending rose from ₹70 crore to ₹104 crore, a figure that covers the company's wider technology bill including work on AI.

The offline business remains under pressure. Revenue from the segment rose 9 per cent to ₹541 crore, while it posted a pre-Ind AS EBITDA loss of ₹108 crore in the first half of FY27, which the company linked partly to lower NEET enrolments and delays in government exam notifications.

PhysicsWallah reported second-quarter revenue of ₹1,242 crore, up 18 per cent, with profit rising 2.1 times. It has earmarked roughly $20 million or less for potential acquisitions of fully online education businesses, and has an investor day scheduled for 18 November. The company listed in November 2025 following a ₹3,480 crore initial public offering.

Two numbers from the same call sit awkwardly together, and the company did not reconcile them.

AI writes around 90 per cent of the code. First-half technology spending went from ₹70 crore to ₹104 crore, an increase of roughly 49 per cent. Part of that rise is the AI itself, since the figure covers the wider technology bill, which is precisely the point. The saving on people who write software is being spent on the systems that write it instead, and then some. Whatever else automation has done here, it has not reduced what the company spends on technology.

The employee line tells a similar story at smaller scale. Expenses excluding ESOP fell 2.3 per cent while revenue grew 18 per cent, which improves the ratio considerably and saves very little in absolute terms. A company that had genuinely removed a large block of labour cost would show more than 2.3 per cent.

What is missing from the disclosure is the part that would settle it. PhysicsWallah has not said how many people were dismissed, how many resigned, how many reached the end of a contract, or what share of the decline AI explains. The headcount fall is a fact. The attribution is management's characterisation of a number it has chosen not to decompose.

It is worth noting when that characterisation arrived. The same quarter showed an offline business losing ₹108 crore at the EBITDA line on ₹541 crore of revenue, hurt by weaker NEET enrolments and delayed exam notifications. An AI efficiency story is a more comfortable thing to discuss on a call than a classroom expansion that is not yet working, and both were on the agenda.

The faculty numbers describe something more specific than automation, and more interesting.

Eighty-four more salaried teachers. Two hundred and thirty-five fewer faculty consultants. That is not AI replacing teaching. It is a company converting a variable, contracted teaching base into a smaller permanent one, while automating the work that sits around the classroom rather than inside it. Content creation, quality checks, test papers and grading are back-office functions of an education business. The teaching itself is being brought closer, not further away.

That distinction matters for anyone reading this as a template. The jobs under pressure at PhysicsWallah are not the ones most people imagine when they hear that AI has arrived in education. They are the people who prepared materials and marked papers, and a large share of them were on contracts rather than payroll, which is why the adjustment can happen quietly and without a redundancy announcement.

It also fits what the Bank of England described four days ago, that firms adopting AI become less likely to hire or replace departing staff without becoming more likely to fire them. A contract that is not renewed produces the same headcount outcome as a dismissal and none of the disclosure.

The harder question is what this does to the product. PhysicsWallah's original advantage was teaching quality at low prices. Grading by model and materials by model are defensible if the output holds, and are the first thing students notice if it does not.

For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.