The end of Venture Capital? (VC Roundtable) | E2285

What It Covers

Three venture capitalists — Michael Eisenberg (Aleph VC), Mike Granoff (Muneeb), and Larry Covert (Oxcart Ventures) — examine how five US firms capturing 73.1% of Q1 LP commitments signals either the end of venture capital as a craft business or a cyclical concentration wave, while debating AI compute costs, autonomous vehicles, defense tech, and global supply chain realignment.

Key Questions Answered

• •VC Concentration Math: Five US firms captured 73.1% of all LP commitments in Q1, up from 12 firms capturing 75% previously. Smaller funds can survive by maintaining double the ownership stakes of larger firms. Ownership percentage is the critical variable — dilution from oversized cap stacks means IPO returns fail to cover losses, breaking the asymmetric upside model that makes venture capital function mathematically.
• •Two-Tier VC Strategy: The industry is splitting into Consensus VC (large firms deploying index-like capital) and Traditional VC (sub-$500M funds taking concentrated, high-conviction bets on non-obvious founders). Smaller funds should target LP bases outside major financial centers — Oxcart raised from large Southern family offices with zero prior venture exposure by traveling the "Southern smile" from Southern California to South Carolina.
• •AI Gross Margin Risk: ARR figures from AI companies obscure negative or thin gross margins because token and compute costs are not software margins. Cursor reportedly ran negative 23% gross margins during rapid growth. Founders and investors should model compute costs as a rising variable tied to energy prices, not a declining one, since grid infrastructure scales far slower than semiconductor demand.
• •Edge AI vs. Data Center: Models trained on real-world data outperform simulation-trained equivalents at a fraction of the size — Nexar's Badass 2.0 model beats NVIDIA's Cosmos by 2.5-3x while being an order of magnitude smaller. Founders building AI products should evaluate whether their use case can run on edge CPUs (Apple M-series, Snapdragon at 50-80 TOPS on 15 watts) rather than defaulting to expensive cloud GPU inference.
• •Defense Tech Valuation Caution: The drone sector now has over 2,000 companies, mirroring the 200 bomb-resistant vehicle firms that emerged during Iraq/Afghanistan — most were absorbed or failed while legacy primes like BAE and Oshkosh waited and acquired best-in-class technology. Investors entering defense tech should prioritize non-obvious infrastructure plays adjacent to national security rather than competing in crowded drone or kinetic hardware categories at inflated valuations.
• •Allied Supply Chain Realignment: Global supply chains are permanently restructuring around sovereign allied networks rather than cost-optimized globalist ones. The UAE is building a 5-gigawatt data center accessible to US companies and NVIDIA chips. Founders in hardware, defense, and materials should map their supply chains now against allied-nation sourcing, as adversarial dependencies in semiconductors, rare earths, and critical components face increasing regulatory and geopolitical disruption over the next decade.