The firms that wrote checks to OpenAI are raising billions while everyone else is explaining why Peloton seemed like a good idea.
The Summary
- Felix Capital is stuck $150 million short of its $600 million fundraising target, with LPs demanding returns from older funds before committing new capital
- AI is creating a two-tier VC market: funds with successful AI portfolio companies are raising at record speed, while everyone else is getting ghosted
- The split isn't about fund size or vintage—it's about whether you own equity in the companies building the agent economy
The Signal
Venture capital is experiencing its sharpest divide in decades. Felix Capital, despite backing household names like Peloton and Deliveroo, is struggling to close its fundraise. The firm set out to raise $600 million early last year but remains $150 million short. Limited partners are withholding capital until they see actual distributions from older funds.
Meanwhile, firms with AI winners in their portfolios are raising at unprecedented speed. The gap isn't about firm reputation or track record anymore. It's about exposure to the AI infrastructure layer that's rewriting software economics.
"Investors want to see returns from older funds before they commit new capital."
The dynamics playing out are instructive:
- Consumer-focused funds are bleeding out slowly
- Infrastructure and B2B software funds with AI holdings are oversubscribed
- LPs are treating AI exposure like a binary filter, not a portfolio consideration
Bloomberg's coverage featuring Julia Moore from Breakout Ventures, Emily Zhao from Salesforce Ventures, and Daniel Docter from Dell Technologies Capital underscores how corporate venture arms with strategic AI investments are consolidating power. These aren't traditional VC firms competing on returns alone. They're offering portfolio companies distribution, infrastructure access, and enterprise customer relationships that matter more than capital.
The problem for firms like Felix isn't that they made bad bets. Peloton was a pandemic winner. Deliveroo built a real business. But neither company is building the rails for autonomous agents or the compute layer for AI training. In 2026, LPs don't want exposure to companies that used software. They want exposure to companies building the software that builds software.
This creates a vicious cycle. Firms without AI winners can't raise. Without fresh capital, they can't compete for AI deals. Without AI deals, their next fundraise will be even harder. The VC industry is stratifying into those who got in early on foundation models and agent platforms, and everyone else.
The Implication
If you're raising a fund right now without AI portfolio companies, expect a brutal 18 months. LPs have decided AI isn't a sector, it's the only sector that matters. That's probably wrong, but it doesn't matter. Capital follows narrative, and the narrative is set.
For founders, this means something different. The VCs desperate for AI deals will pay stupid prices and offer terrible terms just to get logo exposure. The smart play is raising from firms that already have AI winners and don't need you for fundraising optics. They'll be better partners when the hype cycle turns.