Eric Vishria didn't want to take the meeting. That was nine years ago, in 2016 - he was still a young partner at Benchmark, one of the most prestigious investment firms in Silicon Valley, and the firm traditionally doesn't invest in hardware. He even wrote to his assistant asking why she'd put that meeting on his calendar.

Today, after the listing of Cerebras Systems, those 17.6 million shares Benchmark holds began trading at an opening price of 185 dollars - and finished the same day above 300 dollars. Which means Benchmark's stake, worth 3.3 billion dollars at the open, climbed past 5.3 billion dollars by the close. On an investment of around 270 million dollars. That's nearly 20 times the original stake on a big slice of it.

Cerebras makes something very specific - AI chips the size of an entire silicon wafer, instead of small processor units. That's a completely different approach from Nvidia's. When founders Andrew Feldman (CEO) and Sean Lie (CTO) pitched the idea to Vishria in 2016, they argued that graphics processors (GPUs) aren't perfect for deep learning, even if they're better than ordinary CPUs. "Oh dear. Why would a graphics processor be the real tool for AI?" Vishria asked himself after the third slide.

The road wasn't quick. Eight and a half years of engineering, new cooling methods, machines that hold more bolts at once without damaging fragile silicon. A 500 million dollar raise while the product was still being developed. Another round in the catastrophic 2022. A first IPO attempt in 2024 fell through over an American national security audit on investments from G42 - at the time the main client.

The irony is that the delay is what helped the company. OpenAI and AWS became key clients. Revenue doubled. Cerebras became profitable. And now Vishria, the same man who didn't want to take the meeting, sits on a board where 9.5 percent of an AI chip company is worth over five billion dollars. "Persistence, inventiveness, and adaptiveness too," he says today about the team.

For the European startup scene, the story carries one question. How many good ideas in the past nine years weren't financed because conventional investors decided that hardware is "too heavy"? And how many Cerebras-es are sitting in garages and university labs from Zagreb to Sofia, waiting for a partner who'll come to the meeting rather than one who'll fire off a note to the assistant?