Can AI Give India's Edtech Ecosystem a New Lease of Life?

When India’s edtech boom peaked in 2021, the sector appeared unstoppable. 

At its height, Byju’s was valued at $22 billion. Indian edtech startups collectively raised over $4 billion that year as pandemic lockdowns forced classrooms online and venture capital chased digital learning at scale. 

Two years later, the exuberance had evaporated. Funding into Indian edtech fell by more than 80% from its peak. Layoffs swept across major players. Valuations were written down. Aggressive inside-sales models that once symbolised scale became cautionary tales about unsustainable customer acquisition costs and weak retention. 

What followed was not just a funding winter. It was a credibility crisis. 

Now, a new cycle is emerging: more cautious, more metrics-driven, and increasingly shaped by artificial intelligence. The question is whether this second wave represents structural improvement or simply technological optimism layered onto old fragilities. 

This Isn’t the First Cycle 

India has seen this story before. In the mid-2000s, Educomp led the first major wave of education technology through its “smart class” model, selling digital classroom infrastructure to schools. At one point, Educomp’s stock traded at around ₹950 per share. 

Unlike the second wave, including Byju’s and others, the first wave was focused on the institutional market. Schools signed multi-year contracts. Hardware, content, and services were bundled together. Revenue visibility appeared strong. 

But expansion was fuelled by debt. Execution complexity ballooned. Payment cycles stretched. Governance concerns mounted. Over time, the balance sheet buckled. Today, Educomp stock trades at around ₹1. 

Both cycles relied on the belief that technology adoption in education would scale predictably once infrastructure was in place. In both cases, capital extended timelines beyond what underlying economics justified. 

Third Time’s the Charm? 

After getting it wrong twice, the edtech industry is attempting a third act. 

The first attempt focused on bringing technology to classrooms. The second wave was built around personalised learning through at-home devices such as tablets and smartphones. The theme of the third wave is unmistakably AI. 

Unlike earlier platforms built around static video libraries, newer systems are structured around adaptive learning engines. AI generates practice problems dynamically, adjusts difficulty in real time, and provides personalised feedback loops. 

For founders in this camp, AI addresses two structural constraints: engagement and cost. 

For example, if a user is weaker in quadratic equations, the AI engine automatically generates additional step-by-step practice problems, breaks down the underlying concepts into smaller modules, and revisits prerequisite topics like linear equations before progressing further.

“AI helps us turn what used to be static content into something that actually responds to the learner,” says Dhritiman Talukdar, co-founder and CTO of Vidya AI, a company that uses AI technology to provide personalized learning tools to students. 

Such adaptive systems aim to improve retention, the metric that eluded many first-generation platforms. Where several direct-to-consumer players struggled with sub-40% monthly retention during the boom, AI-first companies now pitch retention north of 60% as baseline sustainability. 

The second advantage offered by AI is in costs. While the first two waves of companies spent a large part of their expenditure on creating vast content libraries, the third wave is distinguished by its use of synthetic and on-the-fly content. The creation of content, once dependent on faculty and expensive studio infrastructure, can now be automated to a large degree. 

Learnings from the Past 

One of the key differentiators of the third wave is the incorporation of lessons from the previous two. With the benefit of hindsight, the industry is better placed to avoid the pitfalls that undermined companies in the earlier rounds. 

The first major lesson is, paradoxically, to focus less on growth. Growth at any cost was the mantra. 

According to reports, during the peak years, customer acquisition costs in competitive exam categories were often estimated between ₹6,000 and ₹15,000 per paying user. Marketing frequently consumed more than 40% of revenue. Payback periods stretched well beyond a year. 

“The first wave was all about chasing top-line growth,” says Gaurav Munjal, co-founder of Unacademy. “We were running hard at scale. And honestly, we didn’t pause enough to ask if the engagement was deep enough to justify that pace. Retention should have been the real north star.” 

While the model worked as long as capital was abundant, the dependency made it vulnerable to changing global macroeconomic conditions. 

When the US Federal Reserve began aggressive rate hikes in 2022 and global venture funding slowed, Indian startup funding fell by more than 30% year-on-year. The economics that had been masked by abundant capital were suddenly exposed. Companies that had focused on the top line and neglected the bottom line suddenly found themselves without the funds required to operate. 

Sanjay Nath, co-founder and managing partner at Blume Ventures, has described the shift as overdue. “In 2021, money was easy, and that made it easier to overlook what wasn’t working,” Nath has said in public conversations. “Now we’re asking tougher questions: are users sticking around, is the burn sensible, and most importantly, are they actually learning or just enrolling?” 

Today, therefore, founders are careful to ensure that their operations are not only scalable but also sustainable. 

