Benchmark Raises $2 Billion but Holds Its Slow, San Francisco First Investing Pace
Benchmark keeps its deal pace low after raising $2 billion, betting on concentrated AI winners, founder loyalty and San Francisco’s enduring startup edge.
Summary
Benchmark announced $2 billion across two funds in summer 2026, only its second departure from smaller vehicles in 31 years. Peter Fenton said the core fund, held at $500 million for 20 years, expanded for inflation, while the $1.25 billion growth fund may make four to eight bets, with one or two generating 90% of returns. Equal partners Fenton, Chethan Puttagunta, Eric Vishria, Ev Randall and Jack Altman still target two or three new investments each annually, hold agenda free Monday meetings without formal veto rules, and treat core founder relationships as 10 to 15 year commitments. Randall joined in late 2025; Altman joined in February 2026 after raising a second $275 million fund in one week.
Altman has made disclosed and undisclosed investments across chips, robotics and infrastructure. Benchmark backed Cerebras in spring 2016, ending a 10 year hardware hiatus; its 2026 IPO raised $6 billion, priced at $185, climbed to $350 and returned near $185. Vishria expects capital hungry AI companies to favor IPOs over direct listings. Benchmark preserves sector exclusivity, although near zero coding costs increasingly push startups into conflict, as Factory and Cognition demonstrated.
Randall said Character AI’s Chapter 11 overstates AI weakness because Google acquihired its founders and team 18 months earlier following a $2.3 billion sale. Portfolio demand exceeds available compute at healthy gross margins, but CoreWeave and Nebius valuations depend heavily on estimated OpenAI and Anthropic revenue. Randall expects both private companies could list within 12 months, potentially reducing 5% to 10% market swings around anonymous estimates. Benchmark strongly favors San Francisco, where more than 90% of its historical capital returns originated; Shopify and Ramp are notable exceptions.
Positives
- The $1.25 billion growth fund lets Benchmark pursue opportunities previously excluded by its early stage mandate.
- Cerebras raised $6 billion in its 2026 IPO after Benchmark backed its first round in spring 2016.
- Portfolio companies face more customer demand than available compute while maintaining healthy gross margins.
- OpenAI and Anthropic listings within 12 months could replace anonymous estimates with regular public disclosures.
- More than 90% of Benchmark’s historical capital returns came from companies founded in San Francisco.
Risks & concerns
- One or two investments may generate 90% of the growth fund’s returns, creating substantial concentration risk.
- Cerebras fell from $350 back near its $185 IPO price, illustrating extreme public market volatility.
- CoreWeave and Nebius valuations depend heavily on estimated revenue at private companies OpenAI and Anthropic.
- Near zero coding costs are pushing ambitious AI startups into overlapping markets and creating investor conflicts.
- AI companies need substantial capital, making direct listings less practical and increasing dependence on conventional IPO financing.
- Portfolio companies cannot secure enough compute to satisfy paying customer demand.