The Meeting That Almost Didn't Happen
Silicon Valley's Benchmark is one of the most celebrated venture capital firms in the world — an early backer of eBay, Twitter, Uber, and Snap. But there's one category the firm has historically steered clear of: hardware startups.
So when Cerebras Systems came knocking roughly a decade ago, Benchmark partner Eric Vishria did what most disciplined investors would do. He dragged his feet.
The pitch wasn't easy to ignore for long, though. Cerebras was building something audacious: a computer chip purpose-built for artificial intelligence workloads, designed from scratch to tackle the limitations that were holding back neural network training. It was a big, expensive, capital-intensive bet — exactly the kind of play Benchmark's lean software-focused thesis was built to avoid.
Vishria eventually took the meeting. He liked what he heard enough to break from the firm's usual playbook.
A Decade Later: Billions on the Table
That decision is now paying off in a very big way. Cerebras Systems has gone public, and Benchmark's early stake has grown into a multi-billion-dollar position — a return that ranks among the firm's most significant wins in years.
For the broader venture capital world, the Cerebras IPO is being watched closely. AI infrastructure investments — chips, data centres, networking — have exploded in interest since the ChatGPT era began in late 2022, but many VCs still view hardware as too capital-hungry and too risky compared to software. Cerebras is one of the clearest proof points yet that hardware bets can generate software-scale returns when the timing is right.
What Cerebras Actually Does
Founded in 2016, Cerebras built the world's largest computer chip — the Wafer Scale Engine — which packs an extraordinary number of processing cores onto a single massive die. The design allows AI models to train dramatically faster than on conventional GPU clusters, with lower energy use per computation.
The company's timing aligned almost perfectly with the generative AI wave. As tech giants and startups alike scrambled for AI compute capacity, Cerebras found itself with a differentiated product at exactly the moment demand was surging.
The Lesson for Investors
Vishria's near-miss is a useful reminder that some of the most valuable investment decisions are the ones made in spite of a firm's own rules. Benchmark's hardware hesitation wasn't irrational — most hardware startups do fail, and the capital requirements are brutal compared to SaaS. But Cerebras had a thesis specific enough, a team credible enough, and a market tailwind powerful enough to justify the exception.
For the VC industry, the story will likely be told for years as a case study in when to override pattern-matching with first-principles thinking.
What's Next for Cerebras
With a public market listing secured, Cerebras now faces the scrutiny that comes with life as a public company. Competition in the AI chip space is intensifying — Nvidia remains the dominant force, and AMD, Intel, and a wave of AI chip startups are all fighting for a slice of the market. Cerebras will need to convert its IPO momentum into sustained revenue growth to justify whatever valuation the market assigns.
For Benchmark and Vishria, though, the hard part is already done. They took the meeting — eventually — and it changed everything.
Source: TechCrunch