The second learning has been about the fickle nature of advertising-fuelled, incentive-based growth. The earlier era leaned heavily on direct-to-consumer marketing, scaling tele-sales and performance advertising aggressively; the new cycle is more cautious. 

This is one of the reasons Teachmint, co-founded by Mihir Gupta, built infrastructure tools for teachers and schools rather than chasing individual household subscriptions. “In the end, it’s about building an operating system for schools, not just another app,” Gupta has said in public forums. “We’d much rather rely on steady SaaS revenue than chase unpredictable consumer funnels.” 

Similarly, instead of focusing solely on direct-to-consumer exam prep, Vidya AI emphasises tools that assist teachers in automatically generating assessments, converting lesson content into interactive quizzes, and providing data on student comprehension. 

Finally, there has been a reassessment of the target market, with a shift towards more predictable, outcome-oriented segments and away from the K-12 market. 

Learners paying for software engineering upskilling typically expect tangible career outcomes with higher salaries, better roles, or job transitions which makes them more willing to commit to structured, time-bound cohorts rather than casual self-paced courses. In fact, 85% of Indian professionals say they plan to invest in upskilling in the coming year.  

Hence, software engineering and job-linked training command fees ranging from ₹1.5 lakh to ₹3 lakh, significantly higher than typical K-12 subscription models, leading to players like Scaler to focus on measurable career outcomes in upskilling. 

The insight is that institutional and group-based models involve longer sales cycles, but they offer stronger renewal visibility and revenue predictability. Contracts often run 12-24 months, smoothing cash flows and reducing dependence on high-intensity consumer marketing. 

Challenges 

These learnings are necessary but may not be sufficient. The sector continues to face challenges that make at least some investors nervous, with the biggest being the threat of disintermediation. 

Tools like ChatGPT are already widely used by students and professionals to assist with homework, coding problems, language learning, and interview preparation. Many of these tools are free at the point of use, and foundational models improve meaningfully every eight to nine months with new versions. 

If a student can ask a large language model to explain calculus, debug code, summarise chapters, or simulate interview questions instantly, the question becomes unavoidable: why pay separately for an edtech subscription? 

In practice, many learners are already using large language models as informal tutors. College students use them to understand problem sets. Working professionals use them to refine code and prepare for interviews. For basic doubt resolution and practice, the marginal cost is effectively zero. 

This creates a structural disintermediation risk. 

At Peak XV Partners, the scrutiny reflects this reality. AI in the pitch deck no longer guarantees enthusiasm. Generative tools reduce content costs for everyone. 

The startups believe they have an edge in the form of proprietary data and the ability of their systems to improve with usage. 

But some investors are not convinced. They point out that most startups build their products on top of external foundation models, layering curriculum and dashboards over APIs. These products and services, they argue, remain downstream of the core intelligence layer. Competing directly with global AI systems on raw explanation quality or breadth is unlikely to be sustainable. 

If, alternatively, they attempt to build proprietary models from scratch, they enter a capital-intensive race against some of the best-funded AI labs in the world. 

The realistic moat, therefore, may not lie in outperforming general-purpose AI but in structuring learning journeys around it by spressing accountability, sequencing, peer groups, assessments, and outcome tracking in ways that generic tools do not natively provide. AI may become infrastructure. The defensibility must sit in pedagogy, data loops, and distribution. 

Even if the threat of disintermediation is seen off, more structural limits remain. Technology can personalise instruction, but it cannot generate motivation, discipline, or institutional reform. Adaptive systems, say some, can support teachers but never replace them. 

Cautious Optimism 

Despite such uncertainties, there are early signs of stabilisation in India’s edtech sector. Funding remains below 2021 highs, but capital has not disappeared. The startups attracting investment today tend to demonstrate disciplined growth, clearer unit economics, and narrower focus areas. 

Globally, companies like Duolingo demonstrate that durable edtech businesses are possible, built less on spectacle and more on retention and habit formation. 

In India, the latest cycle appears less exuberant but more measured. The earlier era believed capital would buy scale and scale would buy profitability. The current wave believes intelligent systems will deepen engagement and that engagement will underpin sustainability. Whether AI ultimately delivers structural advantage in a world where foundational models are widely accessible remains to be seen. 

What is undeniable is the opportunity. India today has over 750 million internet users and more than 600 million smartphone users. Yet paid online education penetration remains relatively modest outside test prep and upskilling. Various industry estimates place paid K-12 edtech penetration in the single digits as a percentage of total enrolled students. The addressable base is vast; the conversion to sustained paying users has historically been far narrower. 

Whether it will be the AI-native players, the surviving incumbents or global technology platforms that ultimately capture that opportunity may be the defining question for Indian edtech over the next five years.