What India’s Early Stage AI Investors Want: Depth In Data, Talent And Product

What is the state of early stage AI investments in India and are investors changing the way they evaluate startups and founders amid raging debates on sovereign models, Indian LLMs as a key focus area, beyond GenAI applications and SaaS tools?

More and more investors are factoring in the influence of GenAI and its future potential impact on sectors and industries when evaluating early stage deals. Even two years into the AI revolution, Indian venture capital in AI is concentrated towards startups in the applications and developer tools layer.

Murmurs of sovereign models and made-in-India large language models might yet influence the course of VC dollars flowing into AI this year, but it’s still too early to call this a significant shift.

However, the evolving due diligence models of VCs in the AI space — and even those outside — has reset expectations from early stage founders and startups. And this includes the potential disruption from sovereign models and Indian LLMs in the future.

Early stage fund managers and VCs are digging deeper into the architecture of the AI stack, dependency on APIs, the flexibility and efficiency of AI models, cost overruns and revenue leaks, and whether AI is core to the startup’s product roadmap. While most early stage investments did not carry the full due diligence weight in the past, this is changing, with technical due diligence becoming very critical.

At the same time, investors are reevaluating how they see the team behind the idea or the product, given the need to be highly conversant with the bleeding edge of AI development. Are teams ready for the next big jump, fund managers are asking as they look beyond the founder.

Inflection Point Ventures’ Ankur Mittal told Inc42, “We now evaluate data strategy, model flexibility, and AI-driven defensibility. Startups that capitalise on proprietary data and thoroughly integrate AI into their value offering outperform those who only use AI as an add-on.”

Instead of focussing on whether they have built their AI models from scratch or integrated third-party solutions, the real criterion should be the impact of technology on end users.

“The first question should always be: What problem does it solve, and how does it enhance customer experience? If a task that once took hours can be completed in minutes, that is a game changer,” Kushal Bhagia of All In Capital told us in an interview last month after the firm launched its second early stage fund.

Once the user impact is established, the focus shifts to technical depth and execution. Companies that create highly personalised AI experiences for specific user segments tend to develop more substantial products.

The All In Capital founder cited the example of an AI-powered skilling startup SuperNova, which enables users to speak English through structured interactions, rather than an open-ended course. This directly helps professionals like hotel employees to upskill for their current job, and it’s a more compelling learning experience than Duolingo or something like that.

In a similar vein, investors think that AI that can redefine existing experiences will win in the long run, since this is the intuitive path set by the internet age companies. While AI-driven SaaS solutions for niche B2B applications are gaining traction in sectors like healthcare, fintech and edtech, consumer-facing AI products in India are still nascent.

Where The Money Is Flowing

Marquee VC firms such as Prosus, Tiger Global, Peak XV Partners, Accel, Lightspeed among others have already made several early stage SaaS bets in the GenAI space — across the applications layer as well as the developer tools layer.

Many of these larger funds foresee startups in their portfolio to bag major returns as the AI wave leads to consolidation, but the point is to survive till that stage without needing to constantly evaluate your go-to-market strategy or your product-market fit. But this is the reality in SaaS and the applications layer.

For instance, in the SaaS space, most investors Inc42 spoke to believe things are changing at such a high pace that it’s hard for even practiced founders to keep up. The revenue run rates are growing rapidly across industry and SaaS companies are reaching the $1 Mn ARR holy grail faster than ever.

This revenue spike has created something akin to blinders for founders, so it’s easy to get swept up in this frenzy and not build a real moat.

AI is no longer a moat, it is par for the course in SaaS. So the real product-market fit comes from elsewhere, and many investors believe this is why SaaS is dead, and AI is the new software.

This despite the fact that a majority of the venture capital invested in the past two years has gone for AI adoption in some ways or the other, benefitting the SaaS ecosystem.

AI has become a central pillar for efficiency and startups have looked to adopt GenAI at scale rather than hiring en masse to remove operational inefficiencies. As a result, SaaS-AI hybrid tools have boomed.

For investors like Abhishek Prasad, managing partner at Cornerstone Ventures, the AI wave in SaaS is prompting fundamental changes in deal evaluation. His lens has shifted from traditional SaaS metrics like CAC and ARR growth alone to newer AI-driven qualifiers. “With AI becoming core to SaaS, we are interested in how AI is being leveraged by these companies, what is the impact it is making to the value proposition, is it creating new moats, and is the cost of leveraging AI capabilities delivering the right ROI to both the startups and their customers.”

Inc42 reported that AI startups comprised nearly 40% of new SaaS ventures funded in 2024, a sharp leap from 19% in 2022. In total, Indian SaaS startups raised over $2.1 Bn in 2024, up 31% YoY. A growing chunk of this capital is flowing toward companies that are not just building on AI, but being built by AI.

