(The Weekend Insight) - India’s Startup Salary Bubble

In today’s deep dive, we will look at how startup compensation in India moved away from normal salary benchmarks during the funding boom, what companies actually spent on employees, why some of those salaries were still justified, what happened when valuations fell, how employees should think about ESOPs, and why AI could produce a very different startup organisation over the next few years.

In 2021, switching jobs in India's startup ecosystem became unusually rewarding. Recruiters were reporting engineers getting salary hikes of 50 to 80 percent, and in some cases close to 100 percent. Engineers with multiple offers could keep negotiating until someone gave in. YourStory reported at the time that a software engineer with five years of experience could command around ₹70 lakh a year. Product-management offers at large startups were reportedly reaching ₹70 lakh at Flipkart, ₹90 lakh at Unacademy and as much as ₹1.2 crore at Swiggy.

There was a reason companies behaved this way.

Money was arriving faster than experienced startup talent could be produced. Indian startups raised more than $10 billion in the first half of 2021 alone, almost twice the amount raised during the corresponding period in 2020. Large funding rounds gave founders targets that assumed another two or three years of aggressive growth. A startup that had 300 employees and had just raised $200 million was often planning for the company it expected to become, rather than the company its current revenue could support.

That created a peculiar labour market.

If five startups had each raised large rounds and all wanted the same product manager, engineering leader or growth executive, the market price of that person went up. The employee had done nothing wrong by asking for 60 percent more. The startup had a rational reason to pay it. Losing six months because a critical engineering team could not be built might have been more expensive than adding another ₹20 lakh to someone's compensation.

The problem was somewhere else. Venture funding had started influencing what the ecosystem considered a normal salary.

When VC money became part of salary discovery

Companies usually decide what they can pay employees from some combination of revenue, margins, productivity, available skills and competitive salaries.

Startups added another variable: how much capital they had recently raised. That worked while capital was cheap.

A startup valued at $2 billion could tell itself that spending another ₹100 crore to build product, category, sales and strategy teams was reasonable because revenue might triple. Employees could accept lower cash in exchange for ESOPs because the next funding round might double the company's valuation. Investors could tolerate a high payroll because the company was expected to grow into it.

For a while, everybody's assumptions supported everybody else's assumptions.

Then funding slowed down and the calculation changed.

By 2025, the Deel and Carta State of Global Compensation report showed median compensation for engineering and data roles in India falling from around $36,000 in 2024 to $22,000 in 2025, a decline of roughly 40 percent. Product and design compensation also weakened. At the same time, compensation for specialised AI and machine-learning talent continued to rise, particularly at the top end of the market.

Another salary study from TeamLease put expected increases in ecommerce, fintech and technology startups at roughly 8.1 to 9.5 percent in 2025-26. That is still healthy salary growth. It is very different from a market where changing employers could produce a 50 or 80 percent jump.

This is why describing the current phase simply as a funding winter misses part of what has happened. The price of startup labour itself is being reset.

A ₹1 crore salary tells us very little

There is an easy way to make the argument against startup salaries and a better way.

The easy version is to find an employee earning ₹50 lakh or ₹1 crore and say the number is excessive. But, that does not tell us much.

A ₹1 crore engineering leader who allows a company to run with 15 engineers instead of 30 may be cheap. A senior salesperson earning ₹80 lakh who brings ₹20 crore of high-margin revenue is probably cheap. A product manager earning ₹60 lakh in a business generating very little gross profit from the product being managed can be expensive.

Salary only makes sense in relation to what the organisation produces.

Zepto gives us a good example of why this calculation is complicated.

The company reported ₹22,624 crore of operating revenue in FY26. Employee-benefit expense was approximately ₹1,785 crore, up 44 percent from the previous year. On the surface, employee expense was less than 8 percent of revenue, which looks extraordinarily efficient. Zepto nevertheless lost ₹5,905 crore during the year.

The problem is the denominator.

A rupee of quick-commerce revenue is very different from a rupee of software revenue. Zepto has to pay for merchandise, delivery, dark stores, warehouses and the other physical infrastructure required to move groceries from suppliers to customers.

Its salary numbers are still revealing. In April 2025, Zepto reportedly had a monthly salary bill of around ₹95 crore for roughly 3,000 employees. Earlier months had reportedly been even higher. The company had also developed a reputation for paying aggressively to attract talent from larger rivals. On a crude mathematical basis, ₹95 crore across 3,000 people works out to around ₹3.17 lakh per employee per month. That should not be confused with average employee CTC because senior compensation, variable pay and other payroll items distort the calculation. But the number shows the scale of the organisation Zepto built around the quick-commerce race.

