OpenAI and Anthropic Bets Test Venture LPs as Fundraising Accelerates
Venture LPs embrace concentrated OpenAI and Anthropic stakes, but faster fundraising, paper gains and thin distributions raise a yellow light for 2026.
Summary
As of October 31, 2026, limited partners Beezer Clarkson of Zurich based LGT Capital Partners and Amit Bhatti of Chapel Hill venture specialist TrueBridge see AI venture as lucrative, not frightening, but flag a yellow light. OpenAI and Anthropic now recur across funds, overriding conflict avoidance conventions and concentrating exposure unusually early. They view power law consolidation as normal, like mature portfolios dominated by Databricks, SpaceX and Stripe, while early stage funds remain more diverse. Fundraising cycles fell from four years to three, two, 18 months, and sometimes one year, versus norms of two to two and a half years or 24 to 36 months. Some funds now return in 12 to 18 months or enlarge midcycle as giant rounds absorb 20% of a fund and seed rounds nearly double, although companies reaching $100 million in revenue distinguish this cycle from zero revenue crypto bets.
LPs often halve planned $10 million commitments to $5 million rather than abandon managers and lose access, but faster capital calls without distributions strain endowments, hospitals and foundations. Extreme 2021 vintages remain immature at five years. Liquidity improved after two weak years through SpaceX, Cerebras's 2026 exit, IPOs, M&A above $10 billion, a cited $60 billion Cursor deal and secondaries. Selling 20% of a winner can return a fund. Emerging managers still raise $5 million to $10 million, typical first funds of $25 million to $50 million, or spinouts at hundreds of millions. Half of LP capital may go to large funds, yet smaller vehicles can total $1 billion. TrueBridge backed 22 and 23 year old GPs whose MIT, Harvard and Stanford accelerator found Cursor and Mercor. LGT usually commits at least $5 million and targets a 20% maximum; TrueBridge writes $3 million to $5 million, sometimes half or all, before dilution toward 20%.
Positives
- Startups reaching $100 million in revenue give accelerated AI investing stronger fundamentals than the previous cycle's zero revenue crypto bets.
- SpaceX, Cerebras, IPOs, M&A and secondaries are restoring liquidity after two slower years.
- Selling 20% of a leading company can return an entire venture fund while preserving most of its upside.
- First funds of $25 million to $50 million remain viable even if large firms capture half of committed LP capital.
- TrueBridge backed 22 and 23 year old GPs whose university accelerator identified Cursor and Mercor early.
Risks & concerns
- Fundraising has compressed from four years to as little as one year, making sound deployment across vintages harder.
- OpenAI and Anthropic exposure now overlaps across many funds, creating unusually rapid concentration and weakening traditional diversification.
- Giant rounds can consume 20% of a fund, while seed rounds nearly doubling force managers to write much larger checks.
- Capital calls without matching distributions can strain the operating commitments of endowments, hospitals and foundations.
- The high valuation 2021 vintages remain unproven after five years, reinforcing concerns about one-year deployment cycles.
- Early employee tenders may weaken retention, particularly where powerful founders can override investor objections.