Leading VC firms are targeting startups creating AI-based consumer and business applications and developer tools. They believe such ventures have the potential to launch globally competitive solutions.n“We have not seen many AI startups targeting direct-to-consumer experiences in India yet, but it’s only a matter of time,” said Bhagia.

Unless new consumer products completely reimagine the user experience, they may have little competitive advantage over incumbents.

Plugging The Foundational Gaps

Global tech giants dominate the infrastructure, cloud and foundational model layers. Competing in these areas would require one to build advanced technology and make massive investments over a period, as we explained in our recent look at why India’s deeptech future is so precariously poised.

India’s comparative advantage, as has been made clear over the past two years, lies in the developer tools and applications layer. These segments align well with the depth of talent in India, and is linked to its SaaS and IT services prowess.

There are of course exceptions such as Sarvam and Bhavish Aggarwal’s Krutrim, which are taking their own routes to build large and small models for Indian languages.

Even as India increases its focus on artificial intelligence (AI), the country accounts for just 3% of early stage AI infrastructure and foundational model startups. AI application-focused startups in India are on the rise, capturing a significant 65% market share, as per Sense AI’s Annual Ventures 2025 report.

The report claims AI tooling accounts for 22% of all funded AI startups in India. “India has a large pool of software developers with AI knowledge. Instead of building infrastructure, they rely on global providers like Google, Microsoft, and AWS, which offer free cloud access to startups in their early years,” said Rahul Agarwalla, founding partner, SenseAI Ventures.

DeepSeek Vs Early Stage Indian AI Startups 

Since the entry of DeepSeek in 2025, the race for faster AI adoption and implementation has intensified. As a result, companies have been focusing heavily on LLM and model investments.

DeepSeek was built at a cost of around $5.5 Mn as compared to OpenAI, which invested over $100 Mn to develop its large language model GPT-4. This has fuelled a new race for a sovereign model for India and Indic language models.

“The number of startups attempting to build LLMs in India is growing, especially after DeepSeek’s success. The recognition that building an LLM doesn’t require a billion dollars has fueled this. Instead, startups can do it with much less capital, like around $6 Mn. The number of companies in this space is expected to rise, potentially reaching 20, 30, or even 40 by the end of the year,” said Agarwalla.

From the geopolitical shifts of the past six months, DeepSeek’s disruption to Silicon Valley’s AI giants and the very real possibility of a fragmented AI world, many in the Indian startup ecosystem are finally speaking about sovereign models for the first time.

“If you had asked me just six months ago, I would have said there’s no need for India to develop a ‘sovereign model’, but a lot has changed in that time. And now I definitely believe that we need to take control of our destiny,” Ashwin Raguraman, founder of Bharat Innovation Fund, said.

Now Vertical LLMs and models are becoming all the rage, and the hope is that these could one day become part of a full-blown India-made LLM. Investors are today seeking founders that are capable of  creating solutions for industries like manufacturing, healthcare and legal with data as a moat.

These sectors are seen to be traditionally slower in adopting SaaS or other technology, but the hope or rather the vision is that AI can bring in measurable outcomes in terms of cost efficiencies to be worth the investment. Investors believe domain-specific agentic AI models for finance, hospitality, banking hold greater potential for monetisation and value creation.

The Talent Piece Of The Puzzle

The AI adoption wave is also influencing how investors assess the founding team, another key facet of the early stage investor evaluation kit. Technical depth is perhaps the most essential and paramount element, which bodes well for founders with the right engineering and AI experience, but also complicates the view for investors who are backing a second-time founder now building in AI after starting up again.

In these cases, investors are looking for domain fluency and the ability to iterate quickly when it comes to products or changes in the AI landscape. The ability of the AI product to extract the best ROI for its eventual customer and figuring this out is often critical besides the technical depth.

Inflection Point Ventures’ Ankur Mittal revealed, “We seek entrepreneurs who have not only created AI models but also effectively implemented them at scale, resolving real-world challenges with demonstrable results.”

Besides this, a good AI team consists of more than simply technical talent; it also comprises deep domain knowledge, research-backed capabilities and an understanding of unit economics fundamentals. It’s also about having a team that can sustain itself when there are major disruptions such as DeepSeek.

Building For The Long Haul

If we read between the lines of what VCs and investors are saying about early stage AI deals, the message is that one needs to think long term and not just eye short term gains within the AI hype cycle.

The rapid rise in investments into Indian AI applications and developer tools could become a big burden on the investor ecosystem if these products do not mature in the long run or deliver the results. Ultimately, there is a risk of some of these investments turning out to be FOMO-induced duds.

But at the same time, there is a growing realisation that India can also go beyond this stage. The fight for space in the global AI race has forced Indian companies on the LLM trail and behind foundational models. This will have deeper ramifications on AI applications as well without a formidable product vision and heavy reliance on ARR projections.

As we also noted in our recent deep dive into India’s deeptech woes, VCs and investors are sharpening their knives and looking for genuine technical depth in companies, defensible data moats, and teams that can actually navigate the breakneck speed of AI.