The rare investor who understands that $2M creates more value than $20M for AI companies.

The rare investor who understands that $2M creates more value than $20M for AI companies.

I spoke with Yash Patel (Capra Ventures), who's an active investor in Lean AI companies like Higgsfield. While every major VC firm waits 12-18 months for retention data, Yash has an entirely different approach for Lean AI companies. He knows the timeline has compressed. One month of lean AI data now equals 6-12 months of traditional SaaS data. So, Yash started evaluating companies like consumer products: Day 1, 7, 30 retention, and DAU/MAU ratios (50%+ is a yes). But he could not use the traditional VC fund structures as they are misaligned with lean AI companies

The typical VC model is a $400M~ fund investing in 50~ companies, writing $5-10M checks with 2-3x capital reserved for follow-ons for winners. It's built for capital-intensive SaaS companies that need to burn cash to grow. But lean AI companies don't need $5M at seed. They need $500K to $2M. This creates misaligned incentives where founders are forced to raise more than needed, and VCs can't generate returns from small checks. So Yash structured his fund differently: $100M fund size, 20-25 companies, $500K-$2M initial checks at seed, and $5-$15M at series A for inflecting AI companies, with just a 1-1.2x reserve ratio. He targets 15-20% ownership at seed, but is flexible at Series A. This works because he's not forcing a model designed for capital-intensive SaaS onto capital-efficient AI companies.

To separate billion-dollar outcomes from the rest, Yash developed a 3-moat framework:

1. Proprietary data feedback loops: As users interact with your product, you gather data that continuously improves the experience, creating a flywheel hard to replicate
2. Distribution edges: Unique GTM strategies that give you access to specific customer segments through network effects and community building
3. Product velocity: When you have the above two, you can iterate faster than anyone else, creating a compounding flywheel

If a company has all three, it's an automatic investment. Higgsfield had all three moats, yet most VCs hesitated and asked for more data over a longer time period. Yash invested early while others waited, and by the time the round closed, Higgsfield’s revenue doubled. It went from $0 to $50M+ in 5 months, and this is why waiting for validation means missing the opportunity entirely.

Whether you are a founder or an investor in this space, Yash shared what actually works:

• Founders should seek flexibility in investors who write $500K-$2M checks

• VCs should adjust GP/LP mandates to allow smaller check sizes and lower reserve ratios at seed. When companies inflect at series A, VCs should move fast before valuation grows exponentially

There's a lot more to his story than I could fit here.