The relevant question is whether that organisation eventually produces enough gross profit and operating leverage to justify the payroll.

If it does, the salaries were investments. If it does not, at least some of those salaries were another form of customer-acquisition spending funded by venture capital.

ShareChat shows what happens when people become one of the business model's biggest costs

ShareChat provides a cleaner comparison.

In FY24, the company generated ₹718 crore of revenue. Employee-benefit expense was ₹580 crore, equivalent to around 81 percent of revenue. Of that amount, ₹126 crore represented non-cash ESOP expense. Even after removing that identified ESOP charge, employee expense remained very large relative to revenue.

ShareChat was also becoming more efficient. Revenue grew 33 percent during FY24, while adjusted EBITDA losses fell sharply to ₹793 crore. Server costs were cut significantly and employee costs had already declined from the previous year.

That context matters because the number should not be used to argue that ShareChat employees were individually overpaid.

It says something about the organisation that had been built around the company.

For years, venture-backed consumer internet businesses could support very large engineering, product, moderation, sales, content, growth and management organisations because the objective was to establish scale first. The eventual advertising, transaction or subscription business was supposed to monetise that scale later.

When monetisation arrives more slowly than payroll, the company has only a few choices. Revenue has to accelerate, the organisation has to become cheaper, or both.

That process has been taking place across the ecosystem.

Udaan's smaller payroll came with a smaller business

Udaan is useful because its numbers prevent us from making the opposite analytical mistake.

The company's staff costs fell to around ₹500 crore in FY25, down 22 percent from the previous year. Revenue was ₹4,561 crore, down around 20 percent from ₹5,707 crore. Net losses narrowed 37 percent to ₹1,055 crore.

Go further back and the change is much larger.

Udaan's revenue was close to ₹10,000 crore in FY22. By FY25 it had fallen to less than half that figure. The organisation also became much smaller.

It would therefore be wrong to say Udaan discovered how to produce exactly the same business with far fewer employees. The business itself contracted.

But this gives us another way to understand the salary boom. Companies were hiring for anticipated scale. When that scale did not arrive, the organisation had to move backwards toward the business that actually existed.

This happened in product teams, strategy functions, growth organisations and management layers across the ecosystem.

During the funding boom, startups were rewarded for preparing early for growth. A company expected to triple might hire managers before their teams existed, build categories before demand had matured and create specialist roles that made sense only at much larger revenue.

After 2022, those costs became much harder to defend.

The same employee could go from being considered essential to being considered overhead without becoming any less talented.

What changed was the business assumption around the job.

Unacademy made the salary reset explicit

Few companies illustrated the reversal as clearly as Unacademy.

In 2023, the company informed founders and senior management that compensation cuts could reach 25 percent. The cuts were described as permanent, with salaries to be reconsidered later. Unacademy had already started a large cost-reduction programme as the economics of Indian edtech deteriorated after the pandemic boom.

Its earlier accounts also show why employee-cost numbers have to be read carefully.

Unacademy reported employee-benefit expense of around ₹1,771 crore in FY22, but roughly ₹1,140 crore of that amount related to ESOP expenses settled through cash and equity. Calling the full ₹1,771 crore the company's salary bill would badly misrepresent what employees were receiving as normal cash compensation.

Still, Unacademy's journey captures the larger change.

During the boom, compensation was being constructed around future valuation and future revenue. After the boom, compensation had to survive contact with current revenue.

The company could cut free meals, office expenses, marketing and travel. Eventually, however, a people-heavy technology business has to examine what it pays people and how many people it needs.

That is why salary resets have become more significant than the layoffs themselves. A layoff removes an employee.

A salary reset changes the reference price used by thousands of employees who remain in the market.

Some companies are showing what good payroll leverage looks like

Urban Company gives us a useful counter-example.

Its operating revenue increased 38 percent to around ₹1,144 crore in FY25. Employee-benefit expense stayed almost flat at approximately ₹350 crore, including around ₹72.5 crore of non-cash ESOP expense. On a pre-tax basis the company also moved into profit.

That is a more interesting productivity story than simply reducing salaries.

Revenue rose substantially while employee expenditure barely moved.

