Indian edtech has not recovered from Byju’s collapse. AI offers hope

Early- and growth-stage startups, including Seekho, SpeakX and Arivihan, are using large language models to AI chatbots to offer engaging courses in multiple languages and subjects. And they are reaching audiences beyond India’s top metros.

“We want to move beyond short videos and make learning interactive. While videos will remain the primary mode, we’ll integrate contextual chatbots to make the journey dynamic,” Rohit Choudhary, founder and chief executive officer (CEO), Seekho, toldMint. “Over time, this may evolve into a seamless AI-driven experience where users get personalized prompts in between videos.”

As these startups drive user engagement and grow subscriptions, investors are taking note. That signals a potential revival in the edtech sector that is still reeling from the collapse of hyper-funded giants like Byju’s. But caution still persists. According to data from Venture Intelligence, deal activity has plunged from 117 in 2021 to just 21 so far in 2025, while investment volumes have collapsed from $4.3 billion to $141 million over this period.

Seekho, an edutainment company, has raised $28 million in a Series B funding round led by Bessemer Venture Partners. Elevation Capital, Goodwater Capital, and existing backer Lightspeed participated. With the fresh capital, the company plans to incorporate AI into its content offerings.

The company is also exploring $3-4 million deals to acquire early-stage startups working on AI for learning, according to Choudhary.

SpeakX, which teaches users English through a conversational AI bot, is seeking to raise $15 million in a new round. Arivihan, which provides personalised courses in local languages, just raised $4.17 million in a seed round co-led by Prosus Ventures (its first edtech bet in India after Byju’s) and Accel. SigIQ, an edtech focused on personalised UPSC and GRE prep, raised $9.5 million in seed funding led by The House Fund & GSV Ventures. Similarly, Stimuler, a voice-first tutor for English as a second language, secured $3.75 million in pre-Series A funding from Lightspeed, SWC Global, and others.

Seekho relies on existing large language models to personalize content for users, implementing prompt engineering and fine-tuning where required.“We’re experimenting with ways to make the platform more social, but we want to balance it carefully,” Choudhary said. “Think of YouTube, which is completely social, versus Netflix, which is curated. We want to find a middle ground. Pilots are underway to explore this.”

Business is driven not just by the Indians’ desire for education but also aspiration. SpeakX and Seekho’s target markets aren’t concentrated in urban areas but they have a large userbase in tier-II cities as well.

“By shifting focus from institutions and educators to the individual learner, AI tailors content, pace, and assessments to each student’s need,” said Rahul Chowdri, partner at Stellaris Venture Partners. “This shift from standardized to personalized learning could be one of the most transformative impacts of AI on education.”

The early-stage venture capital firm is bullish on professional upskilling, self-help, and language learning.

Focus on early-stage startups

Mint had reported early signs of edtech recovery last year, but most of the activity was concentrated among market leaders such as Eruditus and Physics Wallah, often with a secondary component that allowed early investors to exit. In 2025, that recovery in late-stage funding has all but dried up, slipping to just two deals this year compared to nine in 2024.

In contrast, activity in 2025 so far has been concentrated at the early and growth stages. Early-stage startups have already seen nine deals in the first eight months—close to last year’s tally of 14, while growth-stage companies have recorded nine deals versus 10 in all of 2024,according to data from Venture Intelligence.

The AI edge

Founders are, however, optimistic about AI supporting both scale and cost efficiency.

“With AI, scaling is much faster and cheaper. We already have users in almost every country, with paying subscribers in 180 countries,” Akshay Akash, co-founder of Stimuler, told Mint. “That’s something impossible to achieve with a traditional teacher-based model, since you’d need to hire teachers in each country for each accent.”

Thanks to cheaper services, Akash sees companies like his acquiring paying customers quickly. “For consumers, it would be much more expensive to hire teachers, which costs around $3,000 a month, versus our $10 a month subscription.”

Companies like Seekho and SpeakX have pivoted their business models to become subscription-first as UPI AutoPay has picked up. More people are willing to make bite-sized payments for things they believe will benefit them in the long run.

“Efficacy must be demonstrated through measurable improvements–in test scores, skills acquired, or fluency levels,” said Chowdri from Stellaris. “Retention will be equally critical–these startups must show that their AI model can sustain long-term engagement while scaling efficiently.”

At the end of the day, these companies still need to prove that their businesses work at scale, and to that end, a full-blown revival might still be at least two years away, according toIvyCap Ventures’ founder and managing partner Vikram Gupta. “It all comes down to the value they’re creating. We’re at a stage of growth where a lot of the products are still unproven. The capital that you need at that stage is higher risk capital.”

That’s also reflected in deal size, which has not recovered after the collapse from pandemic highs. Average ticket sizes fell sharply from $37.2 million in 2021 to just $6.7 million in 2025, underlining investor caution.

Legacy models to stay

Legacy players, too, have acknowledged the potential of such platforms. In July, Mint reported that Gaurav Munjal, co-founder and CEO of test-prep edtech firm Unacademy, has been looking to hive off Airlearn, an AI language learning company, and raise separate funding for it.

Similarly, IPO-bound PhysicsWallah is fine-tuning small language models to solve student queries on physics, chemistry, mathematics and biology. The company also plans to use AI-based voice dubbing to take the lessons of some of its star teachers pan-India.

Sri Chaitanya’s Infinity Learn, too, has introduced an AI mentor, AINA for JEE and NEET test prep.

Still, Akash of Stimuler said, such models work better to supplement learning, like language acquisition. “Older, legacy models will continue to dominate mainstream learning such as test prep and K–12. That requires human touchpoints and teachers, and parents are willing to pay more for it.”