Groq’s early backer: half my AI bets will fail

Sandhya Venkatachalam, founder of early-stage firm Axiom Partners, has built her reputation on backing outliers. An early investor in Groq, the AI chipmaker now valued at over $1 billion, she’s now warning that half of her AI bets will fail—a rare admission in an industry where optimism often outpaces outcomes.

Venkatachalam’s interview with Crunchbase News reveals how Groq’s trajectory reshaped her approach to AI investing. The company, which emerged from stealth with a focus on high-performance inference chips, forced her to reconsider what makes an AI startup durable. It wasn’t just about the technology, she says, but the team’s ability to adapt as the market shifted. That lesson now guides her firm’s thesis: look for founders who can pivot without losing sight of the problem they’re solving.

This perspective arrives as early-stage AI funding faces growing skepticism. While seed rounds for AI startups remain plentiful, the gap between hype and commercial traction is widening. Venkatachalam’s candor about failure rates reflects a broader reckoning among investors who’ve poured capital into AI without clear paths to profitability. Her stance isn’t just defensive—it’s a challenge to the sector. If even a disciplined investor like her expects half her portfolio to collapse, what does that say about the hundreds of AI startups raising pre-seed rounds on little more than a pitch deck?

The timing matters. Europe’s early-stage funding crunch, which we reported last week, is forcing founders to stretch runway further, while India’s spacetech sector tests whether investor appetite extends beyond a few high-profile names. Venkatachalam’s focus on durability suggests AI startups will soon face similar scrutiny. The question isn’t just who can raise the most capital, but who can survive long enough to prove their model works.

Groq’s success looms large in this calculus. The company’s chips, designed for low-latency inference, have found traction in enterprise applications where dominant players don’t hold an unassailable lead. But even Groq’s journey hasn’t been linear—it adapted its approach as the market evolved. Venkatachalam’s takeaway? The best AI startups aren’t just building products; they’re building systems that can outlast the current cycle of hype.

For founders, this is a warning. The days of raising on a demo and a dream may be numbered. Investors like Venkatachalam are demanding more: proof of product-market fit, a clear path to revenue, and a team that can navigate the inevitable pivots. For backers, it’s a signal that the AI gold rush is entering a new phase—one where discipline matters as much as ambition.

What happens next will depend on whether other investors adopt this mindset. If they do, expect fewer seed rounds, longer due diligence, and a sharper focus on unit economics. If they don’t, the AI graveyard could grow faster than anyone expects.

Sources: news.crunchbase.com