The company's service professionals are not all employees, so Urban Company's model naturally has different labour economics from a software company or quick-commerce operator. Even so, the direction is important. The central organisation was able to support considerably more business without employee expense growing at the same rate.

This is what startup investors increasingly want when they talk about operating leverage.

Headcount can grow. Salaries can rise. Expensive senior people can be hired. What becomes harder to defend is payroll growing indefinitely at the same rate as revenue.

For years, revenue growth excused organisational growth.

The next phase is likely to demand revenue growth without proportional organisational growth.

ESOPs made expensive compensation easier to sell

Cash salary explains only part of the startup compensation boom.

ESOPs helped bridge the difference between what startups could afford to pay today and what employees hoped to earn tomorrow.

An offer might be presented as ₹45 lakh cash plus ₹30 lakh of ESOPs. On a compensation spreadsheet, that became ₹75 lakh.

Economically, the two numbers were never equal.

₹45 lakh of salary is cash compensation.

The eventual value of the other ₹30 lakh depends on vesting, exercise price, dilution, taxes, the company's next valuation, liquidation preferences, secondary transactions and whether an exit ever happens.

During the valuation boom, employees had good reasons to treat those options seriously. Flipkart had already created genuine employee wealth. Freshworks' listing created another visible example. Later, startups started conducting large secondary transactions and ESOP buybacks without waiting for an IPO.

That market is becoming more mature. It is also showing how uneven ESOP outcomes can be.

Seven Indian startups conducted nearly $220 million of ESOP buybacks and liquidity transactions in the first quarter of 2026, according to Entrackr. That exceeded the roughly $190 million recorded during all of 2024 and the little over $75 million recorded in 2025.

The headline looks impressive until we examine where the money came from.

BrowserStack announced a $125 million share and ESOP buyback in January 2026 covering more than 500 current and former employees as well as early investors. It was the company's third such programme, taking cumulative buybacks to $275 million.

Innovaccer accounted for another roughly ₹600 crore, or around $75 million, of employee liquidity. Together, BrowserStack and Innovaccer represented most of the quarter's headline number.

So two things can be true at the same time.

Indian startup ESOPs have created real wealth for thousands of people.

An average startup employee should still not treat an ESOP grant like cash.

Astrotalk and Unacademy show the two ends of the ESOP market

Astrotalk became a particularly useful example this week.

The Noida-based astrology platform crossed a reported $1 billion valuation through an ESOP buyback involving more than 100 employees. The transaction was reportedly funded through the company's own profits rather than fresh investor money. In FY25, Astrotalk had generated ₹1,176 crore of operating revenue, up 81 percent from the previous year, and roughly ₹250 crore of net profit.

That is close to the ideal version of startup equity compensation.

Employees receive options while building the company. The underlying company becomes profitable. Cash generated by the business eventually creates liquidity for those employees.

Now compare that with what happened around Unacademy's ESOPs.

The company had been valued at $3.4 billion during its 2021 funding round. By late 2025, a merchant-bank valuation used in connection with the ESOP situation was around ₹2,650 crore, or roughly $230 million. Unacademy also initially reduced the post-employment exercise period for former employees from as much as ten years to 30 days, before the controversy around the decision developed. Founder Gaurav Munjal explained that liquidation preferences attached to the more than $800 million raised by the company could leave employee options with little or no economic value at a sufficiently low exit valuation.

That is what employees rarely saw on the compensation slide.

Startup common equity does not necessarily participate in every rupee of headline valuation.

Investors can have preferential rights. Later rounds dilute earlier holders. An acquisition below the amount of preference capital can leave common shareholders with very little.

The ₹30 lakh ESOP line in a joining offer was therefore always a probability-weighted claim on future value. During 2021, the probability was often treated as though it were close to one.

It never was.

AI is changing the calculation again

The first salary reset came from expensive capital. The second could come from cheaper labour productivity.

Dukaan became an early and controversial example in 2023 when founder Suumit Shah said the company had replaced about 90 percent of its customer-support team with an AI chatbot. He said customer-support costs had fallen roughly 85 percent.

Since then, the technology has become considerably more capable.

Reuters reported in 2025 that Indian AI companies such as LimeChat and Haptik were automating large volumes of customer-service work. LimeChat said its systems had already automated work equivalent to thousands of jobs, while companies in the sector argued that AI could reduce human requirements in parts of customer support by as much as 80 percent.

