Benchmark Capital, the storied Silicon Valley VC firm known for early investments in eBay, Snap, Uber and Twitter, is breaking one of its signature traditions: keeping its capital at around $425 million and backing only new startups. After more than two decades of limiting its vehicles to that amount or less, the firm closed with $2 billion in commitments to two new funds, including a $1.25 billion vehicle dedicated to later-stage investments, according to the Wall Street Journal.
While the capital sizes of many venture capital firms have grown into billions of dollars over the past decade, Benchmark stuck to the strategy that helped make it legendary. By being consistently selective and taking a large—typically 20%—stake in each startup the firm backs, it has maintained a model designed to maximize big returns for its limited partners.
However, Benchmark’s relatively small capital sizes likely prevented the firm from investing in capital-intensive AI startups, particularly institutional model makers, whose round sizes often run into the hundreds of millions. As a result, the company hasn’t invested in Anthropic, OpenAI, or any of the other capital-intensive AI labs like Periodic Labs, Reflection AI, or Retrograde Superintelligence.
Where Benchmark has placed AI bets, the results have been mixed. The firm led a $75 million round in Manus, a Singapore-based AI agent platform that reached $100 million in annual recurring revenue within eight months of launch. When Meta agreed to acquire Manus for about $2 billion late last year, it looked like another Benchmark winner was in the works. But Chinese regulators – arguing that the company, which was founded in China before moving to Singapore, had violated export control laws – blocked the deal in April, leaving Benchmark’s stake in limbo.
Benchmark’s new $750 million early-stage fund will give the company more check-writing flexibility in an environment where early-stage valuations have soared. While the firm traditionally backs companies at the Series A stage, Benchmark has recently given itself more flexibility to invest in companies at other early stages of growth.
In recent months, Benchmark has backed two Series B startups: Gumloop, a platform that lets businesses build AI agents without writing code, and Monaco, an AI sales and CRM platform.
Benchmark general partner Everett Randle previously told TechCrunch that the company wants to build a “meaningful and deep relationship with entrepreneurs, and that can happen relatively early in the company’s lifecycle, in the creation, [Series] Oh, at [Series] SI.”
The company dipped its toe into late-stage investing when it raised a $225 million special purpose vehicle (SPV) to participate in a $1 billion pre-IPO round for Cerebras, as TechCrunch previously reported. Benchmark led the chipmaker’s Series A in 2016. Cerebras held its IPO last month, returning Benchmark $3.25 billion at the IPO price.
This windfall prompted the company to raise a special growth fund. This new vehicle will make five to six major investments in both existing portfolio companies and new startups, according to a person familiar with Benchmark’s strategy.
The two new funds are not the only changes at Benchmark. Over the past two years, the company has undergone a significant change in its general partners.
In 2024, Miles Grimshaw left the company to rejoin Thrive Capital. Then last year, Sarah Tavel—Benchmark’s first and only female general partner to date—took the less involved role of venture partner, while Victor Lazarte left to found his own VC firm.
To replenish its ranks, Benchmark — which traditionally operates with four to six general partners — added two new high-profile investors to its team: Randle, a poacher from Kleiner Perkins, and Jack Altman, brother of OpenAI CEO Sam Altman. The moves suggest that even Benchmark, long defined by its resistance to growth, now sees the age of artificial intelligence as requiring a different playbook — more capital, more stages and fresh blood at the partner table.
When you purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
