Till now, AI startups have been judged by how quickly they could ship impressive demos, whether it’s a working chatbot, generated videos, or an automated workflow, as part of their pitch deck. But as capital tightens and enterprises become more cautious, investors say the bar is moving decisively and fast.
According to a Tracxn report, Indian tech startups raised $10.5 billion in 2025, a 17% decrease from $12.7 billion in 2024 and slightly below $11 billion in 2023. Late-stage funding also dropped to $5.5 billion, reflecting fewer large investments. However, early-stage funding increased by 7% year-on-year to $3.9 billion, indicating investor willingness to support proofs-of-concept as the AI ecosystem continues to experiment.
For Stellaris Venture Partners, a Bengaluru-based venture capital firm investing in early-stage startups, the AI advantage today has far less to do with access to large language models and far more with whether founders can survive the brutal journey from demo to production.
“We’ve always had demos that look good,” saysVardhan Dharnidharka, a principal at Stellaris and a former engineer who built AI systems at scale, in an interview with AIM. “The hard part has never been getting to 80% or 85% quality. The real work is getting to 95% or 99%, where real customers actually trust the system.”
From Abundance to Accountability
The shift comes against a broader reset in private capital following the post-COVID-19 boom. Between 2020 and 2022, abundant capital rewarded speed, storytelling, and rapid experimentation.
However, AI-native funding in India saw a selective recovery in 2025, rising to $643.5 million by mid-December, up from $619.5 million in 2024, according to Tracxn. This figure remains below the peak of $1.1 billion raised in 2021.
But as funding slowed and investors insisted on valuation corrections, firms like Stellaris say their core decision-making philosophy remains intact, even as the advice they give founders has changed.
Early-stage investing is inherently cyclical, explains Naman Lahoty, partner at Stellaris, who also runs a startup navigator programme. Funds are built with an eight- to ten-year horizon, spanning multiple bull and bear markets, he said. What changes is not the bar for investing, but the operating guidance offered to founders.
“In times of capital abundance, our role is often to slow founders down,” Lahoty said. “In tougher times, it’s about helping them stay ambitious without losing discipline. The strategy stays consistent, but the balance shifts.”
That balance has become particularly delicate in AI, where the ease of building prototypes has dramatically lowered the cost of entry.
Why AI Advantage is No Longer Obvious
One of the most striking ideas to emerge from the exclusive interview is that the AI advantage itself has become a fuzzy concept. Models, APIs, and tooling are increasingly commoditised. As Dharnidharka puts it, “The clay is available to everyone.”
What separates enduring companies, he argues, is not whether they use AI, but how deeply they engage with the messy, unglamorous details of real-world use.
“There’s a lot of art involved, not just science, ” he remarks. “Two companies can use the same underlying model, but one goes much deeper into nuance, edge cases, user behaviour, context—and that’s where differentiation shows up.”
Rather than asking whether a startup is “AI-first,” Stellaris increasingly looks for aptitude: how quickly founders can learn, adapt, and rethink their technology choices as the ecosystem evolves.
In an environment where what’s possible today may have been unimaginable six months ago, rigidity is a red flag.
Both Lahoty and Dharnidharka emphasise that defensibility in AI is now more likely to emerge from workflows, integration, and execution quality than from proprietary algorithms alone.
One recurring theme was workflow ownership. AI products that embed themselves deeply into how users actually work, rather than sitting as optional overlays, are harder to displace.
“When you save someone 30–40 hours of work and become part of their daily flow, that’s a moat,” Dharnidharka notes. “Especially in underserved or overlooked roles.”
Beyond workflows, product taste and customer obsession are underappreciated advantages. As the gap between first-best and fifth-best technology narrows, execution discipline and outcome quality matter more. “In the long run, what separates winners is whether they’re solving a deep enough problem and executing relentlessly,” Lahoty notes.
He was blunt about the root cause. Venture capital is designed for extreme outcomes. Out of a portfolio of 30 companies, Stellaris expects only a handful to generate returns.
“That doesn’t mean other businesses are bad,” he adds. “They may be profitable, meaningful companies. But they may not fit a high-risk, high-reward portfolio.”
This structural mismatch between what founders want to build and what venture funds must optimise for often fuels resentment. Stellaris helps early founders understand investor strategy through initiatives like Startup Navigator.
The goal, Lahoty notes, is not to push everyone toward VC funding, but to help founders make informed choices about what kind of business they actually want to build.
Discipline of Detachment
One of the less-discussed tensions in early-stage investing is the emotional pull that founders and investors can feel toward an idea, even when the market signals are weak. Both Lahoty and Dharnidharka acknowledge that this is where judgement is tested most.
For founders, emotional attachment is almost unavoidable. Building a company requires deep belief, persistence, and optimism, often in the face of repeated rejection. But for investors, that same emotional commitment can become a liability.
“As an investor, you can’t afford to be too emotional about a problem statement or a direction,” Lahoty observes. “You may really like the idea, but if you’re not able to take objective calls, you risk compounding mistakes.”
That discipline becomes especially important in early-stage companies, where product-market fit remains uncertain. Stellaris, which primarily invests at seed and Series A stages, often backs companies before validation is complete. This means investors must regularly reassess whether a startup is converging toward a scalable opportunity, or whether it is time to pivot, slow down, or even stop.
“The hardest decisions usually come one or two years in, when you’ve tried multiple iterations and still aren’t seeing product-market fit. That’s when you have to ask whether to change course meaningfully, or whether it’s worth continuing at all,” he adds.
Dharnidharka frames this tension in terms of roles. Founders are builders; investors are capital allocators with fiduciary responsibility.
“We’re fund managers at the end of the day,” he notes. “The founders should get the credit for success, but that also means we have to stay objective, even when we care deeply about the people and the problem.”
The challenge, both argue, is not avoiding emotion altogether, but knowing when to override it. In an ecosystem increasingly shaped by rapid iteration and shrinking feedback loops, that ability to balance conviction with detachment may determine not just which startups survive, but which ideas deserve to be pursued at all.
Leaner Teams, More Startups
AI’s impact on team structure is another area where theory is giving way to practice. Both Lahoty and Dharnidharka agree that startups are becoming leaner, with fewer people doing more work. But this does not necessarily mean fewer startups.
“If anything, you’ll see more startups,” Lahoty forecasts. “Because the cost of experimenting is lower.”
Dharnidharka also notes that while some traditional software roles may shrink, new opportunities are emerging across deep tech, robotics, and applied AI. The shift, he argues, is less about job destruction and more about redistribution.
This mirrors earlier platform transitions, from the internet to mobile, where short-term disruption eventually gave way to new categories for job roles and demand.
However, founders can no longer rely solely on novelty; they must also confront issues like reliability, edge cases, and customer outcomes much earlier in their journey. For investors, this means asking tougher questions sooner. As Dharnidharka notes, “Technology will keep getting cheaper, but judgement about what to build and for whom won’t be as easily commoditised.”
The post Stellaris Has Fallen Out of Love With AI Demos, and Other VCs May Follow Suit appeared first on Analytics India Magazine.