The effect will not be confined to call centres.

Startups employ large numbers of people doing research, first drafts, routine coding, testing, basic analysis, customer communication, sales preparation, reporting, data cleaning and internal coordination. These jobs will not all disappear. The number of people needed to produce the same amount of work can still fall.

That changes the economics of junior hiring.

A startup may previously have hired one senior manager and six junior employees because the manager needed people to prepare spreadsheets, produce first drafts, answer standard queries, perform research and update systems.

If AI removes half of that workload, the company may prefer one very good senior person, two or three stronger junior employees and a collection of AI tools.

The salary bill can fall even if the remaining employees are paid more.

That helps explain the strange compensation market emerging now. Median Indian tech compensation has weakened, while the Deel and Carta data shows AI and ML salaries rising aggressively, particularly for scarce people at the upper end.

AI may therefore make the startup salary market more unequal rather than simply cheaper.

The middle of the organisation could feel the most pressure

The junior employee faces competition from automation. The exceptional senior employee can use automation to control more output.

The difficult position may increasingly be the reasonably good middle.

Indian startups created large numbers of roles during the funding boom that sat between execution and genuine business ownership. Strategy teams produced analyses. Growth teams coordinated agencies. Product organisations acquired multiple management layers. Founder's offices absorbed ambitious young generalists. Category teams expanded as companies entered new businesses.

Many of those employees are capable people.

The question companies are beginning to ask is whether the role itself produces enough incremental value.

AI makes that question easier to ask because the alternative is no longer simply "hire another person or leave the work undone."

The alternative can be to redesign the work.

That is why the next version of the Indian startup organisation may have fewer people earning average startup salaries and more people at both extremes: relatively inexpensive execution roles at one end and highly paid people with genuine revenue, product or technical ownership at the other.

The comfortable middle becomes harder to defend.

Startup salaries were never the real problem

India should not want startup salaries to fall simply because founders once paid too much.

High salaries are generally a good outcome when productivity supports them. India needs companies capable of paying engineers, product managers, scientists and operators far more than traditional employers did. Better pay is one of the reasons talented people left consulting firms, IT services companies and multinational corporations to build Indian startups in the first place.

The mistake was treating compensation itself as proof of value creation.

A company raising $300 million does not suddenly make every job inside it more productive.

A private valuation moving from $1 billion to $4 billion does not mean an employee has become four times more valuable.

A product manager receiving three competing offers does not mean the product being managed has good economics.

For several years, the startup ecosystem confused the market price of scarce labour with the economic productivity of that labour.

Cheap capital made that confusion affordable. It is becoming less affordable now.

The better way to judge startup compensation is through fairly ordinary business questions. How much gross profit does the company generate per employee? Is revenue increasing faster than payroll? Can the company add customers without adding the same percentage of people? How many management layers sit between founders and actual execution? Does a ₹1 crore employee control a function that can produce many multiples of that amount? Is ESOP compensation backed by a plausible path to liquidity?

The answers will differ dramatically by business model.

A quick-commerce company needs more operations. A SaaS company should produce considerably more revenue per corporate employee. A marketplace can keep much of its service workforce outside payroll. A financial platform may have heavy compliance and risk functions. Comparing every company using a single employee-cost ratio would be as misleading as comparing their revenue multiples without considering margins.

But every company eventually faces the same basic constraint.

People have to produce more economic value than they cost.

Venture capital can postpone that calculation. It cannot remove it.

India's startup salary bubble therefore looks less like a bubble that suddenly bursts and more like a market going through a long repricing. The 2021 premium on general startup talent is weakening. ESOPs are being judged more carefully. Companies are learning that a large organisation can become a liability when revenue disappoints. AI is reducing the number of people needed for some kinds of work. At the same time, unusually capable engineers, salespeople and operators may become even more expensive because one strong employee can now do more.

The ₹1 crore startup salary is unlikely to disappear.

What is disappearing is the assumption that working at a well-funded startup is, by itself, enough to justify one.

That is probably a healthier correction for India's startup ecosystem than a simple salary crash would have been. The next generation of compensation will still be aggressive in places, but companies will increasingly have to explain what they receive in return. For employees, the same change means that the value of a role will depend less on how much money the employer recently raised and more on how directly that role connects to product, revenue, margins or operating efficiency.

After years in which capital helped set the price of talent, business fundamentals are slowly taking that job back.